Deep learning for sentiment analysis in textual expressions

In this study, the effectiveness of machine learning and deep learning methods for sentiment analysis in Turkish texts has been investigated. User comments on white goods products were collected from popular e-commerce platforms such as Trendyol and Hepsiburada and analyzed. Using the word cloud method, the most frequently occurring words in the comments were visualized. Positive comments highlighted expressions like "high performance," "good price," "satisfied," and "easy installation," while negative comments included phrases such as "delay," "bad odor," "lack of quality control," and "service problem." These analyses visually represent the overall satisfaction levels of users and the problems they encounter, providing insights for businesses to improve their products and services. Classification processes were carried out using machine learning methods such as SVM, KNN, and Naive Bayes (NB). With an 80-20 training-test ratio, the NB model exhibited the highest performance with a success rate of 94.02%. The SVM model achieved a success rate of 93.42%, while the KNN model achieved 86.75%. The NB model also demonstrated high performance in precision, recall, and F1-score metrics. Deep learning methods such as LSTM and 1D-CNN were utilized, and success rates were examined with different training-test ratios. The LSTM model achieved a success rate of 92.42%, while the 1D-CNN model achieved 91.42%. In conclusion, the Naive Bayes model outperformed other models with the highest success rate. Deep learning methods, specifically LSTM and 1D-CNN, also demonstrated notable performance with high success rates. This study illustrates the effectiveness of machine learning and deep learning methods for sentiment analysis in user comments within the e-commerce sector. Such analyses offer significant advantages for businesses in better evaluating customer feedback and enhancing customer satisfaction.