In online advertising, click-through rate (CTR) is a very important metric for evaluating ad performance. As a result, click prediction systems are essential and widely used for sponsored search and real-time bidding.
Use case
Budget-constrained Real-Time Bidding
| Example observations | time stamp banner_pos site_id / site_domain / site_category app_id / app_domain / app_category device_id / device_ip / device_model / device_type / device_conn_type anonymised features C1, C14…C21 (geo, OS, browser, size, etc.) |
| Example actions | 0 = skip (no bid) 1 = low bid (<30) 2 = medium bid (30–80) 3 = high bid (>80) |
| Example reward | click (0/1) – λ · cost, where cost is the price actually paid (payprice) if the impression was won, else 0. λ is set so that the expected return over a day is positive only if clicks are cost-effective. |