Project on predicting West Nile Virus outbreaks in Chicago
Project is based on a Kaggle competition where three years of mosquito trap data was provided and the goal was to determine when and where West Nile Virus would be detected.
Data was used to create probabilities that traps would test postive for west nile virus during certain parts of the year.
Our team took a five step approach to the model:
- Identify the Problem
Identify probablity that west nile would be present at trap locations across Chicago at diferent times of the year.
- Obtain the Data
The dataset was provided from a Kaggle competition and contained three years of data in the train dataset as well as two years of data in the test dataset.
- Clean the Data
- Merge Weather Data
- Create Model
- Grid Search Base model: Log Reg or ADABoost
- Evaluate Model
- ROC-AUC Score of 0.77