Kaggle H&M Multimodal recommends 5% silver solution
Address of the competition:https://www.kaggle.com/competitions/h-and-m-personalized-fashion-recommendations
Run steps:
1 Download the dataset
2 Run code/data_prepare, the main purpose of this step is to generate parquet files, which runs fast
3 Run code/recall_sort, which contains two parts: recall and sort
Some dependent packages need to be installed by yourself
pandas numpy lightgbm pyarrow pickle tdqm
Running requires a large amount of memory, 64G memory is recommended