Logistic Regression Based on Artificial Fish Swarm Algorithm with T-Distribution Parameters
Peiran Leng, Liya Xiang, Yihan Lin, Wei Xiao, Zan Yang, Dan Li, Wei Nai · 2020
Logistic regression is to establish a cost function in the face of a regression or classification problem, and then iteratively solve the optimal model parameters through optimization method, and then test and verify the quality of the model we solve. It is the primary approach to data classification in the field of big data and machine learning. The traditional logistic regression uses gradient descent method to solve the optimal parameter of the missing function. However, swarm intelligence algorithms, such as artificial fish swarm, can replace the traditional penalty function method to some extent to ensure the global convergence of the optimization algorithm. Artificial swarm intelligence algorithm can not only realize simple iterative optimization, but also has simple environmental adaptability and system self-regulation. In addition, guided random parallel global search makes the algorithm not easy to fall into the local optimal solution and get the global optimal solution. In this paper, the artificial fish swarm theory in the field of swarm intelligence is applied to solve the optimal parameter of the loss function and overcome the limitation of the gradient descent method to the loss function. In this paper, logistic regression and artificial fish swarm are adopted to deal with the actual data classification problem, and the validity of the algorithm is verified.