Intelligent and Evolutionary Systems

Recently, the necessity to resolve the patrolling problem has become pressing. This problem is modeled using an undirected graph structure in which one or more agents patrol the graph and regularly visit each node with the shortest time interval possible. Some central controlled algorithms have been proposed to solve this problem. However, the reliability of these algorithms, which depends on the central controller and communication between the controller and each agent, is considered insufficient. Thus, algorithms with a central controller are not applicable to critical environments. As an alternative approach, some autonomous and distributed algorithms have been proposed to achieve higher reliability and robustness. In a previous paper, we proposed an autonomous and distributed algorithm, called pheromone- and inverse-degree-based Probabilistic Vertex-Ant-Walk (pidPVAW). pidPVAW uses a pheromone model corresponding to fixed points for agent communication and cooperative patrolling as an extension of pheromone-based PVAW (pPVAW). In this paper, we introduce a new parameter k to control the effect of the degree of the neighbor nodes on the agent decision to move. When $$k = 0$$, pidPVAW behaves like pPVAW; therefore, pidPVAW includes pPVAW. The parameter k controls how easily nodes with lower connectivity can be visited. We ran some computer simulations for the parameter k on square grid graphs and scale-free graphs, and showed its effect on the system.

This paper proposes an evacuation guidance method for use in disaster situations. The method is based on ant colony optimization (ACO). We have implemented the method as ACO-based evacuation system in a simulator and examined the feasibility of the system. Since we cannot depend on the communication infrastructures with a disaster occurs, we make the system utilize mobile ad hoc network (MANET). We expect the ACO-based evacuation system produces quasi-optimized evacuation paths by the cooperation of multiple agents, while MANET provides communication between agents in the environment lacking of network infrastructure. Even though a number of ACO-based guidance systems have been developed, there are still some questions whether evacuees who follow the evacuation paths given by ACO are really safe. We examined how safe following these paths is by simulations, and found that they were not safe in some cases. As a result, in this paper, we propose an improved ACO-based evacuation system that equips deodorant pheromone to actively erase ACO pheromone traces when dangerous locations are found. Our simulation results show the use of deodorant pheromone can improve the safety level of the evacuation guidance system without degrading evacuation efficiency.

In online communities like e-marketplaces, the success of business transactions only refers to the cooperation between reputable business agents. While untrustworthy agents will never have an opportunity and are enforced to leave the systems even they are potential to cooperate. In this study, we propose computational models of an exploration strategy based on forgiveness mechanism for potential untrustworthy agents to recover their reputation. The implementation of this mechanism is centralized in nature. Therefore, it can incorporate with existing reputation systems to improve the efficiency of online trading.

Community structures exist in networks which has complex biological, social, technological and so on structures and contain important information. Networks and community structures in computer systems are presented by graphs and subgraphs respectively. Community structure detection problem is NP-hard problem and especially final results of the best community structures for large-complex networks are unknown. In this paper, to solve community structure detection problem a genetic algorithm-based algorithm, AGA-net, which is one of evolutionary techniques has been proposed. This algorithm which has the property of fast convergence to global best value without being trapped to local optimum has been supported by new parameters. Real-world network which are frequently used in literature has been used as test data and obtained results have been compared with 10 different algorithms. After analyzing the test results it has been observed that the proposed algorithm gives successful results for determination of meaningful communities from complex networks.

A channel is a logical spcae where agents make announcements publicly. Examples of such objects are forums, wikis and social networks. Several questions arise about the nature of such a statement as well as about the attitude of the agent herself in doing these announcements. Does the agent know whether the statement is true? Is this agent announcing that statement or its opposite in any other channel? Extensions to Dynamic Epistemic Logics have been proposed in the recent past that give account to public announcements. One major limit of these logics is that announcements are always considered truthful. It is however clear that, in real life, incompetent agents may announce false things, while deceitful agents may even announce things they do not believe in. In this paper, we shall provide a logical framework, called Multiple Channel Logic, able to relate true statements, agent beliefs, and announcements on communication channels. We discuss syntax and semantics of this logic and show the behaviour of the proposed deduction system. Lastly, we shall present a classification of agents based on the above introduced behaviour analysis.

