AnalyticsVidhya organized an online hackathon during the weekend Aug-8,9.
This is the question I tried to answer for this hackathon: How does some of the popular techniques predict on the dataset? I did no feature engineering. I took the data, and ran it on about a dozen models. I did no cross-validation and did no grid-search for hyper-parameters.
Details of the hackathon can be found here
Unsurprisingly, none of the models worked well on the leaderboard. But this can work as a quick-start benchmark code for such data.
The code, in notebook format, can be viewed here