Journal of Artificial Intelligence and Data Mining
Volume:2 Issue: 2, Summer-Autumn 2014

In Wireless Sensor Networks (WSNs), localization algorithms could be range-based or range-free. The Approximate Point in Triangle (APIT) is a range-free approach. We propose modification of the APIT algorithm and refer as modified-APIT. We select suitable triangles with appropriate distance between anchors to reduce PIT test errors (edge effect and non-uniform placement of neighbours) in APIT algorithm. To reduce the computational load and avoid useless anchors selection, we propose to segment the application area to four non-overlapping and four overlapping sub-regions. Our results show that the modified-APIT algorithm has better performance in terms of average error and time requirement for all sizes of network with random and grid deployments. For increasing the accuracy of localization and reduction of computation time, every sub-region should contain minimum 5 anchors. The modified-APIT has better performance for different sizes of network for both grid and random deployments in terms of average error and time requirement. Variations of the size of a network and radio communication radius of anchors affect the value of average error and time requirement. To have more accurate location estimation, 5 to 10 anchors per sub-region are effective in modified-APIT.

An integrated model considers all parameters and elements of different deficiencies in one problem. This paper presents a new integrated model of a supply chain that simultaneously considers facility location, vehicle routing and inventory control problems as well as their interactions in one problem, called location-routing-inventory (LRI) problem. This model also considers stochastic demands representing the customers’ requirement. The customers’ uncertain demand follows a normal distribution, in which each distribution center (DC) holds a certain amount of safety stock. In each DC, shortage is not permitted. Furthermore, the routes are not absolutely available all the time. Decisions are made in a multi-period planning horizon. The considered bi-objectives are to minimize the total cost and maximize the probability of delivery to customers. Stochastic availability of routes makes it similar to real-world problems. The presented model is solved by a multi-objective imperialist competitive algorithm (MOICA). Then, well-known multi-objective evolutionary algorithm, namely anon-dominated sorting genetic algorithm II (NSGA-II), is used to evaluate the performance of the proposed MOICA. Finally, the conclusion is presented.

In this paper after a general literature review on the concept of Efficient Frontier (EF), an important inadequacy of the Variance based models for deriving EFs and the high necessity for applying another risk measure is exemplified. In this regard for this study the risk measure of Lower Partial Moment of the first order is decided to replace Variance. Because of the particular shape of the proposed risk measure, one part of the paper is devoted to development of a mechanism for deriving EF on the basis of new model. After that superiority of the new model to old one is shown and then the shape of new EFs under different situations is investigated. At last it is concluded that application of LPM of the first order in financial models in the phase of deriving EF is completely wise and justifiable.

This paper presents designing an optimal adaptive controller for tracking control of robot manipulators based on particle swarm optimization (PSO) algorithm. PSO algorithm has been employed to optimize parameters of the controller and hence to minimize the integral square of errors (ISE) as a performance criteria. In this paper, an improved PSO using logic is proposed to increase the convergence speed. In this case, the performance of PSO algorithms such as an improved PSO (IPSO), an improved PSO using fuzzy logic (F-PSO), a linearly decreasing inertia weight of PSO (LWD-PSO) and a nonlinearly decreasing inertia weight of PSO (NDW-PSO) are compared in terms of parameter accuracy and convergence speed. As a result, the simulation results show that the F-PSO approach presents a better performance in the tracking control of robot manipulators than other algorithms.

Drug addiction is a major social, economic, and hygienic challenge that impacts on all the community and needs serious threat. Available treatments are successful only in short-term unless underlying reasons making individuals prone to the phenomenon are not investigated. Nowadays, there are some treatment centers which have comprehensive information about addicted people. Therefore, given the huge data sources, data mining can be used to explore knowledge implicit in them, their results can be employed as a knowledge base of decision support systems to make decisions regarding addiction prevention and treatment. We studied participants of such clinics including 471 participants, where 86.2% were male and 13.8% were female. The study aimed to extract rules from the collected data by using association models. Results can be used by rehab clinics to give more knowledge regarding relationships between various parameters and help them for better and more effective treatments. E.g. according to the findings of the study, there is a relationship between individual characteristics and LSD abuse, individual characteristics, the kind of narcotics taken, and committing crimes, family history of drug addiction and family member drug addiction.

In this paper, first, an initial feature vector for vocal fold pathology diagnosis is proposed. Then, for optimizing the initial feature vector, a genetic algorithm is proposed. Some experiments are carried out for evaluating and comparing the classification accuracies which are obtained by the use of the different classifiers (ensemble of decision tree, discriminant analysis and K-nearest neighbours) and the different feature vectors (the initial and the optimized ones). Finally, a hybrid of the ensemble of decision tree and the genetic algorithm is proposed for vocal fold pathology diagnosis based on Russian Language. The experimental results show a better performance (the higher classification accuracy and the lower response time) of the proposed method in comparison with the others. While the usage of pure decision tree leads to the classification accuracy of 85.4% for vocal fold pathology diagnosis based on Russian language, the proposed method leads to the 8.5% improvement (the accuracy of 93.9%).

This paper presets a method including a combination of the wavelet transform and fuzzy function approximation (FFA) for high impedance fault (HIF) detection in distribution electricity network. Discrete wavelet transform (DWT) has been used in this paper as a tool for signal analysis. With studying different types of mother signals, detail types and feeder signal, the best case is selected. The DWT is used to extract the best features. The extracted features have been used as the FFA Systems inputs. The FFA system uses the input-output pairs to create a function approximation of the features. The FFA system is able to classify the new features. The combined model is used to model the HIF. This combined model has the high ability to model different types of HIF. In the proposed method, different kind of loads including nonlinear and asymmetric loads and HIF types studied. The results show that the proposed method is able to distinguish no fault and HIF state with high accuracy.

The uncertainty estimation and compensation are challenging problems for the robust control of robot manipulators which are complex systems. This paper presents a novel decentralized model-free robust controller for electrically driven robot manipulators. As a novelty, the proposed controller employs a simple Gaussian Radial-Basis-Function Network as an uncertainty estimator. The proposed network includes a hidden layer with one node, two inputs and a single output. In comparison with other model-free estimators such as multilayer neural networks and fuzzy systems, the proposed estimator is simpler, less computational and more effective. The weights of the RBF network are tuned online using an adaptation law derived by stability analysis. Despite the majority of previous control approaches which are the torque-based control, the proposed control design is the voltage-based control. Simulations and comparisons with a robust neural network control approach show the efficiency of the proposed control approach applied on the articulated robot manipulator driven by permanent magnet DC motors.

Fuzzy segmentation is an effective way of segmenting out objects in images containing both random noise and varying illumination. In this paper, a modified method based on the Comprehensive Learning Particle Swarm Optimization (CLPSO) is proposed for pixel classification in HSI color space by selecting a fuzzy classification system with minimum number of fuzzy rules and minimum number of incorrectly classified patterns. In the CLPSO-based method, each individual of the population is considered to automatically generate a fuzzy classification system. Afterwards, a population member tries to maximize a fitness criterion which is high classification rate and small number of fuzzy rules. To reduce the multidimensional search space for an M-class classification problem, centroid of each class is calculated and then fixed in membership function of fuzzy system. The performance of the proposed method is evaluated in terms of future classification within the RoboCup soccer environment with spatially varying illumination intensities on the scene. The results present 85.8% accuracy in terms of classification.