The increased dimensionality of genomic and proteomic data produced by microarray and mass spectrometry technology makes testing and training of general classification method difficult. Special data analysis is demanded in this case and one of the common ways to handle high dimensionality is identification of the most relevant features in the data. Wrapper feature selection is one of the most common and effective techniques for feature selection. Although efficient, wrapper methods have some limitations due to the fact that their result depends on the search strategy. In theory when a complex search is used, it may take much longer to choose the best subset of features and may be impractical in some cases. Hence we propose a new wrapper feature selection for big data based on a random search using genetic algorithm and prior information. The new approach was tested on 2 biological dataset and compared to two well known wrapper feature selection approaches and results illustrate that our approach gives the best performances.

Most discretization algorithms focus on the univariate case. In general, they take into account the target class or interval-wise frequency of data. In so doing, useful information regarding natural group, hidden pattern and correlation among the attributes may be inevitably lost. In response, this paper introduces a new pruning method that exploits natural groups or clusters as an explicit constraint to traditional cut-point determination techniques. This unsupervised approach makes use of cluster ensembles to reveal similarities between data belonging to adjacent intervals. To be precise, a cut-point between a pair of highly similar or related intervals will be dropped. This pruning mechanism is coupled with three different univariate discretization algorithms, with the evaluation is conducted on 10 datasets and 3 classifier models. The results suggest that the proposed method usually achieve higher classification accuracy levels, than those of the three baseline counterparts.

A classifier aims to categorize instances into well-defined groups based on a model called classifier. One of the most widely used classifiers is a decision tree built using a recursive partitioning algorithm. This paper applies the recursive partitioning technique based on the series of tubes. A tube is identified from three information; 1) a core vector, 2) a tube length and 3) a tube radius. The first component is the core vector generated by the extreme pole and the centroid of the current dataset and the second component is the tube length which is the maximum magnitude of the projections from all instances onto the core vector and the last component is the tube radius which is the maximum distance of the farthest point away from the core vector. Our experiment was performed on synthesized datasets of varying sizes with 2, 4, 6 and 8 attributes. The results showed the improvement over the conditional inference tree and C4.5 tree via the F-measure and G-measure score.

Outlier detection is one of the knowledge discovery problems that identifies a data point which does not agree with majority data points in a dataset. In the real-world datasets, the majority data points normally line up into patterns that can be captured by some models. In this paper, we propose the new outlier detection algorithm based on the dynamically updated tree model. It composes of two-step processes (1) constructing the extreme-centroid tree from a sampling dataset, and (2) dynamically updated extreme-centroid tree. In the extreme-centroid tree construction step, the root initially identifies two extreme data points from the centroid of a sampling dataset and uses them for splitting data points into groups. It continues splitting until the terminal criterion is met. A leaf node with a single data point is assigned as a suspected outlier in this process. The suspected outliers are trimmed from the tree model and sent back to the rest of a dataset. In the dynamically updated extreme-centroid tree step, a data point from the rest of a dataset will be inserted to the tree model, called the new inserted data point, and a single data point in the tree model is randomly removed from this tree model to maintain the amount of current data points, called the expired data point. The new inserted data point and the expired data point will adjust the tree maintaining the linear time complexity. We compared our algorithm with LOF algorithm and COF algorithm on the synthetic dataset and three UCI datasets. In the UCI datasets, a majority class is selected and other classes are randomly picked as the outliers. The results show that our algorithm outperformed when compared to LOF and COF using precision, recall, and F-measure.

Fuzzy integral is a kind of effective fusion tool. Traditionally, fuzzy integral can project the data with n-dimension into one line, in which the projection is along with a group of linear lines. In reality, data distribution is not regular, so the straight line for projection is too limited. Gaussian function is applied to natural science widely. It is close to normal distribution and can cover more data. In this article, a new generalization of fuzzy integral is proposed. The Gaussian function is used as integrand. A new classifier is constructed based on Gaussian Fuzzy integral and applied into several benchmark data sets. The results show that the new version can improve the property of fuzzy integral and obtain the better performance.

