Advances and Trends in Artificial Intelligence. From Theory to Practice

Weather forecasting is a logical process that consists in evaluating the predictions provided by a set of stochastic models, compare these and take a conclusion about the weather in a given area and a given interval of time. Meteorological forecasting provides reliable predictions about the weather within a given interval of time. The automation of the forecasting process would be helpful in a number of contexts. For instance, when forecasting about underpopulated or small geographic areas is out of the human forecasters’ tasks but is central, e.g., for tourism. In this paper, we start to deal with these challenging tasks by developing a defeasible reasoner for meteorological forecasting, which we evaluate against a real-world example with applications to tourism and holiday planning.

Most of existing event prediction approaches consider event prediction problems within a specific application domain while event prediction is naturally a cross-disciplinary problem. This paper introduces a generic taxonomy of event prediction approaches. The proposed taxonomy, which oversteps the application domain, enables a better understanding of event prediction problems and allows conceiving and developing advanced and context-independent event prediction techniques.

Hydrologists are often encountered problem of missing values in a rainfall and runoff database. They tend to use the normal ratio or distance power method to deal with the problem of missing data in the rainfall and runoff database. However, this method is time consuming and most of the time, it is less accurate. In this paper, two neighbor-based imputation methods namely K-nearest neighbor (KNN) and Gaussian mixture model based KNN imputation (GMM-KNN) were explored for gap filling the missing rainfall and runoff database. Different percentage of missing data entries were inserted randomly into the database such as 2%, 5%, 10%, 15% and 20% of missing data. Pros and cons of these two methods were compared and discussed. The selected study area is Bedup Basin, located at Samarahan Division, Sarawak, East Malaysia. It is observed that the GMM-KNN imputation method results in the best estimation accuracy for the missing rainfall and runoff database.

We consider the unsolved problem of Distance Estimation (DE) when the inputs are the x and y coordinates (i.e., the latitudinal and longitudinal positions) of the points under consideration, and the elevation/altitudes of the points specified, for example, in terms of their z coordinates (3DDE). The aim of the problem is to yield an accurate value for the real (road) distance between the points specified by all the three coordinates of the cities in question (This is a typical problem encountered in a GISs and GPSs.). In our setting, the distance between any pair of cities is assumed to be computed by merely having access to the coordinates and known inter-city distances of a small subset of the cities, where these are also specified in terms of their 3D coordinates. The 2D variant of the problem has, typically, been tackled by utilizing parametric functions called “Distance Estimation Functions” (DEFs). To solve the 3D problem, we resort to the Adaptive Tertiary Search (ATS) strategy, proposed by Oommen et al., to affect the learning. By utilizing the information provided in the 3D coordinates of the nodes and the true road distances from this subset, we propose a scheme to estimate the inter-nodal distances. In this regard, we use the ATS strategy to calculate the best parameters for the DEF. While “Goodness-of-Fit” (GoF) functions can be used to show that the results are competitive, we show that they are rather not necessary to compute the parameters. Our results demonstrate the power of the scheme, even though we completely move away from the traditional GoF-based paradigm that has been used for four decades. Our results conclude that the 3DDE yields results that are far superior to those obtained by the corresponding 2DDE.

Multi-class classification problem is research challenge in many applications. Listing companies’ statuses are signals on different risk levels in China’s stock markets. The prediction of the listing statuses is complex problem due to imbalance in the data, due to different values and features. In the literature when the list status is divided into two categories for simple measurements using binary classification model, accurate risk management cannot achieved correctly. In this work, we have used SMOTE and wrapper feature selection to reprocess data. Accordingly, we have proposed an algorithm named as Twin-KSVC (twin multi-class support vector machine) which is used for multi-class classification problem by “1-versus-1-versus-rest” structure. Our experiments tested on large sample of data set; show that we could achieve better performance, in comparison with other approach. We have tested our algorithm on different strategies of feature selection for comparison purposes.

Reliable query execution time prediction is a desirable feature for modern databases because it can greatly help ease the database administration work and is the foundation of various database management/automation tools. Most exiting studies on modeling query execution time assume that each individual query is executed as serialized steps. However, with the increasing data volume and the demand for low query latency, large-scale databases have been adopting the massive parallel processing (MPP) architecture. In this paper, we present a novel machine learning based approach for building a robust model to estimate query execution time by considering both query-based statistics and real-time system attributes. The experiment results demonstrate our approach is able to reliably predict query execution time in both idle and noisy environments at random levels of concurrency. In addition, we found that both query and system factors are crucial in making stable predictions.

The pertinent problem of Traveling Time Estimation (TTE) is to estimate the travel time, given a start location and a destination, solely based on the coordinates of the points under consideration. This is typically solved by fitting a function based on a sequence of observations. However, it can be expensive or slow to obtain labeled data or measurements to calibrate the estimation function. Active Learning tries to alleviate this problem by actively selecting samples that minimize the total number of samples needed to do accurate inference. Probabilistic Programming Languages (PPL) give us the opportunities to apply powerful Bayesian inference to model problems that involve uncertainties. In this paper we combine Thompson Sampling with Probabilistic Programming to perform Active Learning in the Travel Time Estimation setting, outperforming traditional active learning methods.

Recently, many research results have indicated that diversity is the most important characteristic of crowd-based applications. However, it is a point estimates-based finding in which single values are used as the representation of individual predictions on a real-life cognition task. This paper presents a study on how cardinality and diversity influence the quality of collective prediction using interval estimates. By means of computational experiments, we have found that these factors positively influence the quality of collective prediction. Besides, the results also indicate that the hypothesis “the higher the diversity, the better the quality of collective prediction” is true. Furthermore, the findings also reveal a cardinality threshold in which its increase does not significantly influence the quality of collective prediction.

Our understanding and ability to effectively monitor and manage coastal ecosystems are severely limited by observation methods. Automatic recognition of species in natural environment is a promising tool which would revolutionize video and image analysis for a wide range of applications in marine ecology. However, classifying fish from images captured by underwater cameras is in general very challenging due to noise and illumination variations in water. Previous classification methods in the literature relies on filtering the images to separate the fish from the background or sharpening the images by removing background noise. This pre-filtering process may negatively impact the classification accuracy. In this work, we propose a Convolutional Neural Network (CNN) using the Squeeze-and-Excitation (SE) architecture for classifying images of fish without pre-filtering. Different from conventional schemes, this scheme is divided into two steps. The first step is to train the fish classifier via a public data set, i.e., Fish4Knowledge, without using image augmentation, named as pre-training. The second step is to train the classifier based on a new data set consisting of species that we are interested in for classification, named as post-training. The weights obtained from pre-training are applied to post-training as a priori. This is also known as transfer learning. Our solution achieves the state-of-the-art accuracy of 99.27% accuracy on the pre-training. The accuracy on the post-training is 83.68%. Experiments on the post-training with image augmentation yields an accuracy of 87.74%, indicating that the solution is viable with a larger data set.

