DengAI: Disease spread prediction (DrivenData Competition)
Achieved a rank of 197 in 7000 participants on the leaderboard(on submission).
Following is a summary of predictions done using different algorithms
Typical results when submitted (MAE)
Neural Network Deep (with multiple hidden layers) neural Networks
24 to 29
Random Forest
Random Forest Regressors with few relevant features
24(approx.)
AdaBoost/Boosted Trees
Different software/algorithms of boosting with all features, one week lagged, or most relevant features
24+
Lasso
Lasso Regressor with few relevant features
24+
Ridge
Ridge Regressor with few relevant features
25 to 27
XGBoost
XGBoost Regressor with tuned parameters
24+
SVR(Support vector regressor)
SVR regressors with few relevant features (kernel=’rbf’ and ‘linear’)
24+
Linear regression (dual)
Dual-linear regression models composed to predict seasonal and trend components of total cases
Below 20,
19.8149(best result till now)