I enjoyed learning a lot of new business questions. There are a lot of interesting things to explore further if I have more discussion and time. I got highest score with polynomial regression and random forest as compared to linear regression, SVM, ridge and lasso.
I want to try piple-line using decision tree, lasso and elasticnet, or XGBoost etc. Also I need to do feature engineering (interaction, groupby etc) and parameter tuning for further optimization.
Data is clearly small and our business interests need clarity like what we want to achieve; either metric score of which field (feature or column) is usually more important.
--- The work is done as following: