I was seeing this case:
and I'm wondering which modeling technique is used and why. Could someone explain that to me?
It looks like, from the document, for this particular dataset that light GBT regressor (the eventual best performing model) and XG Boost were both used, as well as a blender of some other models further down the leaderboard. DataRobot will try a number of modeling techniques, and sort them in terms of performance against the backtest(s). Scrolling down will show all of the models that have been tried, and the model repository will expose even more that you can then run in manual mode.
This looks like a demand forecasting use case. You can access a demand forecasting notebook in the DataRobot docs here: https://docs.datarobot.com/en/docs/api/guide/common-case/demand-forecast/index.html