{"id":13490208,"url":"https://github.com/Arturus/kaggle-web-traffic","last_synced_at":"2025-03-28T06:30:44.103Z","repository":{"id":44406604,"uuid":"111151383","full_name":"Arturus/kaggle-web-traffic","owner":"Arturus","description":"1st place solution","archived":false,"fork":false,"pushed_at":"2022-10-09T07:10:28.000Z","size":9185,"stargazers_count":1828,"open_issues_count":13,"forks_count":667,"subscribers_count":75,"default_branch":"master","last_synced_at":"2024-10-31T03:35:34.481Z","etag":null,"topics":["cocob","cudnn","kaggle","kaggle-web-traffic","rnn","rnn-encoder-decoder","seq2seq","tensorflow","time-series","timeseries"],"latest_commit_sha":null,"homepage":null,"language":"Jupyter Notebook","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/Arturus.png","metadata":{"files":{"readme":"Readme.md","changelog":"Changelog.md","contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null}},"created_at":"2017-11-17T21:15:59.000Z","updated_at":"2024-10-25T07:38:17.000Z","dependencies_parsed_at":"2023-01-19T16:33:34.453Z","dependency_job_id":null,"html_url":"https://github.com/Arturus/kaggle-web-traffic","commit_stats":null,"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Arturus%2Fkaggle-web-traffic","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Arturus%2Fkaggle-web-traffic/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Arturus%2Fkaggle-web-traffic/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Arturus%2Fkaggle-web-traffic/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/Arturus","download_url":"https://codeload.github.com/Arturus/kaggle-web-traffic/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":245984241,"owners_count":20704787,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["cocob","cudnn","kaggle","kaggle-web-traffic","rnn","rnn-encoder-decoder","seq2seq","tensorflow","time-series","timeseries"],"created_at":"2024-07-31T19:00:42.846Z","updated_at":"2025-03-28T06:30:41.897Z","avatar_url":"https://github.com/Arturus.png","language":"Jupyter Notebook","funding_links":[],"categories":["Time Serices Forecasting","Jupyter Notebook"],"sub_categories":["[[2017] Web Traffic Time Series Forecasting](https://www.kaggle.com/c/web-traffic-time-series-forecasting/)"],"readme":"# Kaggle Web Traffic Time Series Forecasting\n1st place solution\n\n![predictions](images/predictions.png)\n\nMain files:\n * `make_features.py` - builds features from source data\n * `input_pipe.py` - TF data preprocessing pipeline (assembles features\n  into training/evaluation tensors, performs some sampling and normalisation)\n * `model.py` - the model\n * `trainer.py` - trains the model(s)\n * `hparams.py` - hyperpatameter sets.\n * `submission-final.ipynb` - generates predictions for submission\n\nHow to reproduce competition results:\n1. Download input files from https://www.kaggle.com/c/web-traffic-time-series-forecasting/data :\n`key_2.csv.zip`, `train_2.csv.zip`, put them into `data` directory.\n2. Run `python make_features.py data/vars --add_days=63`. It will\nextract data and features from the input files and put them into\n`data/vars` as Tensorflow checkpoint.\n3. Run trainer:\n`python trainer.py --name s32 --hparam_set=s32 --n_models=3 --name s32 --no_eval --no_forward_split\n --asgd_decay=0.99 --max_steps=11500 --save_from_step=10500`. This command\n will simultaneously train 3 models on different seeds (on a single TF graph)\n and save 10 checkpoints from step 10500 to step 11500 to `data/cpt`.\n __Note:__ training requires GPU, because of cuDNN usage. CPU training will not work.\n If you have 3 or more GPUs, add `--multi_gpu` flag to speed up the training. One can also try different\nhyperparameter sets (described in `hparams.py`): `--hparam_set=definc`,\n`--hparam_set=inst81`, etc.\nDon't be afraid of displayed NaN losses during training. This is normal,\nbecause we do the training in a blind mode, without any evaluation of model performance.\n4. Run `submission-final.ipynb` in a standard jupyter notebook environment,\nexecute all cells. Prediction will take some time, because it have to\nload and evaluate 30 different model weights. At the end,\nyou'll get `submission.csv.gz` file in `data` directory.\n\nSee also [detailed model description](how_it_works.md)\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2FArturus%2Fkaggle-web-traffic","html_url":"https://awesome.ecosyste.ms/projects/github.com%2FArturus%2Fkaggle-web-traffic","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2FArturus%2Fkaggle-web-traffic/lists"}