Conversion Prediction Using Multi-task Conditional Attention Networks to Support the Creation of Effective Ad Creatives

Abstract

Accurately predicting conversions in advertisements is generally a challenging task, because such conversions do not occur frequently. In this paper, we propose a new framework to support creating high-performing ad creatives, including the accurate prediction of ad creative text conversions before delivering to the consumer. The proposed framework includes three key ideas: multi-task learning, conditional attention, and attention highlighting. Multi-task learning is an idea for improving the prediction accuracy of conversion, which predicts clicks and conversions simultaneously, to solve the difficulty of data imbalance. Furthermore, conditional attention focuses attention of each ad creative with the consideration of its genre and target gender, thus improving conversion prediction accuracy. Attention highlighting visualizes important words and/or phrases based on conditional attention. We evaluated the proposed framework with actual delivery history data (14,000 creatives displayed more than a certain number of times from Gunosy Inc.), and confirmed that these ideas improve the prediction performance of conversions, and visualize noteworthy words according to the creatives' attributes.

Unbiased Delayed Feedback Label Correction for Conversion Rate Prediction

“…First, the data for the CVR task are often more sparse than the CTR task. Existing research mitigates this problem through multi-task learning [9,14] and pre-training [18]. Second, CVR prediction suffers from selection bias.…”

Ad Creative Discontinuation Prediction with Multi-Modal Multi-Task Neural Survival Networks

“…The operation process for ad creatives generally includes the following: (1) created and submitted to the ad platform, (2) served to users on the platform, and (3) discontinued when they become less effective. There are several studies to support (1) and ( 2) based on ML methods, such as supporting the creation of high-performing ad creatives [1,2,10,11], determining the best bid price [3,12,13], and deciding to whom to serve the ad creatives [5,14,15]. However, to the best of our knowledge, no studies have been conducted on supporting the discontinuation process.…”

“…These operations are essential for business revenue; however, operators' abilities are dependent on their experience. Supporting ad operations is an important topic of study in the field of machine learning (ML) [1][2][3][4][5]. Discontinuing ad creatives at the appropriate time is a crucial operation and can have a significant impact on sales.…”

RevMan: Revenue-aware Multi-task Online Insurance Recommendation

“…Feature sharing in the embedded layer enforces only simple, low-level feature sharing, while static sharing cannot characterize the diversity in task relatedness due to difference in user behaviours across sales scenarios (Yosinski et al 2014). (ii) Most existing online product recommendation systems focus on conversion rate, rather than the total revenue (Kitada, Iyatomi, and Seki 2019;Bai et al 2009;Chapelle et al 2010;Zhang et al 2019). Recommending high conversion-rate products does not necessarily maximize the total revenue, because products with high conversion rates seldom have high prices (Zhang et al 2016).…”

“…Multi-task Recommendation. Recently, lots of works have been proposed in this scope towards solving data sparsity (Ma et al 2018b;Kitada, Iyatomi, and Seki 2019;Pan et al 2019) or modeling multi-objectives (Ma et al 2018a;Ma et al 2019;Hu et al 2018). In (Ma et al 2018b), an embedding share method is proposed to jointly train clickthrough rate and conversion rate.…”