A brief description of the project: We analyze the trends in Amazon reviews over time to see how the shopping climate is evolving. We trained a classification model using the SKLearn library on a dataset consisting of 21,000 Amazon reviews, half of which were real and half fake. We then tested them on five different classifiers: logistic regression, support vector machine, decision tree, Gaussian Naive Bayes and Multinomial Naive Bayes. The top accuracy after running the data on each of the five classifiers is 65.25% from Multinomial Naive Bayes. Later, we used a large dataset of purely Amazon reviews that didn't have a classifier on whether they were real or fake. We passed each review into the model and it made a prediction on whether it could have been real or fake. Then we graphed the number of fake or real reviews over time using seaborn.
Contributors: Angel, Connor, Kaiyu
Explain any specific project dependancies here.
Explain how to deploy this project. Maybe minimum computer specifications or browser requirements are listed here as well.