/WBDC2021-Rank70-Solution

Primary LanguagePython

2021**高校计算机大赛-微信大数据挑战赛

队伍名:Pogo

Rank：68（初赛）/70（复赛）/6768

1. 环境依赖

  • Python 3.6.5
  • tensorflow-gpu==1.13.1
  • pandas>=1.0.5
  • numpy>=1.16.4
  • numba>=0.45.1
  • scipy>=1.3.1
  • deepctr[gpu]==0.8.6
  • catboost==0.26
  • lightgbm==3.2.1
  • gensim
  • tqdm

2. 目录结构

./
├── README.md
├── requirements.txt, python package requirements 
├── init.sh, script for installing package requirements
├── train.sh, script for preparing train/inference data and training models, including pretrained models
├── inference.sh, script for inference 
├── src
│   ├── prepare, codes for preparing train/inference dataset
|       ├──extract_history_feed_embeddings.py
|       ├──extract_feed_embeddings.py
|       ├──extract_feed_info_features.py
│   ├── model, codes for model architecture
|       ├──wide_and_deep.py  
|   ├── train, codes for training
|       ├──mmoe.py  
|   ├── inference.py, main function for inference on test dataset with mmoe model
|   ├── inference_lgb.py, main function for inference on test dataset with lightgbm model
|   ├── evaluation.py, main function for evaluation 
├── data
│   ├── wedata, dataset of the competition
│       ├── wechat_algo_data1, preliminary dataset
│       ├── wechat_algo_data2, semi dataset
│   ├── submission, prediction result after running inference.sh
│   ├── model, model files (e.g. tensorflow checkpoints)

3. 运行流程

  • 安装环境：bash init.sh
  • 进入目录：cd /home/tione/notebook/wbdc2021-semi
  • 数据准备和模型训练：bash train.sh
  • 预测并生成结果文件：bash inference.sh ../wbdc2021/data/wedata/wechat_algo_data2/test_b.csv

4. 模型及特征

  • 模型：MMOE
  • 参数：
    • batch_size: 8192
    • emded_dim: 192
    • expert_dim: 128
    • num_epochs: 7
    • learning_rate: 0.1
  • 特征：
    • dnn 特征: userid, feedid, authorid, bgm_singer_id, bgm_song_id
    • linear 特征：videoplayseconds, feed_embedding, manual_tag_embedding, history_feed_embedding

5. 算法性能

  • 资源配置：2*P40_48G显存_14核CPU_112G内存
  • 预测耗时
    • 总预测时长: 478 s
    • 单个目标行为2000条样本的平均预测时长: 32.1186 ms

6. 代码说明

模型预测部分代码位置如下：

路径 行数 内容
src/inference.py 139 pred_ans = train_model.predict(test_model_input, batch_size=batch_size, verbose=1)