Return on Advertising Spend Prediction with Task Decomposition-Based LSTM Model

Return on advertising spend (ROAS) refers to the ratio of revenue generated by projects its expense. It is used assess effectiveness marketing. Several simulation-based controlled experiments, such as geo have been proposed recently. This calculating ROAS dividing a geographic region into control group and treatment comparing in each group. However, data collected through these experiments can only be analyze previously constructed data, making it difficult use an inductive process that predicts future profits or costs. Furthermore, obtain for group, must under new experimental setting time, suggesting there limitation using data. Considering these, we present method predicting does not require acquisition validates comparative experiments. Specifically, propose task deposition divides end-to-end prediction two-stage process: occurrence occurred regression. Through reveal approaches effectively deal with which label mainly set zero-label.