In this paper, we investigate the influence of distance metrics on the results of open-set subject classification of text documents. We utilize the Local Outlier Factor (LOF) algorithm to extend a closed-set classifier (i.e. multilayer perceptron) with an additional class that identifies outliers. The analyzed text documents are represented by averaged word embeddings calculated using the fastText method on training data. Conducting the experiment on two different text corpora we show how the distance metric chosen for LOF (Euclidean or cosine) and a transformation of the feature space (vector representation of documents) both influence the open-set classification results. The general conclusion seems to be that the cosine distance outperforms the Euclidean distance in terms of performance of open-set classification of text documents.

This paper demonstrates how deep learning can be used to find optimal reservoir operating policies in hydropower river systems. The method that we propose is based on the implicit stochastic optimization (ISO) framework, using direct policy search methods combined with deep neural networks (DNN). The findings from a real-world two-reservoir hydropower system in southern Norway suggest that DNNs can learn how to map input (price, inflow, starting reservoir levels) to the optimal production pattern directly. Due to the speed of evaluating the DNN, this approach is from an operational standpoint computationally inexpensive and may potentially address the long-standing problem of high dimensionality in hydropower optimization. Further on, our method may be used as an input for decision-theoretic planning, suggesting the policy that will give the highest expected profit. The approach also permits for a broader use of pre-trained neural networks in historical reanalysis of production patterns and studies of climate change effects.

Head poses are a key component of human bodily communication and thus a decisive element of human-computer interaction. Real-time head pose estimation is crucial in the context of human-robot interaction or driver assistance systems. The most promising approaches for head pose estimation are based on Convolutional Neural Networks (CNNs). However, CNN models are often too complex to achieve real-time performance. To face this challenge, we explore a popular subgroup of CNNs, the Residual Networks (ResNets) and modify them in order to reduce their number of parameters. The ResNets are modified for different image sizes including low-resolution images and combined with a varying number of layers. They are trained on in-the-wild datasets to ensure real-world applicability. As a result, we demonstrate that the performance of the ResNets can be maintained while reducing the number of parameters. The modified ResNets achieve state-of-the-art accuracy and provide fast inference for real-time applicability.

When systems become smarter they have to cope with faults occurring during operation in an intelligent way. For example, an autonomous vehicle has to react appropriately in case of a fault occurring during driving on a highway in order to assure safety for passengers and other humans in its surrounding. Hence, there is a need for fail-operational systems that extend the concept of fail-safety. In this paper, we introduce a method that relies on rules for controlling a system. The rules specify the behavior of the system including behavioral redundancies. In addition, the method provides a runtime execution engine that selects the rules accordingly to reach a certain goal. In addition, we present a language and an implementation of the method and discuss its capabilities using a case study from the mobile robotics domain. In particular, we show how the rule-based fail-operational system can adapt to a fault occurring at runtime.

The monitoring of air quality allows to evaluate the amount of harmful particles for health that are being released. Under this paradigm and knowing the current methods to monitor these parameters, this work proposes the use of a UAV for commercial use and the construction of a card for gas measurement. Additionally and with the objective of having complete control over the vehicle, the article proposes the development of a library for the control and monitoring of the instrumentation of a commercial drone, through which the validation of control algorithms is proposed. As a result of this work, two real experiments on a rural environment and an urban environment are carried out to validate both the library created and the method of acquiring information on air quality.

Capability Oriented Requirements Engineering (CORE) is an emerging research area where designers are faced with the challenge of analyzing changes in the business domain, capturing user requirements, and developing adequate IT solutions taking into consideration these changes and answering user needs. CORE aims at providing continuously a certain level of quality (business, service, security, etc.) in dynamically changing circumstances such as smart city operations. Intelligent management of transportation is a complex smart city operation, and hence, an optimal application domain for CORE. A specific application within the domain of intelligent management of transportation is the smart management of parking. In this paper, we propose dealing with the intelligent management of parking spaces by using CORE. In an I-Parking system that offers personalized advice to users, we deal with change and introduce smartness by means of “Goal Models”, “Informational Models”, “Capability Models” and “Actor Dependency Models”.

The article deals with the combined motion control system which provides an autonomous movement of the robot in an uncertain environment. The motion planning level is implemented on a cascade neural network of deep learning. The proposed structure of the network allows decomposing the task of planning a path to the task of deciding whether to maneuver and the task of selecting a direction to bypass an obstacle. The motion control level is implemented in the form of a hybrid system that includes the neural network correction of the path, and the algorithm for avoiding collisions, built on the basis of unstable modes. The control system was modeled and as the result of modeling the quality of control system was obtained. The results of experiments confirming the performance of the control system are presented. It is proposed to classify the environment of operation of the robot according to the complexity of the current situation, depending on the need for maneuver. The environment is classified into complexity classes, the number of which depends on the number of active network cascades.

Upcoming space missions are requiring a higher degree of on-board autonomy operations to increase quality science return, to minimize close-loop space-ground decision making, and to enable new scenarios. Artificial Intelligence technologies like Machine Learning and Automated Planning are becoming more and more popular as they can support data analytics conducted directly on-board as input for the on-board decision making system that generates plans or updates them while being executed. This paper describes the planning and execution architecture under development at the European Space Agency to target this need of autonomy for the ops-sat mission to be launched in 2019.

This research deals with the coverage path problem (CPP) in a given area with several known obstacles for agriculture Unmanned Aerial Vehicles (UAVs). The work takes the geometry characteristics of the field and obstacles into consideration. A practical method of the coverage path planning process is established. An obstacle avoidance path planning is used to find a coverage path for agriculture UAVs. The method has been tested with an Android application and is already applied in reality. The results turn out that the method is complete for this kind of coverage path planning problem.

We consider the case where a knowledge base consists of interactions among parameter values in an input parameter model for web application security testing. The input model gives rise to attack strings to be used for exploiting XSS vulnerabilities, a critical threat towards the security of web applications. Testing results are then annotated with a vulnerability triggering or non-triggering classification, and such security knowledge findings are added back to the knowledge base, making the resulting attack capabilities superior for newly requested input models. We present our approach as an iterative process that evolves an input model for security testing. Empirical evaluation on six real-world web application shows that the process effectively evolves a knowledge base for XSS vulnerability detection, achieving on average 78.8% accuracy.

Mining frequent itemsets (abbr. FIs) from dense databases usually generates a large amount of itemsets, causing the mining algorithms to suffer from long execution time and high memory usage. Frequent closed itemset (abbr. FCI) is a lossless condensed representation of FI. Mining only the FCIs allows to reducing the execution time and memory usage. Moreover, with correct methods, the complete information of FIs can be derived from FCIs. Although many studies have presented various efficient approaches for mining FCIs, few of them have developed efficient algorithms for deriving FIs from FCIs. In view of this, we propose a novel algorithm called DFI-Growth for efficiently deriving FIs from FCIs. Moreover, we propose two strategies, named maximum support selection and maximum support replacement to guarantee that all the FIs and their supports can be correctly derived by DFI-Growth. To the best of our knowledge, the proposed DFI-Growth is the first kind of tree-based and pattern growth algorithm for deriving FIs from FCIs. Experiments show that DFI-Growth is superior to the most advanced deriving algorithm [12] in terms of both execution time and memory consumption.

