Personalized Recommendation after Classification of Tweets to Predict Depression using Sentiment Analysis

The task of extracting user profiles from unstructured and informal data obtained from social networks is a challenging endeavor. The study aims to construct a user profile utilizing Twitter data, which can subsequently facilitate the provision of individualized recommendations to the user. The study involves the retrieval and categorization of tweets pertaining to mental health, followed by the extraction and standardization of the sentiments conveyed within said tweets. The present study employs a domain-specific seed list for the purpose of categorizing tweets which involves conducting semantic and syntactic analyses to mitigate any potential loss of information along with integration of random forest classifier. Through a meticulous process of categorization and sentiment analysis, the paper proposes a user-friendly system by closely examining the user’s Twitter activity to discern their specific fields of interest. Lastly, the proposed system is evaluated by comparing the performance of classifiers namely KNN and Naïve Bayes.