A Semi Supervised Approach to Fake Product Review Detection

As the E-commerce system continues to evolve progressively, maintaining a very good reputation is essential. Hence with regards to these online surveys show a majority role in maintaining and building a strong E - commerce platform. Online reviews play the role of decision making for the consumers at the receiving end of the E - Commerce network. Experiences, suggestions, and opinions on the variety of products available in the market is given through online reviews. A small variation in the way the review is expressed can bring about a positive or negative impact on sales, brand value, reputation of the business etc. There are high chances where falsification of reviews can also take place with the intended motive to bring down the reputation of a targeted brand, organization, or e-commerce platform. Hence detection and classification of reviews based on genuinely is very much essential. This paper uses a semi- supervised machine learning methodology in which traditional ML algorithms like naive bayes and random forest classifiers are used to solve the above stated problem faced by the E- commerce industry. An existing Food review dataset by Amazon has been used to analyse, extract, and interpret diverse review behaviors.