Intelligent Systems and Applications

Prediction of financial time series is described as one of the most challenging tasks of time series prediction, due to its characteristics and their dynamic nature. Support vector regression (SVR), Support vector machine (SVM) and back propagation neural network (BPNN) are the most popular data mining techniques in prediction financial time series. In this paper a hybrid combination model is introduced to combine the three models and to be most beneficial of them all. Quantization factor is used in this paper for the first time to improve the single SVM and SVR prediction output. And also genetic algorithm (GA) used to determine the weights of the proposed model. FTSE100, S&P 500 and Nikkei 225 daily index closing prices are used to evaluate the proposed model performance. The proposed hybrid model numerical results shows the outperform result over all other single model, traditional simple average combiner and the traditional time series model Autoregressive (AR).

In this work we present two methods based on Association Rules (ARs) for the prediction of bladder cancer recurrence. Our objective is to provide a system which is on one hand comprehensible and on the other hand with a high sensitivity. Since data are not equitably distributed among the classes and since errors costs are asymmetric, we propose to handle separately the cases of recurrence and those of no-recurrence. ARs are generated from each training set using an associative classification approach. The rules’ uncertainty is represented by a confidence degree. Several symptoms of low intensity can be complementary and mutually reinforcing. This phenomenon is taken into account thanks to aggregate functions which strengthen the confidence degrees of the fired rules. The first proposed classification method uses these ARs to predict the bladder cancer recurrence. The second one combines ARs and decision tree: the original base of ARs is enriched by the rules generated from a decision tree. Experimental results are very satisfactory, at least with the AR’s method. The sensibility rates are improved in comparison with some other approaches. In addition, interesting extracted knowledge was provided to oncologists.

Extracting critical sections from project management documents is a challenging process and an active area of research. Project management documents contain certain early warnings that, if modelled properly, may inform the project planners and managers in advance of any impending risks via early warnings. Extraction of such indicators from documents is termed as text mining, which is an active area of research. In the context of construction project management, extraction of semantically crucial information from documents is a challenging task that can in turn be used to provide decision support by optimising the entire project lifecycle. This research presents a two-step modelling and clustering methodology. It exploits the capability of a Naïve Bayes classifier to extract early warnings from management text data. In the first step, a database corpus is prepared via a qualitative analysis of expertly fed questionnaire responses. In the latter stage a Naïve Bayes classifier is proposed which evaluates real-world construction management documents to identify potential risks based on certain keyword usages. The classifier outcome was compared against labelled test documents and gave an accuracy of 68.02 %, which is better than the majority of text mining algorithms reported in the literature.

The early detection of anomalies is an important part of a support system to aid human operators in surveillance tasks. Normally, such an operator is confronted with the overwhelming task to identify important events in a huge amount of incoming data. In order to strengthen their situation awareness, the human decision maker needs an support system, to focus on the most important events. Therefore, the detection of anomalies especially in the maritime domain is investigated in this work. An anomaly is a deviation from the normal behavior shown by the majority of actors in the investigated environment. Thus, algorithms to detect these deviations are analyzed and compared with each other by using different metrics. The two algorithms used in the evaluation are the Kernel Density Estimation and the Gaussian Mixture Model. Compared to other works in this domain, the dataset used in the evaluation is annotated and non-simulative.

Wealthy nations continue to demonstrate their unwavering support to improving conditions and the general well-being of poor countries in spite of the recent economic crises. However, as developmental aid relatively shrinks, both Aid donors and recipient countries have shown keen interest in methodologies used in evaluating developmental assistance programs. Evaluation of aid programs is seen as a complex task mainly because of the several non-aid factors that tend to affect overall outcomes. Adding to the complexity are the subjective sets of criteria used in Aid evaluations programs. This paper proposes a two stage framework of fuzzy TOPSIS and sensitivity analysis to demonstrate how aid-recipient countries can be evaluated to deepen transparency, fairness, value for money and sustainability of such aid programs. Using the Organisation for Economic Co-operation and Development (OECD) set of subjective criteria for evaluating aid programs; a numerical example pre-defined by linguistic terms parameterized by triangular fuzzy numbers is provided to evaluate aid programs. Fuzzy PROMETHEE is used in the first stage to evaluate and rank aid-recipients followed by a comparative analysis with Fuzzy VIKOR and Fuzzy TOPSIS to ascertain an accurateness of the method used. A sensitivity analysis is further added that anticipates possible influences from lobbyists and examines the effect of that bias in expert ratings on the evaluation process. The result shows a framework that can be employed in evaluating aid effectiveness of recipient-countries.

