Improved Spider Monkey Optimization Algorithm to train MLP for data classification

• Prabhat Ranjan Singh ^[1] ; Diallo Moussa ^[1] ; Xiong Shengwu ^[1] ; Bikram Prasad Singh ^[1]
  1. [1] Wuhan University of Technology

     Wuhan University of Technology

     China

     
• Localización: 3c Tecnología: glosas de innovación aplicadas a la pyme, ISSN-e 2254-4143, Vol. 8, Nº. Extra 1, 2019 (Ejemplar dedicado a: “2nd International Multi–Topic Conference on Engineering and Science”), págs. 142-165
• Idioma: inglés
• Enlaces
  • Texto completo (pdf)
• Resumen

  • In this paper, the modified Spider Monkey Optimization (SMO) with Multi-Layer Perceptron (MLP) is utilized to solve the classification problem on five different datasets. The MLP is a widely used Neural Network (NN) variant whichrequires training on specific application to tackle the slow convergence speed and local minima avoidance. The original SMO with MLP experiences the problemof finding the optimal classification result; due to that, the SMO is enhanced by other meta-heuristics algorithm to train the MLP. Based on the concept of nofree lunch theorem, there is always a possibility to improve the algorithm. With the same expectation, the performance of the SMO algorithm is improved byusing Differential Evolution (DE) and Grey Wolf Optimizer (GWO) algorithm to train the MLP. Likewise, the SMO-DE and SMO-GWO are two differentconcepts employed to improve efficiency. The results of proposed algorithms are compared with other well-known algorithms such as BBO, PSO, ES, SVM,KNN, and Logistic Regression. The results show that the proposed algorithm performs better than others or they are more competitive.