Periodic-frequent patterns are sets of items (values) that periodically appear in a sequence of transactions. The periodicity of a pattern is measured by counting the number of times that its periods (the interval between two successive occurrences of the patterns) are greater than a user-defined maxPer threshold. However, an important limitation of this model is that it can find many patterns having a periodicity that vary widely due to the strict maxPer constraint. But finding stable patterns is desirable for many applications as they are more predictable than unstable patterns. This paper addresses this limitation by proposing to discover a novel type of periodic-frequent patterns in transactional databases, called Stable Periodic-frequent Pattern (SPP), which are patterns having a stable periodicity, and a pattern-growth algorithm named SPP-growth to discover all SPP. An experimental evaluation on four datasets shows that SPP-growth is efficient and can find insightful patterns that are not found by traditional algorithms.

The main objective of our work is to assess the security of a given real world system by verifying whether this system satisfies given properties and, if not, how far it is from satisfying them. We are interested in performing formal verification of this system based on event sequences collected from its execution. In this paper, we propose a preliminary model-based approach where a Graphical Event Model (GEM), learned from the event streams, is considered to be representative of the underlying system. This model is then used to check a certain security property. If the property is not verified, we also propose a search methodology to find another close model that satisfies it. Our approach is generic with respect to the verification procedure and the notion of distance between models. For the sake of completeness, we propose a distance measure between GEMs that allows to give an insight on how far our real system is from verifying the given property. The interest of this approach is illustrated with a toy example.

Emotions have an impact to almost all decisions. They affect our choices and are activated as feedback during the decision process. This work aims at investigating whether behavior patterns can be learned and used to predict the user’s choice. Specifically, we focused on pairwise image comparisons in a preference elicitation experiment, and exploited a Process Mining approach to learn preferences. We proposed and evaluated a strategy based on experienced emotions and gaze behaviour, whose results show promising prediction performance.

The number of Foreign Independent Tour (FIT) is increasing in the world. This research aims to develop a personal adaptive tourism recommendation system (PATRS) for FIT. This paper describes the concept of PATRS and related researches. In order to develop the PATRS, an easy feature extraction method from a tourism website is required. The classification of photos of tourism spots is an important technology to realize the feature extraction from numerous information in the website. This paper proposes a classification method of photos in a major tourism website by color analysis. From the results on the experiments, we confirmed that the photos in a tourism website can be classified into four classes by the proposed method.

In this paper, we focus on improving the customer’s flow by harnessing the power of analytics and focusing on the arrival process of passengers at one of the busiest airports in Asia. As there is a recent growth in travelers, the airport is undergoing expansion and is thus under tremendous pressure to utilise its resources effectively and efficiently. We first leverage the historical data of the arrival flights, passenger load, and on-time performance flag indicator in order to predict the arriving passenger’ load for the immigration counters and taxi queues. We then build a decision support system using simulation to estimate the optimal number of immigration counter requirements so as to minimize the waiting time at the queues. This is also done to predict the number of taxis required to meet the service level agreement and to ensure the seamless flow of customers at various touch points to improve customer’ satisfaction. The tool developed has benefited the manager in his daily operations, and advanced his decision making process supported by data rather than personal experience or “gut” feeling.

Constraint-based recommender systems help users to identify useful objects and services based on a given set of constraints. These decision support systems are often applied in complex domains where millions of possible recommendations exist. One major challenge of constraint-based recommenders is the identification of recommendations which are similar to the user’s requirements. Especially, in cases where the user requirements are inconsistent with the underlying constraint set, constraint-based recommender systems have to identify and apply the most suitable diagnosis in order to identify a recommendation and to increase the user’s satisfaction with the recommendation. Given this motivation, we developed two different approaches which provide similar recommendations to users based on their requirements even when the user’s preferences are inconsistent with the underlying constraint set. We tested our approaches with two real-world datasets and evaluated them with respect to the runtime performance and the degree of similarity between the original requirements and the identified recommendation. The results of our evaluation show that both approaches are able to identify recommendations of similar solutions in a highly efficient manner.

Customer satisfaction represents a crucial goal for every seller. In e-commerce, it is possible to increase this factor by a better understanding of customers purchasing behavior based on collected historical data. In a period of a continually growing amount of data, it is not an easy task to effectively pre-process and analyses. Our motivation was to understand the buying behavior of the on-line e-shop customer through appropriate analytical methods. The result is a knowledge set that retailers could use to deliver products to specific customers, to meet their expectations, and to increase his revenues and reputation. For recommendations generation, we used a collaborative filtering method and matrix factorization associated with Singular Value Decomposition (SVD) algorithm. For segmentation, we selected the K-Means algorithm and the RFM method. All methods produced interesting and potentially useful results that will be evaluated and deployed into practice.

For any corporation the interaction with its customers is an important business process. This is especially the case for resolving various business-related issues that customers encounter. Classifying the type of such customer service e-mails to provide improved customer service is thus important. The classification of e-mails makes it possible to direct them to the most suitable handler within customer service. We have investigated the following two aspects of customer e-mail classification within a large Swedish corporation. First, whether a multi-label classifier can be introduced that performs similarly to an already existing multi-class classifier. Second, whether conformal prediction can be used to quantify the certainty of the predictions without loss in classification performance. Experiments were used to investigate these aspects using several evaluation metrics. The results show that for most evaluation metrics, there is no significant difference between multi-class and multi-label classifiers, except for Hamming loss where the multi-label approach performed with a lower loss. Further, the use of conformal prediction did not introduce any significant difference in classification performance for neither the multi-class nor the multi-label approach. As such, the results indicate that conformal prediction is a useful addition that quantifies the certainty of predictions without negative effects on the classification performance, which in turn allows detection of statistically significant predictions.

Offline knowledge compilation enables an online diagnosis process that can manage in a linear time any sequence of observables. In a posteriori diagnosis, this sequence, called a symptom, is the input, and the corresponding collection of sets of faults, each set being a candidate, is the output. Since the compilation is computationally hard, we propose to compile only the knowledge chunks that are relevant to some phenomena of interest, each described as a scenario. If, on the one hand, a partial knowledge compilation does not ensure the completeness of the resulting collection of candidates, on the other, it allows attention to be focused on the most important of them. Moreover, the compiled structure, called symptom dictionary, can incrementally be extended over time.