Advanced automation requires complex robotic systems that are susceptible to mechanical, software and sensory failures. While bespoke solutions exist to avoid such situations, there is a requirement to develop generic robotic framework that can allow autonomous recovery from anomalous conditions through hardware or software reconfiguration. This paper presents a novel robotic architecture that combines fuzzy reasoning with ontology-based deliberative decision making to enable self-reconfigurability within a complex robotic system architecture. The fuzzy reasoning module incorporates multiple types of fuzzy inference models that passively monitor the constituent sub-systems for any anomalous changes. A response is generated in retrospect of this monitoring process that is sent to an Ontology-based rational agent in order to perform system reconfiguration. A reconfiguration routine is generated to maintain optimal performance within such complex architectures. The current research work will apply the proposed framework to the problem of autonomous visual navigation of unmanned ground vehicles. An increase in system performance is observed every time a reconfiguration routine is triggered. Experimental analysis is carried out using real-world data, concluding that the proposed system concept gives superior performance against non-reconfigurable robotic frameworks.

In this chapter, the detailed design of a novel adhesion mechanism is described for robots climbing on concrete structures. The aim is to deliver a low-power and sustainable adhesion technique for wall climbing robots to gain access to test sites on large concrete structures which may be located in hazardous industrial environments. A small, mobile prototype robot with on-board force sensor was built which exhibited 360\(^{\circ }\) of manoeuvrability on a 50 \(\times \) 50 mm meshed reinforcement bars test rig with maximum adhesion force of 108 N at 35 mm air gap. The proposed adhesion module consists of three N42 grade neodymium magnets arranged in a unique arrangement on a flux concentrator. Finite Element Analysis (FEA) is used to study the effect of design parameters such as the distance between the magnets, thickness and material of the flux concentrator, use of two concentrators, etc. Using two modules with minimum distance between them showed an increase of 82 N in adhesion force compared to a single module system with higher force-to-weight ratio of 4.36. An adhesion force of 127.53 N was measured on a real vertical concrete column with 30 mm concrete cover. The simulation and experimental results prove that the proposed magnetic adhesion mechanism can generate sufficient adhesion force for the climbing robot to operate on vertical reinforced concrete structures.

Robot hands have attracted increasing research interest in recent years due to their high demand in industry and wide scope in number of applications. Almost all researches done on the robot hands were aimed at improving mechanical design, clever grasping at different angles, lifting and sensing of different objects. In this chapter, we presented the detail classification of control systems and reviewed the related work that has been done in the past. In particular, our focus was on control algorithms implemented on pneumatic systems using PID controller, Bang–bang controller and Backstepping controller. These controllers were tested on our uniquely designed ambidextrous robotic hand structure and results were compared to find the best controller to drive such devices. The five finger ambidextrous robot hand offers total of \(13^\circ \) of freedom (DOFs) and it can bend its fingers in both ways left and right offering full ambidextrous functionality by using only 18 pneumatic artificial muscles (PAMs).

This paper presents a mathematical formulation of the problem of cleaning and delivery task in a large public space with multiple robots, along with a procedural solution based on task reallocation. The task in the cleaning problem is the cleaning zone. A group of robots are assigned to each cleaning zones according to the environmental parameters. Resource constraints make cleaning robots stop operation periodically, which can incur a mission failure or deterioration of the mission performance. In our solution approach, continuous operation is assured by replacing robots having resource problems with standby robots by task reallocation. Two resource constraints are considered in our formulation: the battery capacity and the garbage bin size. This paper describes and compares the performance of three task reallocation strategies: All-At-Once, Optimal-Vector, and Performance-Maximization. The performance measures include remaining garbage volume, cleaning quality, and cleaning time. Task allocation algorithms are tested by simulation in an area composed of 4 cleaning zones, and the Performance-Maximization strategy marked the best performance. Hence, a delivery task is added to the cleaning task. The delivery request operates as a new perturbation factor for the reallocation. The task allocation procedure for the delivery task includes the switching of tasks of the delivery robot itself as well as exchanging among cleaning robots to meet the balance of the cleaning performance. The experiment was conducted with 9 robots with the software architecture that enables multi-functional of a robot and they performed both pseudo-clean and delivery task successfully.

The control of multivariable interacting process is difficult because of the interaction effect between input output variables. In the proposed work, an attempt is made to design a Multivariable Centralized Fractional Order PID (MCFOPID) controller with the use of evolutionary optimization techniques. The Bat Optimization Algorithm (BOA) and Harmony Search algorithm (HS), the evolutionary optimization techniques are used for the tuning of the controller parameters. As the process is a Two-Input-Two-Output process, four FOPID controllers are required for the control of the two interacting conical tank process. Altogether 20 controller parameters need to be tuned. In a single run, all the 20 parameters are founding considering the interaction effect minimizing Integral Time Absolute Error (ITAE). The BOA, HS based MCFOPID controller is validated under tracking and disturbance rejection for minimum ITAE.

