Deepseek App Sentiment Analysis
Sentiment analysis of Deepseek app reviews on Google Play Store using Naive Bayes algorithm.
With the rise of mobile app usage, user reviews on platforms like Google Play Store have become crucial data for product evaluation. This research aims to perform sentiment analysis on the Deepseek app reviews to classify user opinions into positive, negative, and neutral categories. The study utilizes text mining stages starting from data collection, preprocessing, feature extraction using TF-IDF, to classification using the Multinomial Naive Bayes algorithm.
My contribution
Engineering decisions
Validating manual sentiment labeling, especially for neutral categories which are prone to misinterpretation, and handling unstructured text data from user reviews.
Implemented a complete preprocessing pipeline (Case Folding, Tokenization, Stopword Removal, Lemmatization) and used TF-IDF for feature extraction. The Multinomial Naive Bayes model achieved an accuracy of 87.67%.