Online fraud poses a relatively new threat to the revenues of companies. A way to detect and prevent fraudulent behavior is with the use of specific machine learning (ML) techniques. These anomaly detection techniques have been thoroughly studied, but the level of employment is not as high. The airline industry suffers from fraud by parties such as online travel agencies (OTAs). These agencies are commissioned by an airline carrier to sell its travel tickets. Through policy violations, they can illegitimately claim some of the airline’s revenue by offering cheaper fares to customers.This research applies several anomaly detection techniques to detect fraudulent behavior by OTAs and assesses their strengths and weaknesses. Since the data is not labeled, it is not known whether fraud has actually occurred. Therefore, unsupervised ML is used. The contributions of this paper are, firstly, to show how to shape the online booking data and how to engineer new and relevant features. Secondly, this research includes a case study in which domain experts evaluate the detection performance of the considered ML methods by classifying a set of 75 bookings. According to the experts’ analysis, the techniques are able to discover previously unknown fraudulent bookings, which will not have been found otherwise. This demonstrates that anomaly detection is a valuable tool for the airline industry to discover fraudulent behavior.

Configuration systems must be able to deal with inconsistencies which can occur in different contexts. Especially in interactive settings, where users specify requirements and a constraint solver has to identify solutions, inconsistencies may more often arise. In inconsistency situations, there is a need of diagnosis methods that support the identification of minimal sets of constraints that have to be adapted or deleted in order to restore consistency. A diagnosis algorithm’s performance can be evaluated in terms of time to find a diagnosis (runtime) and diagnosis quality. Runtime efficiency of diagnosis is especially crucial in real-time scenarios such as production scheduling, robot control, and communication networks. However, there is a trade off between diagnosis quality and the runtime efficiency of diagnostic reasoning. In this paper, we deal with solving the quality-runtime performance trade off problem of direct diagnosis. In this context, we propose a novel learning approach for constraint ordering in direct diagnosis. We show that our approach improves the runtime performance and diagnosis quality at the same time.

When ontologies reach a certain size and complexity, faults such as inconsistencies or wrong entailments are hardly avoidable. Locating the faulty axioms that cause these faults is a hard and time-consuming task. Addressing this issue, several techniques for semi-automatic fault localization in ontologies have been proposed. Often, these approaches involve a human expert who provides answers to system-generated questions about the intended (correct) ontology in order to reduce the possible fault locations. To suggest as few and as informative questions as possible, existing methods draw on various algorithmic optimizations as well as heuristics. However, these computations are often based on certain assumptions about the interacting user and the metric to be optimized.In this work, we critically discuss these optimization criteria and suppositions about the user. As a result, we suggest an alternative, arguably more realistic metric to measure the expert’s effort and show that existing approaches do not achieve optimal efficiency in terms of this metric. Moreover, we detect that significant differences regarding user interaction costs arise if the assumptions made by existing works do not hold. As a remedy, we suggest a new notion of expert question that does not rely on any assumptions about the user’s way of answering. Experiments on faulty real-world ontologies testify that the new querying method minimizes the necessary expert consultations in the majority of cases and reduces the computation time for the best next question by at least 80 % in all scenarios.

Medical diagnosis can be defined as the detection of a disease by examining a patient’s signs, symptoms and history. Diagnostic reasoning can be viewed as a process of testing hypotheses guided by symptoms and signs. Solutions to diagnostic problems can be found by generating a limited number of hypotheses early in the diagnostic process and using them to guide subsequent collection of data. Each hypothesis, if correct, can be used to pre-dict what additional findings must be present, and the diagnostic process would then be a guided search for these findings. The process depends on the medical knowledge available. Description Logic-based ontologies provide class definitions (i.e., the necessary and sufficient conditions for defining class membership). In medicine, these definitions correspond to diagnostic criteria, i.e., the particular form of diseases should be associated with the relevant disease categories. In this paper, we model medical diagnosis as an (iterative) abductive reasoning process using ALC. ALC is employed to take advantage of its inference services. However, the inference capabilities provided by DL are not sufficient for diagnosis purposes. The contributions of the paper include: (1) arguing for the need for a disease-symptoms ontology, (2) proposing an ontological representation which, beside facilitating abductive reasoning, takes into account the diagnostic criteria such that specific patient conditions can be classified under a specific disease, and (3) employing Abox abduction to capture the process of medical diagnosis (the process of generating and testing hypotheses) on this proposed representation.

Semantic Web is rapidly becoming a reality through the development of Linked Data in recent years. Linked Data uses RDF data model to describe statements that link arbitrary data resources on the Internet. It can facilitate to infer new data resources at runtime through the RDF links, and then provide more complete answers as new data resources appear on the Internet. Linked Data provides the means to reach the goal of Semantic Web. At present, Linked Data being used only in the promotion of information sharing or exchange is not a semantic inference due to the lack of an easily shared inference engine. This study addresses the issue developing a Lightweight Linked Data Reasoner (LLDR) which is based on Jena reasoner and is implemented in the apache Axis2. To illustrate the LLDR application, this study developed the Vehicle Ontology to annotate project document from heterogeneous and distributed project resources as Linked Data.

The research presented here extends a previous prototype that supported human recollection with tag clouds created from the use of a personal calendar and Twitter. That system weighted keywords by combining term frequency and the number of photos taken by users to recall memorable events. The aim in this paper is to improve upon our previous work and present a full system that uses tag clouds for recalling personal memories. The main differences from our previous work are as follows. (1) Multiple information sources such as SNSs or instant messengers can be used. (2) To handle multiple information sources, we present a new unified keyword-weighting algorithm. (3) We implemented new functions, such as keyword search, tag search, and photo display, to form a complete system. Preliminary experiments reveal the usefulness of our system in recalling personal memories.

Skypatterns are important since they enable to take into account user preference through Pareto-dominance. Given a set of measures, a skypattern query finds the patterns that are not dominated by others. In practice, different users may be interested in different measures, and issue queries on any subset of measures (a.k.a subspace). This issue was recently addressed by introducing the concept of skypattern cubes. However, such a structure presents high redundancy and is not well adapted for updating operations like adding or removing measures, due to the high costs of subspace computations in retrieving skypatterns. In this paper, we propose a new structure called Compressed Skypattern Cube (abbreviated CSKYC), which concisely represents a skypattern cube, and gives an efficient algorithm to compute it. We thoroughly explore its properties and provide an efficient query processing algorithm. Experimental results show that our proposal allows to construct and to query a CSKYC very efficiently.

Instance matching is one of the processes that facilitate the integration of independently designed knowledge bases. It aims to link co-referent instances with an owl:sameAs connection to allow knowledge bases to complement each other. In this work, we present VDLS, an approach for automatic alignment of instances in RDF knowledge base graphs. VDLS generates for each instance a virtual document from its local description (i.e., data-type properties) and instances related to it through object-type properties (i.e., neighbors). We transform the instance matching problem into a document matching problem and solve it by a vector space embedding technique. We consider the pre-trained word embeddings to assess words similarities at both the lexical and semantic levels. We evaluate our approach on multiple knowledge bases from the instance track of OAEI. The experiments show that VDLS gets prominent results compared to several state-of-the-art existing approaches.

