Recommending Insurance products by using Users' Sentiments

TL;DR

This paper explores using sentiment analysis of customer feedback combined with demographic and purchase data to recommend insurance products, demonstrating promising results with simple models despite limited data.

Contribution

It introduces a novel approach of integrating sentiment polarity with user profiles for insurance product recommendations, highlighting potential improvements with better data and methods.

Findings

Sentiment polarity analysis effectively informs recommendations.

Combining sentiment with user profiles yields logical product suggestions.

Simple models showed promising results despite limited data.

Abstract

In today's tech-savvy world every industry is trying to formulate methods for recommending products by combining several techniques and algorithms to form a pool that would bring forward the most enhanced models for making the predictions. Building on these lines is our paper focused on the application of sentiment analysis for recommendation in the insurance domain. We tried building the following Machine Learning models namely, Logistic Regression, Multinomial Naive Bayes, and the mighty Random Forest for analyzing the polarity of a given feedback line given by a customer. Then we used this polarity along with other attributes like Age, Gender, Locality, Income, and the list of other products already purchased by our existing customers as input for our recommendation model. Then we matched the polarity score along with the user's profiles and generated the list of insurance products to be recommended in descending order. Despite our model's simplicity and the lack of the key data sets, the results seemed very logical and realistic. So, by developing the model with more enhanced methods and with access to better and true data gathered from an insurance industry may be the sector could be very well benefitted from the amalgamation of sentiment analysis with a recommendation.