دانلود رایگان مقاله انگلیسی رویکرد تکاملی تعاملی برای انتخاب ویژگی های چند هدفه - وایلی 2017

Abstract

In feature selection problems, the aim is to select a subset of features to characterize an output of interest. In characterizing an output, we may want to consider multiple objectives such as maximizing classification performance, minimizing number of selected features or cost, etc. We develop a preference-based approach for multiobjective feature selection problems. Finding all Pareto-optimal subsets may turn out to be a computationally demanding problem and we still would need to select a solution. Therefore, we develop interactive evolutionary approaches that aim to converge to a subset that is highly preferred by the decision maker (DM). We test our approaches on several instances simulating DM preferences by underlying preference functions and demonstrate that they work well.

5. Conclusions

Feature selection is an important problem as its results have major impacts on the performance, storage requirements, and computational efforts of learning algorithms. In this study, we have implemented several variations of a preference-based evolutionary algorithm, iTDEA-fs, on the feature selection problem. Noting the special characteristics of the problem, we developed a new preference-based evolutionary algorithm, iWREA-fs. In addition to the traditional objectives defined for the feature selection problem in the literature, we define additional objectives that can be useful within different contexts of the problem.

Feature selection is used in many applications of classification problems. The DM of the problem can be different agencies or customers depending on the scope of the application area. For example, in health care, association of medical doctors, governmental agencies, or patients could be the DM of the problem whose concerns are selecting a set of tests that provides accurate diagnosis while being cost-efficient and/or while minimizing health-related risks involved in the tests. It may also be possible to select several meaningful subsets and then involve the patient in the final decision of which subset to use.