Mercedes-Benz Greener Manufacturing

Analysis and submissions code for the Kaggle competition. The idea is to predict the time it takes a car to pass testing given different permutations of Mercedes-Benz car features.

Ideas

• Get rid of columns with no useful information (only one level, either all 0's or all 1's)
• Combine complementary columns (maybe even more than 2) into a single factor
  • Complemetary groups of 3:
    • "X111" "X112" "X113"
    • "X130" "X189" "X232"
    • "X136" "X162" "X310"
    • "X136" "X232" "X236"
    • "X186" "X187" "X54"
  • Complementary groups of 2:
    • "X120" "X52"
    • "X128" "X130"
    • "X136" "X54"
    • "X142" "X158"
    • "X156" "X157"
    • "X186" "X194"
    • "X204" "X205"
    • "X232" "X263"
• Remove outliers
• Try not using variables X0 through X8 as they might be features derived from the rest
• If X0 through X8 are used, group long tail features to avoid trouble when new levels are seen on test data
• Try some interactions between variables in linear models
• Use ID column as a feature
• Add vector decomposition features (SVD, PCA, etc.)
• Cluster the data (maybe using k-means) and train a different model on each cluster
• Use an algorithm for tuning the xgboost regressor

Useful links

• https://www.kaggle.com/c/mercedes-benz-greener-manufacturing

Timeline

• July 3, 2017 - Entry deadline. You must accept the competition rules before this date in order to compete.
• July 3, 2017 - Team Merger deadline. This is the last day participants may join or merge teams.
• July 10, 2017 - Final submission deadline.

All deadlines are at 11:59 PM UTC on the corresponding day unless otherwise noted. The competition organizers reserve the right to update the contest timeline if they deem it necessary.