Predicting the credibility of user evaluations of products in e-marketplaces with machine learning

With developing technologies, the use of online shopping platforms has increased significantly. The fact that these platforms have a wide range of users increases the impressive power of the platforms. Therefore, the credibility of the evaluations made by users on these platforms may also affect the opinions of users who are considering purchasing these products. The general purpose of this study is to determine the credibility values of online reviews. For this purpose, online evaluations accessed between May 2023 - Jan 2024 were collected. Sentiment analysis was applied on the collected data and the sentiment class of each evaluation was determined. Machine learning models were compared and the appropriate model was trained with the obtained data. The prediction success of the trained model was tested and predictions were made on environmentally friendly products. The findings obtained in this study show the effects of the emotion class of online evaluation on credibility. In the analysis of environmentalism, it was observed that environmentalism had an effect on credibility. Analyzes made on a product basis according to environmentalism status show that there are significant differences on the credibility index.