The Internet of Things (IoT) paradigm has so far been investigating into designing and developing protocols and architectures to provide connectivity anytime and anywhere for anything. IoT is currently fast forwarding towards embracing a paradigm shift namely Internet of Everything (IoE) where making intelligent decisions and providing services remains a challenge. Context plays an integral role in reasoning the collected data and to provide context-aware services and is gaining growing attention in the IoT paradigm. To this end, a Context-Aware reasoNer (CAN) has been proposed and designed in this chapter. The proposed CAN is a generic enabler and is designed to provide services based on context reasoning. Discovering and filtering entities, i.e. entity configuration, become pivotal in analysing context reasoning to provide right services to right context entities at the right time. This chapter leverages the concept of entity configuration and CAN towards enabling an IoT controller. The chapter further demonstrates use cases and future research directions towards generic CAN development and facilitating context-aware services to IoE.

We present an approach for predicting continuous pedestrian trajectories over a time horizon of 2.5 s by means of polynomial least-squares approximation and multilayer perceptron (MLP) artificial neural networks. The training data are gathered from 1075 real urban traffic scenes with uninstructed pedestrians including starting, stopping, walking and bending in. The polynomial approximation provides an extraction of the principal information of the underlying time series in the form of the polynomial coefficients. It is independent of sensor parameters such as cycle time and robust regarding noise. Approximation and prediction can be performed very efficiently. It only takes 35 \(\upmu \)s on an Intel Core i7 CPU. Test results show 28 % lower prediction errors for starting scenes and 32 % for stopping scenes in comparison to applying a constant velocity movement model. Approaches based on MLP without polynomial input or Support Vector Regression (SVR) models as motion predictor are outperformed as well.

Traditionally, assembly instructions are written in the form of paper or digital manuals. These manuals contain descriptive text, photos or diagrams to guide the user through the assembly sequence from the beginning to the final state. To change this paradigm, an augmented reality system is proposed to guide users in assembly tasks. The system recognizes each part to be assembled through image processing techniques and guides the user through the assembly process with virtual graphic signs. The system checks whether the parts are properly assembled and alerts the user when the assembly has finished. Some assembly assisted by augmented reality systems use some kind of customized device, such as head mounted displays or markers to track camera position and to identify assembly parts. These two features restrict the spread of the technology whence, in this work, customized devices and markers to track and identify parts are not used and all the processing is executed on an embedded software in an off-the-shelf device without the need of communication with other computers to any kind of processing.

This paper introduces a novel study on the performance of Stochastic Diffusion Search (SDS)—a swarm intelligence algorithm—to address DNA sequence assembly problem. This is an NP-hard problem and one of the primary problems in computational molecular biology that requires optimisation methodologies to reconstruct the original DNA sequence. In this work, SDS algorithm is adapted for this purpose and several experiments are run in order to evaluate the performance of the presented technique over several frequently used benchmarks. Given the promising results of the newly proposed algorithm and its success in assembling the input fragments, its behaviour is further analysed, thus shedding light on the process through which the algorithm conducts the task. Additionally, the algorithm is applied to overlap score matrices which are generated from the raw input fragments; the algorithm optimises the overlap score matrices to find better results. In these experiments real-world data are used and the performance of SDS is compared with several other algorithms which are used by other researchers in the field, thus demonstrating its weaknesses and strengths in the experiments presented in the paper.

In late 1940s and with the introduction of cellular automata, various types of problems in computer science and other multidisciplinary fields have started utilising this new technique. The generative capabilities of cellular automata have been used for simulating various natural, physical and chemical phenomena. Aside from these applications, the lattice grid of cellular automata has been providing a by-product interface to generate graphical patterns for digital art creation. One notable aspect of cellular automata is symmetry, detecting of which is often a difficult task and computationally expensive. This paper uses a swarm intelligence algorithm—Stochastic Diffusion Search—to extend and generalise previous works and detect partial symmetries in cellular automata generated patterns. The newly proposed technique tailored to address the spatially-independent symmetry problem is also capable of identifying the absolute point of symmetry (where symmetry holds from all perspectives) in a given pattern. Therefore, along with partially symmetric areas, the centre of symmetry is highlighted through the convergence of the agents of the swarm intelligence algorithm. Additionally this paper proposes the use of entropy and information gain measure as a complementary tool in order to offer insight into the structure of the input cellular automata generated images. It is shown that using these technique provides a comprehensive picture about both the structure of the images as well as the presence of any complete or spatially-independent symmetries. These technique are potentially applicable in the domain of aesthetic evaluation where symmetry is one of the measures.

