Customer Sentiment Analysis and Prediction of Insurance Products’ Reviews Using Machine Learning Approaches

Abstract

The study’s goal is to analyse and predict customer reviews of insurance products using various machine learning techniques. We gathered consumer rating data from the Yelp website and filtered the initial data set to only include insurance reviews. Following cleaning, the filtered summary texts were graded as positive, neutral or negative sentiments, and the AFINN and Valence Aware Dictionary for Sentiment Reasoning (VADER) sentiment algorithms were used to rate those sentiments. Furthermore, the current investigation employs five supervised machine learning approaches to divide customer ratings of insurance companies into three sentiment groups. The results of the current study revealed that the majority of customer reviews for the insurance products were negative, with the average number of words with negative sentiment being higher. In addition, current research discovered that while all of the approaches (decision tree, K Neighbours classifier, support vector machine (SVM), logistic regression and random forest classifier) can correctly classify review text into sentiment class, logistic regression outperforms in high accuracy. We analysed and predicted customer review messages using a variety of machine learning methods, which could help companies better understand how customers respond to their products and services. As a result, companies can learn how to use machine learning methods to better understand the behaviour of their customers.