A challenge was in selecting which features to use for a given user. I assembled a predictive model based on these features. Model based recommendations- *Games that share characteristics with games you like are.* Games in the database are tagged with features such as the game designer, themes, or rules mechanisms. Collaborative filtering at the user level- *Users like you also liked.* Rather than make recommendations based on the similarity between games a user likes, instead find users that are similar to the target user, and recommend games that these similar users like but the target user has not rated. ', I leaned on the () in addition to the Yelp project from before. There is extensive research on collaborative filtering and feature selection for models. Item-based collaborative filtering builds on the () we previously used with the Yelp restaurants dataset. Collaborative filtering (CF) at an item-level- *If you liked game X, you may also like.* This asks which games in the dataset are most similar, picks the top-rated games for a user, and makes recommendations based on which games are most similar to those.

'My goal is to make predictions about which boardgames users of the website might like based on their ratings history. There was formerly a black-boxed () section, but it was widely regarded in the site's forums as useless, and has since been shut down. ', However, the website does almost nothing to use its data on a user's preferences to make personalized recommendations. General statistics are readily presented on the site, such as an overall ranking of the () or () of user-logged play-counts and ratings. It functions as a user- and administrator-curated database of boardgame data, as a social network for gamers, and as a journal tracking game ratings and comments by users.

' Boardgame Geek (BGG) is a huge site with 500 million pageviews and 45 million returning visitors in 2012 (()). 'The website associated with this project can be found ().

Download dungeon bowl bgg