The number of music in digital format increases years after years. The amount of data available allows streaming services to offer wide variety of music. This makes these services attractive. Automatic classification process is required to manage and structure all files available. Automatic music classification is an active field of research. Researches rely on machine learning techniques such as deep neural network. These techniques used features extracted from raw data to generate classification. Features have an important impact on results. Selecting the right descriptors is one of the main difficulties associated with automatic music classification. Opacity of current methods makes it difficult to evaluate the contribution of descriptors in the classification process. In this paper, we propose to use association rules to add more transparency and interpretability.

Due to the growing aging of the population and the impact of falls on the health and autonomy of the older people, the development of cost-effective non-invasive automatic fall detection systems (FDS) has gained much attention. This work proposes and analyzes the capability of convolutional deep neural networks to detect fall events based on the measurements captured by wearable tri-axial accelerometers that are transported by the user to characterize the mobility of the body. The study is performed on a long public data repository containing the traces obtained from a wide group of experimental users during the execution of a predetermined set of Activities of the Daily Living (ADLs) and mimicked falls. The system is evaluated in term of accuracy, sensitivity and specificity when the network is alternatively fed with the module of the acceleration and the with the tri-axial components of the acceleration.

Indoor positioning methods make it possible to estimate the location of a mobile object in a building. Many of these methods rely on fingerprinting approach. First, signal strength data is collected in a number of reference indoor locations. Frequently, the vectors of the strength of the signals emitted by WiFi access points acquired in this way are used to train machine learning models, including instance-based models.In this study, we address the problem of signal strength data acquisition to verify whether different strategies of selecting signal strength data for model testing are equivalent. In the analysed case, the content of a testing data set can be created in a variety of ways. First of all, leave-one-out approach can be adopted. Alternatively, data from randomly selected points or same grid points can be used to estimate method accuracy. We show which of these and other approaches yield different accuracy estimates and in which cases these differences are statistically significant. Our study extends previous studies on analysing the performance of indoor positioning systems. At the same time, it illustrates an interesting problem of testing data acquisition and balancing the conflicting needs of collecting testing data in similar, yet different conditions compared to how training data was acquired.

Nowadays, artificial intelligence makes a great success in our modern social life. The human should be prepared to be able to live with social robots that can provide him comfort and help in solving complex processes. Extending the use of robot’s technology is certainly desirable, but preventing certain catastrophes from the misdeeds of artificial intelligence is crucial. One of these troubles could be for instance the creation of robots’ coalitions to impose pernicious decisions. As a contribution to cope with such issue, we propose a parallel approach for the detection of cultural coalitions based on Bat Algorithm. This GPU-based bat algorithm approach can treat very large datasets due to the possibility of launching several artificial bats simultaneously, which contribute to reducing the runtime without affecting the performance. To proof the effectiveness of the parallel detection coalition method, we conducted several experiments on datasets of different sizes. These datasets represent the result of cultural artificial agents playing the colored trails (CT) game. For the creation of agents’ profiles, we use real cultural datasets generated based on the WV survey. The experimental analysis demonstrates that the use of the proposed method will considerably reduce the runtime.

The classical problem of balancing an inverted pendulum is commonly used to evaluate control learning techniques. Traditional learning methods aim to improve the performance of the learned controller, often disregarding comprehensibility of the learned control policies. Recently, Explainable AI (XAI) has become of great interest in the areas where humans can benefit from insights discovered by AI, or need to check whether AI’s decisions make sense. Learning qualitative models allows formulation of learned hypotheses in a comprehensible way, closer to human intuition than traditional numerical learning. In this paper, we use a qualitative approach to learning control strategies, which we demonstrate on the problem of balancing an inverted pendulum. We use qualitative induction to learn a qualitative model from experimentally collected numerical traces, and qualitative simulation to search for possible qualitative control strategies, which are tested through reactive execution. Successful behaviors provide a clear explanation of the learned control strategy.

In agriculture, the detection of parasites on the crops is required to protect the growth of the plants, increase the yield, and reduce the farming costs. A suitable solution includes the use of mobile robotic platforms to inspect the fields and collect information about the status of the crop. Then, by using machine learning techniques the classification of infected and healthy samples can be performed. Such approach requires a large amount of data to train the classifiers, which in most of the cases is not available given constraints such as weather conditions in the inspection area and the hardware limitations of robotic platforms. In this work, we propose a solution to detect the downy mildew parasite in sunflowers fields. A classification pipeline detects infected sunflowers by using a UAV that overflies the field and captures crop images. Our method uses visual information and morphological features to perform infected crop classification. Additionally, we design a simulation environment for precision agriculture able to generate synthetic data to face the lack of training samples due to the limitations to perform the collection of real crop information. Such simulator allows to test and tune the data acquisition procedures thus making the field operations more effective and less failure prone.

Distributed embedded systems for safety-critical applications demand reliable communication with real-time characteristics. Switched Ethernet-based network is a practical and scalable solution that allows high-reliability level through path redundancy. The Time-Sensitive Networking (TSN) standard is being developed to support real-time communication which supports a deterministic and low latency communication for safety-critical control applications, namely, Time-Triggered (TT) traffic class. In addition, a bounded-latency traffic class, namely, Audio Video Bridging (AVB) class is introduced. In this paper, we propose a multipath routing technique which tackles both TT and AVB traffic simultaneously. The proposed approach investigates satisfying path redundancy requirements for each message while the imposed interference from TT traffic on AVB traffic is minimized. The considered routing problem is formalized as an Integer Linear Programming (ILP) optimization problem. The Worst Case end-to-end Delay (WCD) is the optimization objective. 50 test cases of various network size and number of messages are solved to evaluate the performance, i.e., interference reduction, and the scalability of the proposed technique. Results demonstrate WCD reduction up to 90% comparing to the typical routing approach that determines TT and AVB routing in separate steps.

Traffic sub-area division is an important problem in traffic management and control. This paper proposes a clustering-based approach to this problem that takes into account both temporal and spatial information of vehicle trajectories. Considering different orders of magnitude in time and space, we employ a z-score scheme for uniformity and design an improved density peak clustering method based on a new density definition and similarity measure to extract hot regions. We design a distribution-based partitioning method that employs k-means algorithm to split hot regions into a set of traffic sub-areas. For performance evaluation, we develop a traffic sub-area division criterium based on the $$S_Dbw$$ indicator and the classical Davies-Bouldin index in the literature. Experimental results illustrate that the proposed approach improves traffic sub-area division quality over existing methods.

