In this short exercise, I will demonstrate how you can create a model for product recommendation. The main aim is to find the important features of the reviews that lead people to make a recommendation. In other terms, it will also show us what are the strengths and weakness of the product, what features make people like or dislike it. This analysis can be also considered a sentiment analysis but for its design, it is a classic text classification model.
I also analyzed the same data from sentiment analyis perspective, using Bing and AFINN dictionaries. If you would like to see my previous post, please check out this page.
Basically, this analysis includes.
- Tokenization.
- Tidy style model and workflow creation.
- Tunning.
- Identifying the predictor importance.
- Model validation.
To see the codes and graphs, please open md file or click here.