Sentiment Analysis of Twitter Data of Hepsiburada E-commerce Site Customers with Natural Language Processing

Social media usage has significantly increased, allowing people to freely express their wishes and complaints through these platforms. Consequently, this has caused a significant increase in data, providing information about users, companies, services, and products. However, making sense of this data with human effort is impossible, necessitating various methods. Sentiment analysis is one such method that helps us understand customers’ thoughts on products and companies. In this study, the emotions of e-commerce site users were analyzed using Turkish Twitter data. Text mining techniques were applied to Twitter data, which were analyzed and classified as positive and negative. It was found that 66.9% of the tweets were negative, and 33.1% were positive. The classification results were evaluated using precision, Recall, and F1 criteria. As a result of the evaluation, the sensitivity criterion for negative comments was 94%, and the precision criterion for positive comments showed the highest value with 86%. When looking at the F1 score, 85% for negative comments and 69% for positive comments were calculated. The accuracy rate of the model was found to be 79%.