Sentiment Classification Based on Public Reviews of Different States Towards the Police

Abstract

Sentiment analysis can be referred to as an Opinion Mining which is an exemplary example of machine learning and is a significant element of NLP that is agitated with the persistence of opinion and individuality in a text, which also includes many applications. This analysis not only identifies the opinions but also extracts the attributes for the expressions. This research will enlighten and set forth the facts and statistics about classifiers for sentiment analysis of public opinion towards police working in different states through comments and tweets via Support Vector Machine (SVM). The proposed with the help of NLP classifies the text into positive and negative tweets where processes like stemming and tokenisation are used to remove the unwanted characters and patterns that are neither positive nor negative in the document. Usage of logistic regression is intended to classify the document into sentence level to consider the polarity of the document. The model has indicates to extract the features from the dataset used and extract the positive and negative tweets as to how they feel about the working of the police in their states. Through SVM and Logistic regression we intend to find the accurate score for the analysis that is classified from the input text to the classes of interest which specifies the need to use the keywords mentioned. After training the model results show that Delhi is the one of the top state with less working of the police whereas Chennai becomes the state where the work of the police is appreciated by their citizen’s which is shown with the help of pie and bar chart. Comparing the accuracies of logistic regression and the SVM techniques accuracies, SVM accuracy made out to be better with 55.4% and for logistic regression gives 56.7% accuracy.