Renal failure is one of major medical diseases that is recently on the rise, especially in Thailand. In general, patients with hypertension and diabetes are at high risk of encountering this disorder. The medical cost for a large group of chronic-disease patients has been the burden not only to the local hospitals, but also the country as a whole. Without forward planning, the allocated budget may not cover the expense of increasing cases. This research aims to develop an intelligent model to predict the duration to progress kidney disease in those patients with hypertension and diabetes. As such, the predictive model can help physicians to acknowledge patients’ risk and set up a plan to prolong the progression duration, perhaps by modifying their behaviors. The methodology of data mining is employed for such cause, with records of 360 patients from Phan hospital’s database in Chiang Rai province between 2004 and 2014. Prior model generation, the underlying data has gone through conventional steps of data cleaning and preparation, such that the problems of incomplete and biased data are resolved. To explore the baseline of prediction performance, four classical classification techniques are exploited to create the desired model. These include decision tree, K-nearest neighbor, Naive Bayes, and Artificial Neural Networks. Based on 10-fold cross validation, the overall accuracy obtained with the aforementioned techniques is around 70% to 80%, with the highest of 86.7% being achieved by Artificial Neural Networks.

The problem of student dropout has steadily increased in many universities in Thailand. The main purpose of this research is to develop a model for predicting dropout occurences with the first-year students and determine the factors behind these cases. Despite several classification techniques being made available in the literature, the current study focuses on using decision trees and rule induction models to discover knowledge from data of students at Mae Fah Luang University. The resulting classifiers that is interpretable and analyzed by those involved in the assistant and consultation aid, are built from the collection of different attributes. These include student’s academic performance in the first semester, student social behavior, personal background and education background. With respect to the experiments with various classifiers and the application of data rebalancing algorithm, the results indicate a promising accuracy, hence the reliability of this study as a decision support tool.

Deep Belief Network (DBN) is widely used for modelling and analysis of all kinds of actual problems. However, it’s easy to have a computational bottleneck problem when training DBN in a single computational node. And traditional parallel full-batch gradient descent exists the problem that the speed of convergence is slow when we use it to train DBN. To solve this problem, the article proposes a parallel mini-batch gradient descent algorithm based on Spark and uses it to train DBN. The experiment shows the method is faster than parallel full-batch gradient and the convergence result is better when batch size is relatively small. We use the method to train the DBN, and apply it to text classification. We also discuss how the size of batch impacts on the weights of network. The experiments show that it can improve the precision and recall of text classification compared with SVM when batch size is small.

Recently, retweeting is found to be an important action to understand diffusion in microblogging sites. There have been studies on how tweets propagate in networks. Previous studies have shown that history of users interaction and properties of the message are good attributes to understand the retweet behavior of users. Factors like content of message and time are less investigated. We propose a model for predicting users who are more likely to retweet a particular tweet using tweet properties, time and estimates of pairwise influence among users. We have analyzed retweet cascades and validated that structural, social, behavioral and history of nodes are equally important for influence estimation among users. We develop a model which ranks the users based on the likelihood of the users to be potential retweeters. We have performed experiments on real world Twitter sub-graphs and our results validate our proposed work satisfactorily. We have also compared our results with existing works and our results outperform them.

One of the recent proposed population-based heuristic search algorithms is tree-seed optimization algorithm, TSA for short. TSA simulates the growing over on a land of trees and seeds and it has been proposed for solving unconstrained continuous optimization problems. The trees and their seeds on the D-dimensional solution space correspond to the possible solution for the optimization problem. At the beginning of the search, the trees are sowed to the land, and a number of seeds for each tree are produced during the iterations. The tree is removed from the stand and its best seed is added to the stand if the fitness of the best seed is better than the fitness of this tree. In the present study, a constraint optimization problem, the well-known pressure vessel design-PVD problem, is solved by using TSA. To overcome the constraints of the problem, a penalty function is used and the problem is considered as a single objective optimization problem. The experimental results obtained by the TSA are compared with the results of state-of-art methods such as artificial bee colony (ABC) and particle swarm optimization (PSO). Based on the solution quality and robustness, the promising and comparable results are obtained by the proposed approach.

