Feature Optimization in Sentiment Analysis by Term Co-occurrence Fitness Evolution (TCFE)

Abstract

The opinion of a target audience is a major objective for the assessing state of efficacy pertaining to reviews, business decisions surveys, and such factors that require decision making. Feature selection turns out to be a critical task for developing robust and high levels of classification while decreasing training time. Models are required for stating the scope for depicting optimal feature selection for escalating feature selection strategies to escalate maximal accuracy in opinion mining. Considering the scope for improvement, an n-gram feature selection approach is proposed where optimal features based on term co-occurrence fitness is proposed in this article. Genetic algorithms focus on determining the evolution and solution to attain deterministic and maximal accuracy having a minimal level of computational process for reflecting on the sentiment scope for sentiment. Evaluations reflect that the proposed solution is capable, which outperforms the separate filter-oriented feature selection models of sentiment classification.