Abstract:

Internet reviews are mostly regarded as a critical element in establishing and sustaining a positive reputation for the businesses. Also, they will make it simpler for end users to make decisions. A favorable review for a specific product typically draws more customers and results in a significant increase in sales. Today, reviews are made on purpose to create a false reputation and draw in more clients. So, our aim is to find false reviews. The ability to spot false reviews depends on both the key characteristics of the reviews as well as the reviewers' actions. This project proposes using machine learning to detect fake reviews. In addition to the review feature extraction process, this project employs several feature engineering techniques to extract various reviewer behaviors. The project compares the results of several experiments conducted on a real Reviews dataset of amazon reviews with and without features derived from user behavior compare the performance of several classifiers in both cases, including Naive Bayes (NB), SVM, and Logistic Regression. During the evaluations, various language models' confusion matrix, f1 score, and precession are also taken into account.

Description:

machine learning, logistic regression, Support Vector Machine, Streamlit, numpy,imutils, pandas, matplotlib, seaborn, textblob and nltk