Sentiment Analysis of Online Customer Reviews for Handicraft Product using Machine Learning: A Case of Flipkart

Abstract

Online customer reviews have been recognised as a vital source of market information related to customer preferences and customer experience. In the e-commerce era, online review content helps customers for developing purchase intention and purchase decision. However, revealing meaningful insights from the large volume of online reviews is the major challenge faced by majority of customers. Hence, extracting the online customer reviews and analyzing such online database are crucial pace for developing understanding of customer preferences and customer experience. Analysing online customer review also helps entrepreneurs to develop new products and understand their product in the customer preference perspective. The online customer comments can be segregated into three classes such as positive, negative and neutral. Employing classifiers give signals to the new customers regarding for the particular product. This paper classifies the online review dataset into the positive, neutral and negative based on the frequency of the words associated with respective sentiments. The bag-of-word is constructed using online customer review from Flipkart. This paper employed classifier algorithms such as logistic regression, K nearest neighbours, Multilayer perceptron and Support vector analysis to segregate the online comments as positive, neutral and negative online reviews. In the current research, 28995 online customer reviews for handicraft products are extracted from Flipkart using Python. This research aims to understand the perception of the customers for handicraft products in e-commerce platform. This paper employed machine learning algorithms such as logistic regression, KNN, Multilayer perceptron (MLP) and Support Vector Machine (SVM) and simulated by using Python. Accuracy of LR, KNN, MLP and SVM is also estimated using TF-IDF and Count Vectorizer. With respect to TF-IDF and count vectorizer, the accuracy of the SVM is the higher than the KNN, MLP and LR.