DrivenData Competition & Final project for Machine Learning

DengAI: Predicting Disease Spread

Competition Site

Directory Structure

• shared/: Training code and models shared between project members.
• src/
  • data/: Raw trainging and testing data cs files.
  • arc/: Random forest models with lagging labels.
  • rfr/: Random forest models without lagging labels.
  • rnn/: RNN models.
  • training/: Training code.
  • ensmeble.sh: Shell script for reproduction.
  • arcanin_rf.py: Testing code for random forest models with lagging labels. (preprocessing included)
  • merge_test.py: Testing code for random forest models without lagging labels and final ensemble. (preprocessing included)
  • rnn2221.py: Testing code for RNN ensmeble models. (preprocessing included)
• Report.pdf**
• requirements.txt**
• README.md**

Reproduce Prediction

1. Under current directory, install required python3.6 packages with requirements.txt.
2. Change directory to src/.
3. Make sure shell script ensmeble.sh is executable.
4. Execute ensemble.sh.
5. The final prediction file is generated under src/ and is named ensemble_result.csv. In addition, two intermediate csv files arc.csv and rnn2221.csv will be generated under the same directory.