The growing use of computer-like tablets and PCs in educational settings is enabling more students to study online courses featuring computer-aided tests. Preparing these tests imposes a large burden on teachers who have to prepare a large number of questions because they cannot reuse the same questions many times as students can easily memorize their solutions and share them with other students, which degrades test reliability. Another burden is appropriately setting the level of question difficulty to ensure test discriminability. Using magic square puzzles as examples of mathematical questions, we developed a method for automatically preparing puzzles with appropriate levels of difficulty. We used crowdsourcing to collect answers to sample questions to evaluate their difficulty. Item response theory was used to evaluate the difficulty of the questions from crowdworkers’ answers. Deep learning was then used to build a model for predicting the difficulty of new questions.

Skyline queries constitute an appropriate tool that can help users to make intelligent decisions in the presence of multidimensional data when different, and often contradictory criteria are to be taken into account. Based on the concept of Pareto dominance, the skyline process extracts the most interesting (not dominated in sense of Pareto) objects from a set of data. However, this process often leads to a huge skyline, which is less informative for the end-users. In this paper, we propose an efficient approach to refine the skyline and reduce its size, using the principle of the formal concepts analysis. The basic idea is to build a formal concept lattice for skyline objects based on the minimal distance between each concept and the target concept. We show that the refined skyline is given by the concept that contains k objects (where k is a user-defined parameter) and has the minimal distance to the target concept. A set of experiments are conducted to demonstrate the effectiveness and efficiency of our approach.

Fault tree (FT) model is one of the most popular techniques for probabilistic risk analysis of large, safety critical systems. Probabilistic graphical models like Bayesian networks (BN) or Probabilistic Relational Models (PRM) provide a robust modeling solution for reasoning under uncertainty. In this paper, we define a general modeling approach using a PRM. This PRM can represent any FT with possible safety barriers, spatial information about localization of events, or resources management. In our proposed approach, we define a direct dependency between the resources allocated to one location and the strength of the barriers related to this same location. We will show how this problem can be fully represented with a PRM by defining its relational schema and its probabilistic dependencies. This model can be used to estimate the probability of some risk scenarios and to assess the presence of resources on each location through barrier’s efficiency on risk reduction.

In this paper, we apply a new promising tool for pattern classification, namely, the Tsetlin Machine (TM), to the field of disease forecasting. The TM is interpretable because it is based on manipulating expressions in propositional logic, leveraging a large team of Tsetlin Automata (TA). Apart from being interpretable, this approach is attractive due to its low computational cost and its capacity to handle noise. To attack the problem of forecasting, we introduce a preprocessing method that extends the TM so that it can handle continuous input. Briefly stated, we convert continuous input into a binary representation based on thresholding. The resulting extended TM is evaluated and analyzed using an artificial dataset. The TM is further applied to forecast dengue outbreaks of all the seventeen regions in Philippines using the spatio-temporal properties of the data. Experimental results show that dengue outbreak forecasts made by the TM are more accurate than those obtained by a Support Vector Machine (SVM), Decision Trees (DTs), and several multi-layered Artificial Neural Networks (ANNs), both in terms of forecasting precision and F1-score.

The maintenance of Case-Based Reasoning (CBR) systems has attracted increasing interest within current research since they proved high-quality results in different real-world domains. This kind of systems stores previous experiences, which are described by a vocabulary (e.g., attributes), incrementally in a case base. Actually, the vocabulary presents one among the most important maintenance targets, since it highly contributes in providing accurate solutions and in improving systems’ performance, especially within high-dimensional domains. However, there is no policy, in the literature, that offers the ability to exploit prior knowledge (e.g., given by domain-experts) during the maintenance of features describing cases. In this paper, we propose a flexible policy for the most relevant attribute selection based on the attribute clustering concept. This new policy is able, on the one hand, to manage uncertainty using the belief function theory based tools, and on the other hand, to make use of domain-experts knowledge in form of pairwise constraints: If two attributes offer the same information without any added-value, then a Must-link constraint between them is generated. Otherwise, if there is no relation between them and they offer different information, then a Cannot-link constraint between them is created.

Assumptions are needed to perform non-monotonic reasoning in assumption-based argumentation (ABA), while hypotheses are used to perform abductive or hypothetical reasoning. However not only were hypotheses sometimes confused with assumptions when representing them in ABA frameworks but no work has been done to perform extended abduction in ABA frameworks whose languages contain explicit negation $$\lnot $$ . Hence first we define consistency of ABA frameworks w.r.t. $$\lnot $$ . Second based on it, we present the framework to perform extended abduction in ABA while treating hypotheses along with assumptions in the framework. Theoretically it is shown that Sakama and Inoue’s extended abduction w.r.t. an abductive logic program (ALP) containing classical negation can be captured by our extended abduction in ABA instantiated with the ALP. Finally we provide the method to compute extended abduction in ABA based on answer set programming.

In this paper a computational model of Deep Brain Stimulation (DBS) therapy for post-traumatic stress disorder is presented. The considered therapy has as a goal to decrease the stress level of a stressed individual by using electrode which placed in a specific area in brain. Several areas in brain have been used to decrease the stress level, one of them is Amygdala. The presented temporal-causal network model aims at integrative modeling a Deep Brain Stimulation therapy where the relevant brain areas are modeled in a dynamic manner.

The use of fuzzy logic to generate optimal actions for hardware architecture reconfiguration offers flexible and efficient solutions. In this paper, a new fuzzy approach is proposed in order to guarantee the balance between real time periodic application schedulability and energy consumption optimization under multi-core architecture. Dynamic voltage/frequency scaling (DVFS) has been a key technique in exploiting the processors configurable characteristics. However, for large class of applications in embedded real time systems, the variable operating frequency interferes with tasks deadline respect. The problem is seen as multi-criteria multi-objective decision making issue with dependent criteria. The approach calculates, in offline mode and in online mode, the optimal number of activated homogenous cores and their frequency. Simulated and tested on periodic task sets generated with different system charges, the proposed intelligent technique is support decision system that shows significant results.

This is the 5th paper in our series of papers on hierarchical learning for classification. Hierarchical learning for classification is an automated method of creating hierarchy list of learnt models that are on the one hand capable of partitioning the training set into equal number of subsets and on the other hand are also capable of classifying elements of each corresponding subset into classes of the problem. In this paper, the probabilistic hierarchical learning for classification has been formalized and presented as a theory. The theory asserts that the accurate models of complex datasets can be produced through hierarchical application of low complexity models. The theory is validated through experiments on five popular real-world datasets. Generalizing ability of the theory is also tested. Comparison with the contemporary literature points towards promising future for this theory. The theory is covered by four postulates, which are carved out elegantly through mathematical formalisms.

Nowadays, we note the dominance of the online reviews which become an essential factor in customers’ decision to purchase a product or service. Driven by the immense financial profits from reviews, some corrupt individuals or organizations deliberately post fake reviews to promote their products or to demote their competitors’ products, trying to mislead or influence customers. Therefore, it is crucial to spot these spammers in order to detect the deceptive reviews, to protect companies from this harmful action and to ensure the readers confidence. In this way, we propose a novel approach able to detect spammers and to accord a spamicity degree to each reviewer relying on some spammers indicators while handling the uncertainty in the different inputs through the strength of the belief function theory. Tests are conducted on a real database from Tripadvisor to evaluate our method performance.

