Key points as apart of Exploratory Data Analysis:
- Collect data from source in csv or excel format or in a database.
- Make sure the customer and product id are well maintained in the table (perform cleansing on null rows).
- Make the list of most frequent items association rule with conf 0.02 ( standard form) (perform market basket analysis using apriori or ftp growth alogirthm.)
- Make a datalist with the query having the list of customer and recommended items which are associated strongly with the items that have been purchased by the customer before.
- Now we have a list of customer and recommended items.
- Now, make a list of products along with sentiment polarity value which has bad feedback from customer (value less than 0.1).
- We shall filterout that items from the previous build datalist of recommendation.
Whatever I have shared is totally my own way to build a product recommendation system. If you have any different approach, please share in the comments section below.