Abstract
If you ask what a statistic student can do, you ask the wrong question. Instead, you should ask: what can't a statistic student do? There are tons of jobs in the Bay Area yearly for statistic-graduated students. You can become a biostatistician in a pharmaceutical company or a data scientist/machine learning engineer in an IC company. The logic and thought process training in the Statistics Department lets you approach the problem from a different perspective and makes you unique in a company.
This talk contains two parts. In the first half, I will briefly introduce the different industry opportunities for statistics students and what life looks like for each position. The second half of the talk is a deep dive into a problem we solved for the real-time bidding area.
Real-time bidding has become one of the primary online purchasing advertising spaces. Advertisers construct campaigns that impose specific targeting criteria on Demand Side Platforms they would like to reach. Forecasting such delivery of these criteria enables advertisers to do a more comprehensive budget and campaign planning. Our forecasting method proposes a new real-time approach to predicting campaign delivery using the sketching technique and modeling methods. We reach this goal through two key steps: we apply an extended KVM sketching method, named Running-Average Sketch, to summarize the information of some dimensions (e.g., Site, Inventory, Geo-Targeting) from the raw bids data, and we build a regression model to incorporate those filter dimensions further to reach the final predictions.