- Regression for the yield_anomaly of a corn farm dataset. Input : CSV - 2300 rows, 58 predictors
- Classification for astronomical elements (3 classes). Input : CSV dataset with 5000 rows, 17 predictors
- Natural images recognition (3 classes). Input : JPG images of different sizes
Best models for now :
- Regression : Random Forest regression (Mean squared error ~= 0.30)
- Classification : SVM with polynomial kernel and C=10 (performance >99%)
- Image recognition : Neural network Resnet50, to be replaced by Inception Resnet V2 network(https://github.com/tensorflow/models/blob/master/research/slim/nets/inception_resnet_v2.py, https://cran.rstudio.com/web/packages/keras/vignettes/applications.html)