Publications

Abstract Quantitative software engineering is aimed at designing models describing software processes and products. While being noticeably dominated by statistical regression models, this area also embraces advanced techniques of computational intelligence and knowledge-based engineering including rule-based models, fuzzy models, and neural networks. The rationale behind their usage in the setting of Software Engineering is threefold: (a) the underlying distributions of datasets may not adhere to general assumptions that are usually made in the setting of linear regression models: (b) it is beneficial to build interpretable models that in some sense are user-friendly; (c) the models should be advanced and computationally appealing so that they exhibit some nonlinear characteristics as well as are fully equipped with learning abilities. The proposed architecture comprises alogic-based skeleton (blueprint) and an array of generic perceptrons as being commonly encountered in neurocomputing. The hybrid nature...

Information granularity augments a variety of schemes of representation of time series, helps quantify the quality of models of the series and supports a thorough analysis of their parameters. This study introduces a concept of a granular representation of time series. We show that information granules formed on a basis of a given original numeric representation of the series can be optimized through a process of allocation (distribution) of information granularity being regarded here as an essential design asset. We formulate an optimization criterion and utilize a Particle Swarm Optimization (PSO) as an optimization vehicle to distribute a predefined level of information granularity. An optimization criterion used in the formation of the granular representation scheme is concerned with expressing and maximizing coverage of available temporal data by their granular representation. Experimental results in which we focus on the Piecewise Aggregate Approximation (PAA) offer details of the optimization of the allocation of granularity completed for some synthetic and real-world time series and quantify the performance of the resulting granular schemes of representation of time series.

This article investigates group decision-making (GDM) problems, where the decision makers' (DMs) preference information is represented by incomplete interval-valued intuitionistic fuzzy preference relations (IVIFPRs). First, a multiplicative consistency property and an acceptably multiplicative consistency property for IVIFPRs are offered. Then, an optimization model to estimate the missing values in an incomplete IVIFPR is constructed. Subsequently, two optimization models are, respectively, established to derive a perfectly consistent IVIFPR and an acceptably consistent IVIFPR from a given inconsistent IVIFPR. Furthermore, a model is offered to gain the DMs' weights. Afterward, the consensus index is defined. When the consensus for IVIFPRs is unacceptable, a model is presented to reach the consensus requirement. Moreover, a novel GDM method for incomplete IVIFPRs is presented. Finally, the presented method is applied to an illustrative example that shows the feasibility of the offered method.

Energy saving becomes a central issue in the design of wireless sensor networks routing algorithms. The objective of this study is to maximize the survival time of wireless sensor networks routing with a sink node by developing an efficient routing algorithm based on the elite hybrid metaheuristic optimization algorithm. The proposed algorithm comes as an original method that innovatively brings together the global search abilities of the particle swarm optimization algorithm, difference operator of differential algorithm and pheromones of ant-colony optimization algorithm in order to avoid local search and retain diversity of the population. As a result, the method quickly finds an optimal solution. A novel routing algorithm based on the proposed elite hybrid metaheuristic optimization algorithm is designed. Comprehensive simulation studies show that the proposed algorithm can increase maximum lifetime of wireless sensor networks by 38% in comparison with the results being produced with the state-of-the art algorithm routing algorithms based on other population-based optimization algorithms.

Fuzzy models are constructs relying heavily on a qualitative domain knowledge and diverse optimization techniques. What makes them different from other models is their inherent embedding in the context of nonnumeric set or fuzzy set-oriented information. One can also look at the development of the fuzzy models from the perspective of data mining-a prudent and user-oriented sifting of data, qualitative observations and calibration of commonsense rules in an attempt to establish meaningful and useful relationships between system's variables. Having accepted this point of view, we analyze various methods of fuzzy clustering and make them uniform enough so that they can constitute a viable design platform. Several fuzzy clustering methods (especially Fuzzy C-Means) have been already exploited in the context of fuzzy modelling. Our claim is that these methods need some conceptual shift that makes them possible to cope with a notion of "directionality" of any model, namely its ability to determine the values of the output variable(s) given the actual values of the inputs and state variables. This aspect of directionality along with the assumed specificity of modelling, is addressed in depth and leads to a series of detailed algorithms.

Transfer learning has emerged as a solution for the cases where little or no labeled data are available in the training process. It leverages the previously acquired knowledge (a source domain with a large amount of labeled data) to facilitate solving the current tasks (a target domain with little labeled data). Many transfer learning methods have been proposed, and especially fuzzy transfer learning method, which is based on fuzzy systems, has been developed because of its capability to deal with the uncertainty in transfer learning. However, there is one issue with fuzzy transfer learning that has not yet been resolved: the domain selection problem, which is heavily depended on the knowledge transfer method and the applied prediction model. In this work, we explore the domain selection problem in TakagiSugeno fuzzy model when multiple source domains are accessible, and define the similarity between the source and target domains to provide guidance for the domain selection. The experiments on synthetic datasets are designed to simulate the situations of multiple sources in transfer learning, and demonstrate the rationality of the proposed similarity in selecting the source domain for the target domain. Further, the real-world datasets are used to validate the proposed domain adaptation method, and verify its capability in solving practical situations.