Reinforcement Learning has achieved an exceptional performance in the last decade, yet its application to robotics and control remains a field for deeper investigation due to potential challenges. These include high-dimensional continuous state and action spaces, as well as complicated system dynamics and constraints in robotic settings. In this paper, we demonstrate a pioneering experiment in applying an existing model-based RL framework, PILCO, to the problem of time-optimal control. At first, the algorithm models the system dynamics with Gaussian Processes, successfully reducing the effect of model biases. Then, policy evaluation is done through iterated prediction with Gaussian posteriors and deterministic approximate inference. Finally, analytic gradients are used for policy improvement. A simulation and an experiment of an autonomous car completing a rest-to-rest linear locomotion is documented. Time-optimality and data efficiency of the task are shown in the simulation results, and learning under real-world circumstances is proved possible with our methodology.

Flexible gripping mechanisms are advantageous for robots when dealing with dynamic environments due to their compliance. However, a major obstacle to using commercially-available flexible fingers is the lack of appropriate feedback sensors. In this paper, we propose a novel integration of flexible fingers with commercial off-the-shelf proximity sensors. This integrated system enables us to perform non-interfering measurements of even minor deformations in the flexible fingers and consequently deduce information about grasped objects without the need of advanced fabrication methods. Our experiments have demonstrated that the sensor is capable of robustly detecting grasps on most test objects with an accuracy of 100% without false positives by relying on simple, yet powerful signal processing and can detect deformations of less than 0.03 mm. In addition, the sensor detects objects that are slipping through the flexible fingers.

Smart factories are on the verge of becoming the new industrial paradigm, wherein optimization permeates all aspects of production, from concept generation to sales. To fully pursue this paradigm, flexibility in the production means as well as in their timely organization is of paramount importance. AI planning can play a major role in this transition, but the scenarios encountered in practice might be challenging for current tools. We explore the use of SMT at the core of planning techniques to deal with real-world scenarios in the emerging smart factory paradigm. We present special-purpose and general-purpose algorithms, based on current automated reasoning technology and designed to tackle complex application domains. We evaluate their effectiveness and respective merits on a logistic scenario, also extending the comparison to other state-of-the-art task planners.

Soft biometric analysis aims at recognizing personal traits that provide some information about an individual. In this paper, we present a real-time system able to automatically recognize soft biometric traits to enhance the capability of a social robot, Pepper, in this case, to understand characteristics of people present in the environment and to properly interact with them. In particular, the proposed system is able to estimate several traits simultaneously, such as gender, age, the presence of eyeglasses and beard, of people in the field of view of the robot. Our hypothesis is that adding these capabilities to a social robot improves and makes more believable its social behavior. Results of the preliminary evaluation seems to support this hypothesis.

Robot localization is a fundamental capability of all mobile robots. Because of uncertainties in acting and sensing and environmental factors such as people flocking around robots there is always the risk that a robot loses its localization. Very often behaviors of robots rely on a valid position estimation. Thus, for dependability of robot systems it is of great interest for the system to know the state of its localization component. In this paper we present an approach that allows a robot to asses if the localization is still valid. The approach assumes that the underlying localization approach is based on a particle filter. We use deep learning to identify temporal patterns in the particles in the case of losing/lost localization in combination with weak classifiers from the particle set and perception for boosted learning of a localization monitor. The approach is evaluated in a simulated transport robot environment where a degraded localization is provoked by disturbances cased by dynamic obstacles.

In most companies Business-Management software has become omnipresent in recent years. These systems have been introduced to streamline productivity and handle data in a more centralized fashion. Yet, these systems are often modelled by complex processes which makes navigating through them a challenging task. In this work, we introduce a text-based chatbot to a mid-sized Austrian company to facilitate the interaction with these software solutions and thus support them in managing their customer relationships. To learn more about positive as well as negative effects the chatbot has on the employees, we conduct a technical as well as an empirical evaluation. We, for instance, hypothesize that younger staff, who grew up with computers and smart phones, are more open to the conversational system than older employees. We implement a customized solution which integrates seamlessly into the company’s system. The implementation process is informed by related work on Customer-Relationship-Management software, structures of a conversation as well as the typical architecture of a chatbot. In the technical evaluation, objective metrics such as average response time were measured. For the empirical evaluation, a questionnaire was sent out to 15 participating employees asking about subjective metrics such as task ease or user experience. In sum, employees were satisfied with Usage, Interaction Pace and System Response.

In a university context, discussion forums are mostly available in Learning and Management Systems (LMS) but are often ineffective in encouraging participation due to poorly designed user interface and the lack of motivating factors to participate. Our integrated platform with the Telegram mobile app and a web-based forum, is capable of automatic thoughtfulness assessment of questions and answers posted, using text mining and Natural Language Processing (NLP) methodologies. We trained and applied the Random Forest algorithm to provide instant thoughtfulness score prediction for the new posts contributed by the students, and prompted the students to improve on their posts, thereby invoking deeper thinking resulting in better quality contributions. In addition, the platform is designed with six features to ensure that students remain actively engaged on the platform. We report the performance of our platform based on our implementations for a university course in two runs, and compare with existing systems to show that by using our platform, students’ participation and engagement are highly improved, and the quality of posts will increase. Most importantly, our students’ performance in the course was shown to be positively correlated with their participation in the system.

Text normalisation is an important task in the context of Natural Language Processing. By normalisation, free text is mapped into dictionaries, i.e. indexed collections of locutions recognised as typical of a particular jaergon. In general, technical dictionaries are difficult to build and validate. They are typically constructed by hand on the basis of everyday human work and they are agreement-based. This is indubitably time consuming and the approach requires a strong human supervision and does not provide a general methodology. In this paper, we perform the first steps towards the to automatic building of a dictionary for Italian journalistic lexicon, called NewsDict, based on sub dictionaries able to characterise main topics occurring in newspaper articles. We exploit a dataset of annotated documents from some Italian newspapers and a statistical techniques based on the Mutual Information Principle. Documents contains information such as the release date and the topic of the article and has been directly annotated by the author. To check the accuracy of the dictionary we built, we develop an initial test. We normalise a control set of journal article into NewsDict. Crossing results presented in this paper against the human annotation, we provide a fist measure of performances of the described methodology.

As more and more users express their opinions on many topics on Twitter, the sentiments contained in these opinions are becoming a valuable source of data for politicians, researchers, producers, and celebrities. These sentiments significantly affect the decision-making process for users when they assess policies, plan events, design products, etc. Therefore, users need a method that can aid them in making decisions based on the sentiments contained in tweets. Many studies have attempted to address this problem with a variety of methods. However, these methods have not mined the level of users’ satisfaction with objects related to specific topics, nor have they analyzed the level of users’ satisfaction with that topic as a whole. This paper proposes a decision-making support method to deal with the aforementioned limitations by combining object sentiment analysis with data mining on a binary decision tree. The results prove the efficacy of the proposed approach in terms of the error ratio and received information.

