DengAI: Predicting Disease Spread

Anders Poirel, Julian Lehrer

Repository for our work on the DengAI: Predicting Disease Spread competition on Driven Data

• write-up link

⚙ Setup

Requirements:

• conda

Setting up your environment

Run

conda env create -f environment.yml

Activate the environment

conda activate dss-dengueai

Downloading original dataset

To get a fresh copy of the original dataset, go to src/data and run

python get_data.py

⚒ Work

Data exploration:

• Anders notebooks/apoirel-exploration-01.ipynb
• Julian notebooks/jlehrer-exploration-01.ipynb

Gradient boosted models:

• notebooks/jlehrer-models-01.ipynb

Baselines using linear models:

• notebooks/apoirel-models-01.ipynb 01 through 04

Linear model with time-lagged features

• notebooks/apoirel-improved-fe-02.ipynb

Linear model with time-lagged features and predictions using target values with time lag 1

• notebooks/apoirel-models-05.ipynb

FBProphet experiments with more or less models:

• notebooks/apoirels-prophet-01.ipynb 01 through 03

🏆 Results

Current leaderboard position : 692 (top 7.6%) Previous submission position: 1100 (top 12%)