Decoding Authenticity in Online Review using Machine Learning

The popularity of the internet has grown significantly in recent years. In today's e-commerce, online reviews are crucial in helping businesses make decisions. Before choosing where and what to buy, a customer looks through product or store reviews. There has been a sharp rise in fake spam views on online review platforms since creating fraudulent reviews yields large profits. Consumer decision-making has led to an increase in the generation of fake reviews, which can significantly impact businesses and consumer trust. This study provides an extensive and well-known machine learning algorithm to address this problem: Stochastic Gradient Descent (SGD), for the task of fake review detection. Fake_reviews_dataset collected by Joni Salminen and have a license :CC-By Attribution 4.0 International used for experimentation consists of a diverse collection of reviews labeled as genuine or fake. Feature extraction techniques are applied to represent textual reviews as numerical vectors, enabling the algorithms to process the data effectively. The range of performance metrics, including accuracy, precision, recall, and F1score, are used in the implementation, training, and evaluation of the algorithms. In the fake review identification procedure, the accuracy of the Random Forest (RF) method is 80.56%, SVM (Support Vector Machine) has an accuracy of 81%, and KNN (K-Nearest Neighbor) has an accuracy of 71%. In the fake review identification procedure, the SGD (Stochastic Gradient Descent) algorithm has an accuracy of 86.7%, which is 15.7% higher than KNN, 5.7% higher than SVM, and 6.14% higher than RF.