International Journal of Reasoning-based Intelligent Systems (11 papers in press)
A fuzzy knowledge-based inference system for detecting human skin color
by Mehdi Neshat, Ghodrat Sepidname
Abstract: Abstract: One of the most important applications of intelligent systems in medicine is using expert systems to help with the diagnostic-treatment decision making process. Identification of a human has many various methods that one of the best ways is detecting the type of human skin. The processing speed and having high accuracy play an important role in this way. Designing a new Fuzzy Expert System is aimed at for detecting human skin from other materials intelligently. Skin detection is a challenging problem, because certain factors such as brightness change, a complicated background, and objects with colors similar to that of human skin represent the obstacles thereof. The dataset used in this research was adopted from the UCI accredited database having 245057 records, with each record having four fields. The skin is made up of R, G, and B color space segmentation obtained from the facial images of both genders in different age and racial groups. The results achieved show that this system has managed to detect human skin with 80% accuracy using the help of experts knowledge on human skin detection such as dermatologists and digital image processing experts.
Keywords: knowledge-based system; fuzzy logic; intelligent decision; human skin detection.
Multi-objective design optimization of four-bar mechanisms using a hybrid ICA-GA algorithm
by Nejlaoui Mohamed
Abstract: This work presents a novel approach to the multi-objective optimal design of four-bar mechanisms. Three conflicting objective functions are considered simultaneously, i.e., the tracking error (TE), the transmission angle deviation from 90
Keywords: Design mechanism; ICA-GA; Hybrid algorithm; imperialist competitive algorithm; Genetic algorithm.
Community Detection Using Intelligent Water Drops Optimization Algorithm
by Iyad Abu Doush, Saba ElMustafa, Ameera Jaradat, Nahed Mansour
Abstract: Community structure means the existence of densely connected subgroups in the networks. It is a surprising property that appears in complex and naturally constructed networks. We are proposing a novel heuristic approach to the community detection problem. In this paper, the community detection problem is solved using the intelligent water drop heuristic on a group of real life networks. The proposed heuristic succeeded in grouping the nodes in the network into sets of densely connected subgroups. Our approach uses the modularity value as an optimization criterion. The quality of the resulting division in the network was proven using measures like modularity and NMI. The experimental results verify that our algorithm is highly efficient at discovering quality community structure.
Keywords: Complex network ; community detection ; intelligent water drop ; modularity Q ; social network ; community structure ; metaheuristic.
Special Issue on: CIIA'2015 Advances in Computational Intelligence for Big Data
An Improved HBA Metaheuristic
by Bekaddour Fatima, Chikh Amine
Abstract: As simple and effective optimization approach, HBA (Homogeneity Based Algorithm) is one of the recent metaheuristics, proposed to minimize the total misclassification cost of data mining approaches. However, one problem is that HBA does not adopt computational complexity of the used data mining technique. This is due to the way objective function is defined. So, in this paper, we have proposed an improved HBA (IHBA), which is utilizing a modified objective function that compute the computational complexity of the used classification method. We also test several clus-tering techniques as HBA parameters tuning, in order to enhance classifiers performance. We have tested IHBA on different benchmarks and the obtained results show the effectiveness of the pro-posed method.
Keywords: Metaheuristics; Performance; HBA; Optimization; Data Mining.
Rotation-invariant method for texture matching using model based histograms and GLCM.
by Izem Hamouchene
Abstract: Nowadays, the research is interested in informatics systems that process automatically the information without human intervention. Image is an interesting research area due to the growth of the technologies. Thus, very large data are generated. This represents industrial and economic problem. Texture is one of the important and complex field of image processing. As all surfaces of objects are textured in nature, we have proposed a new texture analysis method. One of the key problem in image processing is the rotation. Therefore, the proposed method is robust against rotation. The goal of this study is to construct a model from each texture. After that, the system classifies the query texture based on the extracted texture models. In this work, we applied a recent and efficient feature extraction method called Rotation Invariant Neighborhood-based Binary Pattern (RINBP). The RINBP method extract relative and invariant patterns from the textured image. The proposed system combines between two parts. First, extract the RINBP model from the texture to describe the local variation of the texture. Second, we apply the GLCM method in order to extract statistical measures from the texture. Thus, efficient combination between the model histogram and statistical measures represents an efficient and robust feature descriptor of the texture. In the experiments, we have used the Brodats album database, which is a reference texture database. Experimental parts illustrates the efficiency and the robustness of the proposed system against rotation.
Keywords: Rotation invariance; Model based histograms; Texture matching; Feature extraction; Neighborhood-based binary pattern.
