xon235 / bigdata

Game Recommendation using Collaborative filtering with K-Nearest Neighbor

Game Recommendation System using Collaborative Filtering with KNN

2020-2 CAU BigData(#50648) Team Project

팀명 : Big Cram
팀원 : 민준홍(팀장) 김경현 이수호 정영진 이혜성

Dataset - Steam Review

• Steam game review 데이터를 직접 크롤링하여 사용
• Google Drive

Environment

• nltk
• scikit-learn
• pandas
• http-server

Resources

• Crawler
  • steam-review-crawler.py - Code to fetch reviews as html format from Steam
  • steam-review-extractor.py - Parse the Steam review files in html format and save them in csv format
  • sentiments.ipynb - Get comment sentiments using nltk library
• Recommender
  • attempt_1_KNN_all_reviews.ipynb - 1st attempt at Item-based CF with KNN using all reviews and binary rating
  • attempt_2_KNN_with_sentiment_all_reviews.ipynb - 2nd attempt at Item-based CF with KNN using all reviews and comment sentiment
  • attempt_3_KNN_with_sentiment_reviews_gt_1.ipynb - 3rd attempt at Item-based CF with KNN using filtered reviews and comment sentiment
  • unused_KNN_reviews_gt_1.ipynb - Item-based CF with KNN using filtered reviews and binary rating, but not mentioned in final report
  • user-based_knn_with_sentiment.ipynb - User-based CF with KNN using comment sentiment
• Visualization
  • KNN_with_sentiment_add_visul1.ipynb - Bubble chart generation code created by modifying Matplotlib unreleased code
  • KNN_with_sentiment_add_visul2.ipynb - Force-directed Graph Using D3.js to Visualize Recommended Results

References

• aesuli's Steam-Crawler
• Matplotlib Packed-bubble chart
• D3.js v4 Force Directed Graph with Labels