2015 Ali Tianchi Big Data Contest-solution

Introduction to the contest: links

This article records 2015 Ali Tianchi Big Data Contest, some of my code, because currently in the game, only share a naive solution, based on rules, code home on my GitHub: link, below is the code description. If you are interested, please read the code comments without further details.

Description of the Repo catalogue

• Data Storing Data
• preprocess Data preprocessing
• Rule generate submission files according to rules
• Model Training Machine learning model (temporarily not shared)

Code usage Instructions

• Fork Ben Repo, non-GitHub users please click on the lower right corner of theDownlown ZIP

• After decompression, tianchi_mobile_recommend_train_user.csv and tianchi_mobile_recommend_train_item.csv put into the /data/ directory

• Only two steps to get a submission,F1 can reach 7.6%

  • First step, go to /preprocess/ directory, rundata_preprocess.py
  • Step two, go to /rule/ directory, rungen_submission_by_rule.py

• After completing the two steps above, /rule/ a file is generated in the directory tianchi_mobile_recommendation_predict.csv and submitted.

Additional Information

• Pure Python, without any dependencies.

• About the function of the code implementation, in each code file has comments, the code may be written more chaotic, there may be bugs, welcome issues.

• If you want to get higher F1 value, modify gen_submission_by_rule.py This document, add some rules,F1 can reach more than 9% .

• recommended to run under Linux ; on my PC (8 cores), the above two steps took a total of less than 20 minutes.

• On the basis of the rules, do feature engineering, training model, this is the purpose of the competition.

• After entering the second season, please delete this code, not suitable for processing big data.

2015 Ali Tianchi Big Data Contest-solution