Classification of customer comments in Turkish using text-mining techniques

Today, the rapid development of technology has led people to carry out their buying and selling activities on e-commerce platforms. This has increased the importance of customer reviews written on products on e-commerce platforms. It has helped people to get accurate information about the product or service. It has also allowed sellers to measure and improve customer satisfaction. In particular, companies use various text mining techniques to identify negative reviews as quickly as possible and thus make business improvements. This thesis aims to use text mining methods to analyze customer reviews in the Turkish language and classify them into 32 different classes. Two other machine learning models, artificial neural networks (ANNs) and convolutional neural networks (CNNs), are used for this classification. Both models were first trained on the training dataset and then evaluated on the test dataset. According to the results, 92.34% Accuracy and 92.40% F1-score were obtained for the ANN model. The CNN model performed better with 92.78% accuracy and 92.83% F1-score. Both models achieved very high success rates. The higher performance of the CNN model shows that convolutional structures are more effective, especially on text data.