This paper presents a named entity recognition method which finds predetermined entities in an unstructured text. The method uses word similarities based on typical word transformations (lemmatization and stemming), word embeddings and character level based similarity to map those entities onto words in the text. The approach is language independent, though language-dependent components are used for lemmatization, stemming and word embedding, and works on any given set of entities. Special attention is given to the entities which are represented in a hierarchical form with the hypernymy-hyponymy relation. The proposed method has the following advantages: it finds the normalized form of the recognized entity name; it is easy to adjust to a new domain; it respects the hierarchical organization of entities; and due to the modular approach can be constantly improved just by updating components for lemmatization, stemming or word embedding. The proposed entity recognition method was tested on a test set of tourist queries and hierarchical entities collected from Slovenia.info tourist portal.

This paper provides a framework to efficiently discover production performance and its application plan by analyzing massive amounts of comments from online review data, especially in the field of hospitality. In order to achieve the goal, two stages of text analytics of sentiment analysis and structural topic model estimating are integrated to classify sentimental polarity of each reviews and elicit hidden dimensions of products or services. Based on these dimensions and polarities, this paper verifies key attributes which impact customer satisfaction by adapting logistic regression. This study extends prior research limitation which focused on discovering the product defect by (1) strength detection, (2) time series analysis, and (3) explanation of the relationship between crucial factors and polarity of the review as a proxy of customer satisfaction. By integrating text analytics from computational linguistic and a traditional statistical method, this paper is expected to contribute on both academical and practical implications.

Chatbots as conversational recommender have gained increasing importance for research and practice with a lot of applications available today. In this paper, we present the methods to support conversational defaults within a human-chatbot conversation that simplifies communication with the purpose of improving the overall recommendation process. In particular, we discuss our model-based reasoning approach for easing user experience during a chat, e.g., in cases where user preferences are mentioned indirectly causing inconsistencies. As a consequence of inconsistencies, it would not be possible for the chatbot to provide answers and recommendations. The presented approach allows for removing inconsistencies during the interactions with the chatbot. Besides the basic foundations, we provide use cases from the intended tourism domain to show the simplification of the conversation process. In particular, we consider recommendations for booking hotels and planning trips.

In this paper, to determine the optimal parameters of the product acceptance process under parametric uncertainty of underlying models, a new intelligent technique for optimization of product acceptance process on the basis of misclassification probability is proposed. It allows one to take into account all possible situations that may occur when it is necessary to optimize the product acceptance process. The technique is based on the pivotal quantity averaging approach (PQAA) which allows one to eliminate the unknown parameters from the problem and to use available statistical information as completely as possible. It is conceptually simple and easy to use. One of the most important features of the proposed new intelligent technique for optimization of product acceptance process on the basis of misclassification probability is its great generality, enabling one to optimize diverse problems within one unified framework. To illustrate the proposed technique, the case of log-location-scale distributions is considered under parametric uncertainty.

Bi-Level Optimization Problems (BLOPs) belong to a class of challenging problems where one optimization problem acts as a constraint to another optimization level. These problems commonly appear in many real-life applications including: transportation, game-playing, chemical engineering, etc. Indeed, multi-objective BLOP is a natural extension of the single objective BLOP that bring more computational challenges related to the multi-objective hierarchical decision making. In this context, a well-known algorithm called NSGA-II was presented in the literature among the most cited Multi-Objective Evolutionary Algorithm (MOEA) in this research area. The most prominent features of NSGA-II are its simplicity, elitist approach and a non-parametric method for diversity. For this reason, in this work, we propose a bi-level version of NSGA-II, called Bi-NSGA-II, in an attempt to exploit NSGA-II features in tackling problems involving bi-level multiple conflicting criteria. The main motivation of this paper is to investigate the performance of the proposed variant on a bi-level production distribution problem in supply chain management formulated as a Multi-objective Bi-level MDVRP (M-Bi-MDVRP). The paper reveals three Bi-NSGA-II variants for solving the M-Bi-MDVRP basing on different variation operators (M-VMX, VMX, SBX and RBX). The experimental results showed the remarkable ability of our adopted algorithm for solving such NP-hard problem.

Nowadays there are studies that show that bridges collapse mainly by scour caused by hydraulic action which implies high reconstruction costs. We can avoid the collapse of the bridge, reinforcing it by means of a system of cable-stayed arches, which will make it possible to hold the deck and eliminate the damaged elements. To this end, it is necessary to optimize the order and the magnitudes of adjustment of the tension of the hangers. In this paper, the use of the Animal Migration Optimization algorithm is proposed for the search of an optimal solution in a gradual way, which consists of two processes: the simulation of how groups of animals move from a current position to a new, and how during migration some animals leave the group and join others. Finally, we present experimental results, where the performance of the Animal Migration Optimization algorithm can be observed.

Although ridesharing is a potential transport model for reducing fuel consumption, green-house gas emissions and improving efficiency, it is still not widely adopted due to the lack of providing monetary incentives for ridesharing participants. Most studies regarding ridesharing focus on travel distance reduction, cost savings and successful matching rate in ridesharing systems, which do not directly provide monetary incentives for the ridesharing participants. In this paper, we address this issue by proposing a performance index for ridesharing based on monetary incentives. We formulate a problem to optimize monetary incentives in ridesharing systems as non-linear integer programming problem. To cope with computational complexity, an evolutionary computation approach based on a variant of PSO is adopted to solve the non-linear integer programming problem for ridesharing systems based on cooperative coevolving particle swarms. The results confirm the effectiveness the proposed algorithm in solving the nonlinear constrained ridesharing optimization problem with binary decision variables and rational objective function.

A Conditional Preference Network with Comfort (CPC-net) graphically represents both preference and comfort. Preference and comfort indicate user’s habitual behavior and genuine decisions correspondingly. Given that these two concepts might be conflicting, we find it necessary to introduce Pareto optimality when achieving outcome optimization with respect to a given acyclic CPC-net. In this regard, we propose a backtrack search algorithm, that we call Solve-CPC, to return the Pareto optimal outcomes. The formal properties of the algorithm are presented and discussed.

The Set Covering Problem (SCP) is an important combinatorial optimization problem that finds application in a large variety of practical areas, particularly in airline crew scheduling or vehicle routing and facility placement problems. To solve de SCP we employ the Spotted Hyena Optimizer (SHO), which is a metaheuristic inspired by the natural behavior of the spotted hyenas. In this work, in order to improve the performance of our proposed approach we use Autonomous Search (AS), a case of adaptive systems that allows modifications of internals components on the run. We illustrate interesting experimental results where the proposed approach is able to obtain global optimums for a set of well-known set covering problem instances.