Real time bidding is complex data science task, with multiple failure modes. One can miss-predict the click-rate, mess up targeting, or spend all the budget prematurely. In this talk Doron from Intango would demonstrate how modelling a DSP (bidder) as a reinforcement learning agent could come handy when trying to model budget pacing with bid price prediction.
![](https://s2.save4k.ru/pic/kptxrinMJNc/maxresdefault.jpg)