Customer Churn Analysis and Prediction

Customer churn is one of the big issues for businesses these days, especially in competitive markets such as telecommunications. This paper introduces a sentiment analysis-based model using Twitter data to determine customer churn. The dataset is obtained through the Tweepy API and comprises thousands of tweets referring to major telecom companies. These tweets undergo preprocessing through various NLP techniques, including tokenization, removal of stopwords, and lemmatization. A lexicon-based approach determines the sentiment labels. The labeled data train an ML classifier based on the XGBoost algorithm. The evaluation metrics accuracy, precision, recall, and F1-score confirm the adequacy of the proposed model. By integrating social sentiment into churn prediction models, telecom providers can gain valuable insights into factors that lead to customer dissatisfaction so they can adopt proactive retention strategies.