PREDICTION AND RECOMMENDATION OF CUSTOMER RESPONSE USING SENTIMENT ANALYSIS

Abstract

With the widespread availability of the internet, human behavior is significantly influenced by individual emotional states and opinions on social media and e-commerce platforms. These opinions, which reflect attitudes, feelings, or sentiments regarding services, products, current events, political issues, are expressed through various customer responses in the form of reviews, image and brain signal. Sentiment analysis, plays a crucial role in interpreting the vast amount of data generated by users on these platforms. Whether through tweets, reviews, or comments, understanding the sentiment behind these interactions is vital for companies looking to enhance customer experience and tailor their offerings. By analyzing sentiments expressed in various customer response, businesses can gain a comprehensive view of consumer feelings and opinions regarding products, services, or brands. Predicting and recommending through sentiment analysis based on customer responses is still important and challenging. There are various problems that need to be addressed and that have not been explored well by the research community. Customer often face the problem of excessive available information through social media platforms. Recommendation systems are deployed to help customer cope up with the information explosion. Recommendation systems have garnered immense interest for applications in e-commerce and digital media. Traditional approaches in recommendation systems include such as collaborative filtering and content-based filtering through these approaches that have certain limitations, such as the necessity of prior user history and habits for performing the task of recommendation. To minimize the effect of such limitation, a hybrid recommendation systems for the movies that leverages the best of concepts used from collaborative filtering and content-based filtering, along with sentiment analysis of tweets from microblogging sites, is proposed. It has been observed that a unimodal framework relies on the strengths of one modality, which limits the system’s overall performance. Multimodal frameworks, on the other hand, integrate two or more modalities to enhance overall prediction performance. Incorporating multiple modalities improves usability by leveraging the strengths of one to compensate for the weaknesses of another. We have proposed a novel multimodal framework for rating prediction of consumer products by fusing different data sources, namely EEG signals and global reviews obtained separately for the product and its brand. The proposed results are encouraging compared to individual unimodal schemes This thesis addresses various challenges in prediction and recommendation systems through sentiment analysis of customer responses. Additionally, it opens new directions in the domain of sentiment analysis. It includes predicting customer satisfaction through the fusion of different modalities and movie recommendations based on microblogging data. The performance of the proposed methods across different experiments and datasets shows the potential of this thesis.