Certain improvements in multi label classification algorithm using group search optimization strategic adaptation technique

Abstract

Hierarchical Multi-label Classification (HMC) is a complex classification problem where the classes are hierarchically structured. This task is very common in protein function prediction, where each protein can have more than one function, which in turn can have more than one sub-function. The main aim of this investigation is decreasing data dimensionality with the K-Nearest Neighbor (KNN) method and also predicting class value based on similar training data using Expectation Maximization (EM) algorithm the output of which is used for predicting the instance. In this work, proposed a novel global HMC with Probabilistic Clustering (HMC-PC) method. HMC-PC works according to the following assumption: instances that belong to a given cluster have similar class vectors, and hence the training instances are clustered following an EM scheme, and the average class vector of the training instances from a given cluster is used to classify new unseen instances associated to the same cluster. The cluster membership probabilities are also used to tune the average class vector in each cluster. Results show that HMC-PC achieves superior or comparable results compared to the state-of-the-art method for HMC. Optimization problems in the real worlds are extremely tough. There are several applications which have to deal with the Non-deterministic Polynomial (NP)-hard problems. In this work, the HMC-PC is proposed using Group Search Optimizer (GSO) method. This GSO algorithm has members in their population that generate effective strong and efficient behaviour from the entire population generating an effective and strong solution for the problem of optimization in a large search space. newline newline newline