Sentiment Analysis of Text Classification Algorithms Using Confusion Matrix

Abstract

Sentiment analysis on text mining has a vital role in the process of review classification. Text classification needs some techniques like natural language processing, text mining, and machine learning to get meaningful knowledge. This paper focuses on performance analysis of text classification algorithms commonly named Support vector machine, random forest and extreme Gradient Boosting by creating confusion matrices for training and testing applying features on a product review dataset. We did comparison research on the performance of the three algorithms by computing the confusion matrix for accuracy, positive and negative prediction values. We used unigram, bigram and trigrams for the future extraction on the three classifiers using different number of features with and without stop words to determine which algorithms works better in case of text mining for sentiment analysis.