The uncapacitated facility location problem (UFLP) is a location-based binary optimization problem investigated by using various methods in the literature. This study demonstrates a solution methodology for UFLP by a binary version of a novel swarm intelligence method namely bat algorithm (BA). BA is an optimization method employed for solving continuous optimization problems in the literature, suggested by inspiring the echolocation of microbats in nature. As implemented within some studies, sigmoid function is used in BA in order to obtain binary version of the algorithm (BBA) in this study, and then BBA is used for solving UFLP. According to the experimental results, BBA acquires successful results for solving UFLP in terms of solution quality.

This paper suggests utilization of a novel metaheuristic method namely bat algorithm (BA) in order to solve 2D rectangular strip packing problem. Although BA is proposed for solving continuous optimization problems, a discrete version of BA is developed by being used neighborhood operators to solve the problem dealt with this study. Firstly, bottom left approach is used as the placement algorithm in the problem, then, discrete BA is used for obtaining the proper sequence of the rectangular object list. The performance of the proposed approach is investigated on 9 different problems on well-known 2D rectangular problem literature. Experimental results show that discrete BA is effective and alternatively usable in solving 2D rectangular strip packing problems.

This research article presents a hybrid approach based on an intelligent combination of artificial ants and neurons. Research on different parameter combinations are performed, in order to find the best performing parameter settings. The obtained insights are then subsumed into an intelligent architecture consisting of Ant Colony Optimization and Self Organizing Map.

Migrating birds optimization (MBO) algorithm is a new meta-heuristic algorithm inspired from behaviors of migratory birds during migration. Basic MBO algorithm is designed for quadratic assignment problems (QAP) which are known as discrete problems, and the performance of MBO algorithm for solving QAP is shown successfully. But MBO algorithm could not achieve same performance for some other benchmark problems like traveling salesman problem (TSP) and asymmetric traveling salesman problem (ATSP). In order to deal with these kinds of problems, neighborhood operators of MBO is focused in this paper. The performance of MBO algorithm is evaluated with seven varieties of neighborhood operators on symmetric and asymmetric TSP problems. Experimental results show that the performance of MBO algorithm is improved up to 36% by utilizing different neighborhood operators.

Evolutionary algorithms have been extensively used to solve static and dynamic single objective optimization problems, and static multiobjective optimization problems. However, there has only been tepid interest to solve multiobjective optimization problems in dynamic environments. It is only in the past few years that evolutionary algorithms have been used to solve dynamic multiobjective optimization problems and comprehensive benchmark suites have been proposed for testing the performance of algorithms. Prediction based algorithms may be able to provide information about the location of the changed optima and thereby assisting the evolutionary algorithm in the non-trivial task of tracking the changing Pareto Optimal Front or Set. Kalman filter is one of the widely used techniques in prediction scenarios for state estimation. A Dynamic Multi-objective Evolutionary algorithm was proposed in which the Kalman Filter was applied to the whole population to direct the search for Pareto Optimal Solutions in the decision space after a change in the problem has occurred. In this work, the Kalman Filter assisted Evolutionary Algorithm is tested on the IEEE CEC 2015 Benchmark problems set and the results are presented. It is observed that while the proposed algorithm performs well on some problems, more efficient strategies are required to supplement the algorithm in cases of high change severity, isolated and deceptive fronts.

This paper proposes an ideation (idea generation) support process based on Channel Theory, which is a qualitative information theory. Channel Theory formulates information flow by establishing infomorphism between two classifications. By encoding a set of examples into a classification and the target of ideation into another classification, the infomorphism between two classifications guides analogical reasonings. This paper employs fuben-eki systems as the appropriate target of ideation by our proposed process. Fuben-eki denotes the benefits of inconvenience, and fuben-eki systems give users beneficial inconvenience. Our proposed process was implemented in a experimental system as a web application. The experimental result shows that the system improved not the amount but the experiment quality of ideas. Furthermore, questionnaire answers from the participants elucidated how the process helped them conceive ideas.

Recommender systems are a popular approach for dealing with the problem of product overload. Collaborative Filtering (CF), probably the best known technique for recommender systems, is based on the idea of determining and locating like-minded users. However, CF suffers from a common phenomenon known as the cold-start problem, which prevents the technique from effectively locating suggestions for new users. In this paper we will investigate how to provide a recommendation to a new user, based on a previous group of users opinions, by utilizing techniques from social choice theory. Social choice theory has developed models for aggregating individual preferences and judgments, so as to reach a collective decision. We then determined how these can best be utilized to establish a collective decision as a recommendation for new users; hence, a solution to the cold start problem. This solution not only solves the cold-start problem, but can also be used to give existing users more accurate suggestions. We focused on models of preference aggregation and judgment aggregation; specifically, by using the judgment aggregation model to solve the cold-start problem, which is a novel approach.

