Statistical Validation of ACO-KNN Algorithm for Sentiment Analysis

Abstract

This research paper aims to propose a hybrid of ant colony optimization (ACO) and k-nearest neighbour (KNN) algorithms as feature selections for selecting and choosing relevant features from customer review datasets. Information gain (IG), genetic algorithm (GA), and rough set attribute reduction (RSAR) were used as baseline algorithms in a performance comparison with the proposed algorithm. This paper will also discuss the significance test, which was used to evaluate the performance differences between the ACO-KNN, the IG-GA, and the IG-RSAR algorithms. The dependency relation algorithm was used to identify actual features commented by customers by linking the dependency relation between product feature and sentiment words in customers sentences. This study evaluated the performance of the ACOKNN algorithm using precision, recall, and F-score, which was validated using the parametric statistical significance tests. The evaluation process has statistically proven that this ACO-KNN algorithm has been significantly improved compared to the baseline algorithms. In addition, the experimental results have proven that the ACO-KNN can be used as a feature selection technique in sentiment analysis to obtain quality, optimal feature subset that can represent the actual data in customer review data.