|کد مقاله||کد نشریه||سال انتشار||مقاله انگلیسی||ترجمه فارسی||نسخه تمام متن|
|382318||660757||2016||20 صفحه PDF||سفارش دهید||دانلود رایگان|
• We propose VbBoPSO and IVbBoPSO for feature selection problem.
• In VbBoPSO, a new parameter Vmin is introduced to explore more regions in the search space.
• In the IVbBoPSO, stagnation is resolved, if exists.
• The scalability of VbBoPSO and IVbBoPSO is superior to BPSO and Boolean PSO.
• Elite features selected are more suitable to classify liver and kidney diseases.
Cancer is one of the foremost causes of death and can be reduced by early diagnosis. Computer Aided Diagnostic system plays an important role in the detection of cancer. Feature selection is an important preprocessing step in the classification phase of the diagnostic system. The feature selection is a NP – hard challenging problem that have many applications in the area relevant to expert and intelligent system. In this study, two new modified Boolean Particle Swarm Optimization algorithms are proposed namely Velocity Bounded BoPSO (VbBoPSO) and Improved Velocity Bounded BoPSO (IVbBoPSO) to solve feature selection problem. Compared to the basic Boolean PSO, these improved algorithms introduce Vmin parameter that makes it more effective in solving feature selection problem. The performance of VbBoPSO and IVbBoPSO are tested over 28 benchmark functions provided by CEC 2013 session. A comparative study of proposed algorithms with the recent modification of Binary Particle Swarm Optimization and Boolean PSO (BoPSO) is provided. The results prove that the proposed algorithms improve the performance of BoPSO significantly. In addition, the proposed algorithms are tested in the feature selection phase of intelligent disease diagnostic system. Experiments are carried out to select elite features from the liver and kidney cancer data. Empirical results illustrate that the proposed system is superior in selecting elite features to achieve highest classification accuracy.
Journal: Expert Systems with Applications - Volume 56, 1 September 2016, Pages 28–47