Maintaining a clean, reliable and efficient electric power system has become more challenging than ever due to the widespread use of solid-state power electronic controlled equipment in industrial, commercial and domestic applications. Non-linear loads draw non-sinusoidal current and reactive power from the source causing voltage and current distortion, increased losses in the power lines and deterioration of the overall power quality of the distribution grid. Devices such as tuned passive filters are among the oldest and most widely used techniques to remove power line harmonics. Other solutions include active power filters which operate as a controllable current source injecting a current that is equal but with opposite phase to cancel the harmonic current. This chapter deals with the design of fuzzy control strategies for a three-phase shunt active power filter to enhance the power quality in a hybrid wind-diesel power system operating in standalone mode. The proposed control scheme is based on Interval Type 2 fuzzy logic controller and is applied to the regulation of the DC bus voltage to compensate for real power unbalances during variable load conditions. A simulation study is performed under Matlab/Simulink to evaluate the performance and robustness of the system under different wind speed conditions.

High Efficiency Video Coding (HEVC/H.265), the latest standard in video compression, aims to halve the bitrate while maintaining the same quality or to achieve the same bitrate with an improved quality compared to its predecessor, AVC/H.264. However, the increase in prediction modes in HEVC significantly impacts on the encoder complexity. Intra prediction methods indeed iterate among 35 modes for each Prediction Unit (PU) to select the most optimal one. This mode decision procedure which consumes around 78 % of the time spent in intra prediction consists of the Rough Mode Decision (RMD), the simplified Rate Distortion Optimisation (RDO) and the full RDO processes. In this chapter considerable time reduction is achieved by using techniques that use fewer modes in both the RMD and the simplified RDO processes. Experimental results show that the average time savings of the proposed method indeed yields a 42.1 % time savings on average with an acceptable drop of 0.075 dB in PSNR and a negligible increase of 0.27 % in bitrate.

In this paper, we have used the multi-agent system (MAS) and the Khalimsky theory to optimize the routing paths and to improve the data management in wireless sensor network (WSN). We have used hierarchical architecture composed by three layers. In the first layer, the sensors are randomly deployed and grouped into clusters for data gathering. The second layer contains the leaders having the highest remaining energy in the clusters. These sensors ensure the fusion and the aggregation of collected data. The third layer contains a set of sensors deployed according to the Khalimsky topology to transmit data packets to the Sink using an optimal path. Each sensor node incorporates an agent to manage the communications and the collected data. In the third layer, mobile agents are used to balance the selection of nodes forming paths between the Sink and the gateways. Our approach showed a reduction of the energy consumption up to 40 % compared with LEACH protocol. Mobile agents were more effective than stationary agents. They allowed to save approximately 5 % of energy. Also, they allowed a balance of tasks distribution and reduced energy gap between nodes which consequently permit the prolongation of the network lifetime.

The paper presents an indirect method learning weights for intuitionistic statement networks. The intuitionistic statement network is a graphical model which represents some associations between statements in the form of the graph. An important element of this model apart from structure represented by the graph are also its parameters that determine the impact of individual statements to other statements. The values of weights are computed for different values of primary statements associated with external sources of observed data. Primary statement values are represented by intuitionistic values and can represent accurate values, which correspond to logical true or false or they represent approximate values. The presented method allows to identification of weights’ values for every instance of evidence of primary statements regardless whether it is an accurate or an approximate value. The estimation of these weights is performed with the application of substitutional generative model represented by Bayesian network. The presented method allows the identification of weights also based on incomplete data.

We present in this chapter a novel method for detecting intrusion into host systems that combines both data and execution flow of programs. To do this, we use sequences of system call traces produced by the host’s kernel, together with their arguments. The latter are further augmented with contextual information and domain-level knowledge in the form of signatures, and used to generate clusters for each individual system call, and for each application type. The argument-driven cluster models are then used to rewrite process sequences of system calls, and the rewritten sequences are fed to a naïve Bayes classifier that builds class conditional probabilities from Markov modeling of system call sequences, thus capturing execution flow. The domain level knowledge augments our machine learning-based detection technique with capabilities of deep packet inspection capabilities usually found, until now, in network intrusion detection systems. We provide the results for the clustering phase, together with their validation using the Silhouette width, the cross-validation technique, and a manual analysis of the produced clusters on the 1999 DARPA data-set from the MIT Lincoln Lab.