{ "id": 228887, "title": "Floors as a Feature", "url": "/competitions/indoor-location-navigation/discussion/228887", "author_name": "Ravi Shah", "post_date": "2021-03-27T02:33:05.738000", "votes": 13, "comment_count": 0, "views": 0, "content": "

One feature that I have been using that has surprisingly significantly increased my score is floors. While at first I did not use this feature because it is one of the outputs I am predicting, after reading discussions and notebooks, I gave it a try. It definitely helps.

\n

Here is the notebook that produces the floor predictions I use: https://www.kaggle.com/nigelhenry/simple-99-accurate-floor-model

\n

Here is an example of a rnn that uses floors as a feature: https://www.kaggle.com/therocket290/lstm-unified-wi-fi-training-x-and-y-with-floor
\nHere is an example of a lgbm that uses floors as a feature: https://www.kaggle.com/oxzplvifi/indoor-gbm-postprocessing-xy-prediction

\n

Did using floors as a feature improve your results? What other unexpected features should we try?

", "messages": [ { "id": 1253759, "postDate": "2021-03-27T02:33:05.737Z", "content": "

One feature that I have been using that has surprisingly significantly increased my score is floors. While at first I did not use this feature because it is one of the outputs I am predicting, after reading discussions and notebooks, I gave it a try. It definitely helps.

\n

Here is the notebook that produces the floor predictions I use: https://www.kaggle.com/nigelhenry/simple-99-accurate-floor-model

\n

Here is an example of a rnn that uses floors as a feature: https://www.kaggle.com/therocket290/lstm-unified-wi-fi-training-x-and-y-with-floor
\nHere is an example of a lgbm that uses floors as a feature: https://www.kaggle.com/oxzplvifi/indoor-gbm-postprocessing-xy-prediction

\n

Did using floors as a feature improve your results? What other unexpected features should we try?

", "rawMarkdown": "One feature that I have been using that has surprisingly significantly increased my score is floors. While at first I did not use this feature because it is one of the outputs I am predicting, after reading discussions and notebooks, I gave it a try. It definitely helps.\n\nHere is the notebook that produces the floor predictions I use: [https://www.kaggle.com/nigelhenry/simple-99-accurate-floor-model ](url)\n\nHere is an example of a rnn that uses floors as a feature: [https://www.kaggle.com/therocket290/lstm-unified-wi-fi-training-x-and-y-with-floor ](url)\nHere is an example of a lgbm that uses floors as a feature: [https://www.kaggle.com/oxzplvifi/indoor-gbm-postprocessing-xy-prediction ](url)\n\nDid using floors as a feature improve your results? What other unexpected features should we try?", "votes": 12 } ], "comments": [], "raw_markdown_by_id": { "1253759": "One feature that I have been using that has surprisingly significantly increased my score is floors. While at first I did not use this feature because it is one of the outputs I am predicting, after reading discussions and notebooks, I gave it a try. It definitely helps.\n\nHere is the notebook that produces the floor predictions I use: [https://www.kaggle.com/nigelhenry/simple-99-accurate-floor-model ](url)\n\nHere is an example of a rnn that uses floors as a feature: [https://www.kaggle.com/therocket290/lstm-unified-wi-fi-training-x-and-y-with-floor ](url)\nHere is an example of a lgbm that uses floors as a feature: [https://www.kaggle.com/oxzplvifi/indoor-gbm-postprocessing-xy-prediction ](url)\n\nDid using floors as a feature improve your results? What other unexpected features should we try?" } }