Building large scale cloud system for product sentiment analysis using genetic algorithm based feature selection

Abstract

In Sentiment analysis, any data driven approach involves changing a piece of text into a feature vector. An optimization scheme of the best-first search which decreases the amount of memory required is referred to as beam search. The possibility of the Beam Search finding the goal can be improvised using a more precise heuristic function as well as a greater beam width. This work covers the local beam search based on feature selection and Genetic Algorithm (GA). A subset of features can be found utilizing the GA where, the bits of chromosomes indicate the presence or the absence of features. Also, for obtaining the best sub-optimal set, the global maximum for the objective function can be discovered. Here, the performance of the predictor is the objective function. As the performance of Support Vector Machine (SVM) in real-world applications is relatively greater than in case of pattern classification, this has been widely investigated in case of machine learning.