Named Entity Recognition (NER) is a foundational technology for systems designed to process Natural Language documents. However, many existing state-of-the-art systems are difficult to integrate into commercial settings (due their monolithic construction, licensing constraints, or need for corpuses, for example). In this work, a new NER system is described that uses the output of existing systems over large corpuses as its training set, ultimately enabling labelling with (i)better F1 scores; (ii)higher labelling speeds; and (iii)no further dependence on the external software.

Humanitarian missions are complex operations that require emergency resources to be delivered in a timely fashion to a disaster area. This article describes the development of a decision-support tool to improve the effectiveness of humanitarian operations through efficient inventory management and quick distribution of emergent resources for disaster areas. Such humanitarian logistics necessitate better coordination and planning. Unlike commercial logistics, humanitarian logistics demands from disaster areas cannot be predicted. Thus, to support a quick and efficient relief operations is important by developing ICT-based decision aids. A decision-support tool is just such an attempt, which allows the design of logistics networks for effective disaster responses. Such a decision-support tool may require the following two key decisions: determining temporary warehouse locations and deciding the means of transportation to points of destination (POD). The design of the humanitarian logistic networks includes: (a) a supply chain network that consists of inventory and distribution management, and (b) a logistic network that includes multimodal transportation of different scales for transportation times.

In this paper, estimated curve Knot points are found for B- Spline Curve by using Niched (Celled) Pareto Genetic Algorithm which is one of the multi objective genetic algorithms. It is necessary to know degree of the curve, control points and knot vector for drawing B-Spline curve. Some knot points are of very few or no effect at all on the drawing of B-Spline curve drawing. Omitting such points will not effect the shape of curve in curve drawing. In this study, it is aimed to find and omit these ineffective curve points from drove of curve. Performance of proposed method are compared with selected studies from literature.

In this paper we provide an insight into the skill representation, where skill representation is seen as an essential part of the skill assessment stage in the Computational Red Teaming process. Skill representation is demonstrated in the context of Sudoku puzzle, for which the real human skills used in Sudoku solving, along with their acquisition, are represented computationally in a cognitively plausible manner, by using feed-forward neural networks with back-propagation, and supervised learning. The neural network based skills are then coupled with a hard-coded constraint propagation computational Sudoku solver, in which the solving sequence is kept hard-coded, and the skills are represented through neural networks. The paper demonstrates that the modified solver can achieve different levels of proficiency, depending on the amount of skills acquired through the neural networks. Results are encouraging for developing more complex skill and skill acquisition models usable in general frameworks related to the skill assessment aspect of Computational Red Teaming.

In this paper, we explore the visual effects of animated 2D line strokes and 3D cubes. A given 2D image is segmented into either 2D line strokes or 3D cubes. Each segmented object (i.e., line stroke or each cube) is initialised with the position and the colour of the corresponding pixel in the image. The program animates these objects using the boid framework. This simulates a flocking behavior of line strokes in a 2D space and cubes in a 3D space. In this implementation the animation runs in a cycle from the disintegration of the original image to a swarm of line strokes or 3D cubes, then the swarm moves about and then integrates back into the original image.

This paper presents the results of an empirical analysis of mobile augmented reality (AR) games focusing on elements of user engagement. The area covered in the analysis was based on the user engagement literature review in classical games as well as AR technology. The results showed that five major elements that affected the user engagement were social, perceived usability, challenge, satisfaction and clear goals. This finding is suggested as one of the key considerations prior in developing mobile AR game.

