新年开工大吉!本周的精选内容如下,请特别关注加粗的内容。
研究推荐系统的9个必备数据集
电影 (MovieLens):http://grouplens.org/datasets/movielens/
笑话 (Jester):http://eigentaste.berkeley.edu/dataset/
图书 (Book-Crossing):http://www2.informatik.uni-freiburg.de/~cziegler/BX/
音乐 (Last.fm):http://grouplens.org/datasets/hetrec-2011/
百科 (Wikipedia):https://en.wikipedia.org/wiki/Wikipedia:Database_download#English-language_Wikipedia
地图 (OpenStreetMap):http://planet.openstreetmap.org/planet/full-history/
开源 (Python Git Repositories):操作源码见 https://github.com/lab41/hermes
编者附---竞赛 (Kaggle):https://www.kaggle.com/kaggle/meta-kaggle
源自:https://gab41.lab41.org/the-nine-must-have-datasets-for-investigating-recommender-systems-ce9421bf981c
应用:
YouTube: Most recommended videos;https://t.co/sstFKLrUEb
Email attachments recommendation from RecSys
论文:
Gilotte et al., "offline A/B testingfor recommender systems", https://arxiv.org/abs/1801.07030,WSDM 2018
Quadrana et al., "Sequence-aware Recommender Systems", https://mquad.github.io/static/papers/2018-seqrec_survey.pdf
Trattner & Elsweiler, "Food Recommender Systems: Important Contributions, Challenges and Future Research Directions", https://www.researchgate.net/publication/320944468_Food_Recommender_Systems_Important_Contributions_Challenges_and_Future_Research_Directions
Trattner et al., "Investigating the utility of theweather context for point of interestrecommendations", https://link.springer.com/article/10.1007/s40558-017-0100-9
观点:
源自:https://medium.com/the-graph/insights-from-an-evening-with-recommender-systems-experts-ab44d677dc5e
RMSE is never an appropriate evluation metric
In most casesImplicit feedback is far more valuablethan explicit feedback
Ratingsthat are not observedare not missing at random
视频:
A LinearReinforcement LearningAlgorithm for Non Stationary Actions:https://www.youtube.com/watch?v=HUabXYYWHYs
Recurrent Neural Network for Session-based Recommendations:https://www.youtube.com/watch?v=M7FqgXySKYk
领取专属 10元无门槛券
私享最新 技术干货