/Landslide-Prevention-and-Innovation-challenge

The core of this challenge is to design and shape the future of landslide prevention and management with the example of Hong Kong.

Primary LanguageJupyter Notebook

Landslide-Prevention-and-Innovation-challenge

Description:

This repository contains the code for the Landslide Prevention and Innovation Challenge on Zindi Africa. You can check out the challenge here 👉 Zindi Competetion.

Approach:

For this Challenge, I used two approachs and an ensemble of both, which are included in the following notebooks:

  • Landslide prediction V1.ipynb - This notebook uses the oversampling method for the data imbalance problem. the model used are CatBoostClassifier and LGBMClassifier. It produes the submission1.csv output file.
  • Landslide prediction V2.ipynb - This notebook uses threshold variation for the data imbalance problem. same models as in V1 are used, but with different parameters. It produces the submission2.csv output file.
  • Ensembling Landslide prediction.ipynb - This notebook uses the ensemble of the two approaches above, i.e the submission1.csv and submission2.csv files. It produces the ensemble.csv output file.

Results:

The results of the two approaches are as follows:

  • Landslide prediction V1.ipynb - The model achieved a F1 score of 0.7624.
  • Landslide prediction V2.ipynb - The model achieved a F1 score of 0.7613.
  • Ensembling Landslide prediction.ipynb - The model achieved a F1 score of 0.7652.

Rank on the leaderboard: 3rd place using the ensemble.csv file.

Please Leave a⭐️ to my repo, if you find it useful😊