This project was both part of an academic work and a contest created by the Rakuten group.
The objective was to find a solution to identify review that help other users in their decision to buy an article.
To achieve this I used machine learning based on the Random Forest and SVM model.
A model of predicting helpfulness in reviews was developed with an accuracy of 65% on Priceminister reviews. The most important features were review length, polarity score, TFIDF and readability score
The complete report can be found here: Prediction of Priceminister’s products reviews interests
- review length
- stars rating
- writing style
- sentimental analysis
- TFIDF
- readability score
During my work I drew different graph to visualise the distribution of the data and help me making the right choice in the development of the model.