A 0 -1 Bat Algorithm for Cellular Network Optimisation: A Systematic Study on Mapping Techniques
by Zakaria Abd El Moiz Dahi, Chaker Mezioud, Amer Draa
Abstract: Many research efforts are deployed today in order to design techniques that allow continuous metaheuristics to also solve binary problems. However, knowing that no work thoroughly studied these techniques, such a task is still difficult since these techniques are still ambiguous and misunderstood. The Bat Algorithm (BA) is a continuous algorithm that has been recently adapted using one of these techniques. However, that work suffered from several shortfalls. This paper conducts a systematic study in order to investigate the efficiency and usefulness of discretising continuous metaheuristics. This is done by proposing five Binary variants of the BA (BBAs) based on the principal mapping techniques existing in the literature. As problem benchmark, two optimisation problems in cellular networks, the Antenna Positioning Problem (APP) and the Reporting Cell Problem (RCP), are used. The proposed BBAs are evaluated using several types, sizes and complexities of data. Two of the top-ranked algorithms designed to solve the APP and the RCP, the Population-Based Incremental Learning (PBIL) and the Differential Evolution algorithm (DE), are taken as comparison basis. Several statistical tests are conducted as well. Experiments show that the angle modulation and the nearest-integer techniques are the best ones. In addition, results of the evaluation demonstrate that the BBAs based on these techniques still have some shortcomings, but they could outperform the PBIL in 5 out of 13 of the APP instances and achieve results as good as the ones obtained by the DE in 3 out of 12 of the RCP instances.
Keywords: Bat Algorithm; Binary Problems; Mapping Techniques; Antenna Positioning Problem; Reporting Cell Problem.
Special Issue on: ICEST'2015 Information, Communication and Energy Systems and Technologies
Comparison of different methods for text skew estimation
by Darko Brodic, Ivo Draganov, Zoran Milivojevic, Visa Tasic
Abstract: This paper analyzes different methods for the evaluation of the text skew. The comparison is based on a dataset that consist of the printed text samples. These image samples are given in the resolution of 25, 50 and 300 dpi. Tested algorithms show different skew accuracy for different resolution of document images. The method with the smallest accuracy deviation demonstrates benefits over the other methods. Furthermore, this contributes to its robustness in applications.
Keywords: binarization, initial skew rate, moment based method, printed text documents, projection profiles methods, text skew
Special Issue on: ICCMIT'16 Decision-support Systems Based on Intelligent Techniques
Social Network Analysis: Friendship inferred by chosen courses, Commuting time and Student Performance at University
by Lionel Khalil, Marie Khair
Abstract: Our Social Network Analysis (SNA) evaluates the performance of students taking courses with a group of friends versus students used to take courses alone. We evaluate the probability to be friend by comparing the number of courses shared by students with the probability to be assigned in the same classroom randomly based on curriculum constraints. A minimum of courses taken in common is used as a criterion to identify students belonging to a tribe of friends. The main findings are that students in tribes over perform other students by about half point of GPA, and are dropping and repeating fewer courses. Considering student without friends, we measured the impact of the commuting distance on GPA and drop off rate: students with very low GPA and high drop off are mostly students with significant higher commuting time.
Keywords: Social Network Analysis; Friendship; Student Performance; GPA; drop off; commuting time.
Special Issue on: ICEST'2015 Information, Communication and Energy Systems and Technologies
Experimental determination of soil electrical parameters for creation of a computer model of a grounding system for lightning protection
by Rositsa Dimitrova, Marinela Yordanova, Margreta Vasileva, Milena Ivanova
Abstract: The paper presents multifactor experimental studies for determining the apparent soil resistivity and the dielectric permittivity depending on the frequency of the electromagnetic field, the multilayered structure, moisture content and density of the soil. The gravimetric method for considering the soil moisture during the experimental researches was chosen. The received experimental results were statistically processed and a mathematical modeling of the controlled parameters was performed considering the specifics of the examined soil. These analytically obtained results of the dependences would contribute to more precise sizing of the grounding systems and could be used for creation of accurate simulation models for study of wave processes in them.
Keywords: soil electrical parameters; soil resistivity; dielectric permittivity; multilayered soil; gravimetric method; mathematical modelling; grounding system; grounding rod; lightning protection.
An approach to Transformation of Data into Knowledge for Power Control in Smart Homes
by Ivaylo Atanasov, Anastas Nikolov, Evelina Pencheva
Abstract: Internet of Things (IoT) encompasses information and networking technologies which allow connected devices gathering data from their environment to exchange information with network applications. The increased number of diverse devices and the variety of multimodal data make interoperability a challenging task. Synthesis of semantic information from raw IoT data enables sharing of common data models between different applications. The paper presents an approach to modeling semantic annotation for power control in smart homes and then converting it to knowledge. The approach includes context aware models as well as a knowledge base describing behavior of an autonomous agent. The context aware models representing remote device management are formalized and verified using the concept of bisimulation. Temporal logic is used for specifying the agent behavior and reasoning about power control of home appliances.
Keywords: Internet of Things; Semantic annotation; Remote device management; Formal model verification; Weak bisimulation; Autonomous service model; Temporal logic.
Solving Medical Classification Problems with RBF Neural Network and Filter Methods
by Jasmina Novakovic
Abstract: This paper evaluates classification accuracy of radial basis function (RBF) neural network and filter methods for feature selection in medical data sets. To improve the diagnostic procedure in the daily routine and to avoid wrong diagnosis, machine learning methods can be used. Diagnosis of tumors, heart disease, hepatitis, liver and Parkinson's diseases are a few of the medical problems which we have used in artificial neural networks. The main objective of this paper is to show that it is possible to improve the performance of the system for inductive learning rules with RBF neural network for medical classification problems, using the filter methods for feature selections. The aim of this research is also to present and compare different algorithm approach for the construction system that learns from experience and makes decisions and predictions and reduce the expected number or percentage of errors.
Keywords: medical classification problems; classification accuracy; feature selection; filter methods; machine learning; RBF neural network.