metacritic game review analyze project

• Recommendation system
• Sentiment Analysis

score base recommendation system

Surprise, Mongodb, Flask, Scrapy

1. Receive the user score of game

   • score : 0 - 10
   • random pick the game
   • which had been scored at least 30 users

2. Insert user score on data base

3. Train model (SVD)

   • data size ~ 100k

4. Recommend the 6 highest expected score game

   • which had not been scored by user

• ipython note

• receive multiple scores (done)
• make loading page (done)
• make it OOP

review base sentiment prediction

Scikit-learn, sqlite, Flask, Scrapy

1. Train model by scraped data
   • SGD classifier(logistic regression)
   • data size ~100k
2. Receive user comment
3. Predict label
   • positive or negative
4. Receive feedback from user
   • collect or incorrect
5. Insert user comment with feedback on data base
6. update model with new data
   • partial_fit

• note - scikit
• note - doc2vec

• try other model
• try multiple label (sentiment)
• combine with recommendation system