The objective of a dynamic economic dispatch problem is to determine the optimal power generation from a number of generating units by minimizing the fuel cost. The problem is considered a high-dimensional complex constrained optimization problem. Over the last few decades, many differential evolution variants have been proposed to solve this problem. However, such variants were highly dependent on the search operators, control parameters and constraint handling techniques used. Therefore, to tackle with this shortcoming, in this paper, a new differential evolution framework is introduced. In it, the appropriate selection of differential evolution operators is linked to the proper combination of control parameters (scaling factor and crossover rate), while the population size is adaptively updated. To add to this, a heuristic repair approach is introduced to help obtaining feasible solutions from infeasible ones, and hence enhancing the convergence rate of the proposed algorithm. The algorithm is tested on three different dynamic dispatch problems with 12 and 24 hours planning horizons. The results demonstrate the superiority of the proposed algorithm to the state-of-the-art algorithms.

A global industrial enterprise is a complex network of different distributed production plants producing, inventory, and distributing products. Agent-based model provides the approach to prove complex network problems of independent actors. A global economy and increase in both demand fluctuation and pressures for cost decreasing while satisfying customer services have put a premium on smart supply chain management. It is important to make risk-benefit analysis of supply chain design alternatives before making a final decision. Simulation gives us an effective approach to comparative analysis and evaluation of the alternatives. In this paper, we describe an agent-based simulation tool for designing smart supply chain networks as well logistic networks. Using an agent-based approach, supply chain models are composed from supply chain agents. The agent-based simulation tool can be very useful for predicting the effects of local and system-level activities on multi-plant performance and improving the tactical and strategic decision-making at the enterprise level. Specifically, this model can reveal the optimal method to ship the inventory on some situations which are demand fluctuation and network disruption. The demand fluctuation effects the inventory management. The network disruption restricts the logistics. This model evaluates supply chain management from the viewpoints of the amount of inventory, the way of shipping and cost.

Managing parking lots usually involve tasks that should provide important information such as parked car counts, and available parking spaces and their locations. This can be used to direct drivers in real-time towards empty spaces which will minimise the time spent looking for one and thus reduce traffic congestions. Using an image-based integrated parking system is an effective way to automatically track a parking lot without exhausting time and manual resources. In this paper, we present a low-cost vision-based parking system to manage a closed area parking lot by using cameras that takes real-time footage of the parking lot. The footage is processed using HSV-based histogram technique and the resulting models are compared against pre-trained models. These models define either a Parked or an Empty class. The parking spaces within the processed footage are then categorised using this two classes based on their matching probability.

Advanced centralised access control (CAC) systems are widely deployed on campuses as a means to provide security and track movements. There is concern that some campuses are not using such protective systems, hence this paper attempts to resolve this weaknesses in such institutions’ by developing a simpler CAC system using the Radio Frequency (RF) and contactless smart card technologies. The scope of the developed system is not only limited to access control but also to utilise the gathered data to automate and potentially to support other processes of the institution, such as lecture scheduling and attendance tracking.

Heuristic parameters such as width and height are usually obtained in audio-visual speech recognition. However, the presence of noise has an impact on such system. In the paper, we present a mathematical study investigating whether descriptive parameters derived from lip shapes can improve the performance of the system through the use of a mathematical model. The video database used consists of five separate pronunciations of the numbers ranging from 0 to 9. Three categories of data have been successfully classified; the polynomial coefficient (curving of the lips), width and height (both inner and outer) and also the raw data (coordinates). The results showed that the best classifier is the curving of the bottom lip contour with an accuracy of 90.91% and the weakest classifier is from points on the right upper lip contour with accuracy of 12.24%.

Autonomous object counting system can help industries to keep track of their inventory in real time and adjust their production rate suitably. In this paper we have proposed a robust algorithm which is capable of detecting all the instances of a particular object in a scene image and report their count. The algorithm starts by intelligently selecting Speeded Up Robust Feature (SURF) points on the basis stability and proximity in the prototype image, i.e. the image of the object to be counted. SURF points on the scene image are detected and matched to the ones on the prototype image. The notion of Feature Grid Vector (FGV) and Feature Grid Cluster (FGC) is introduced to group SURF points lying on a particular instance of the prototype. A learning model based on Support Vector Machine has been developed to separate out the true instances of the prototype from the false alarms. Both the training and inference occur almost in real time for all practical purposes. The algorithm is robust to illumination variations in the scene image and is capable of detecting instances of the prototype having different distance and orientation w.r.t. the camera. The complete algorithm has been embodied into a desktop application, which uses a camera feed to report the real time count of the prototype in the scene image.