Unbiased Filtering Approach for Handling Ad Clicks in Online Advertising

In this article, we explore a novel approach to predicting accidental clicks on web ads based on dwell-time feature. The proposed method leverages machine learning algorithms to classify clicks into two categories: intentional (IC) and accidental (AC). By analyzing the dwell-time of each click event, we can determine whether the user intended to interact with the ad or not.
To understand how this works, imagine you are browsing a website and an ad catches your eye. You may spend some time looking at it without clicking, which is known as dwell-time. If you dwell on the ad for more than three seconds, chances are you intended to click on it. On the other hand, if you quickly glance at the ad and move on without interacting with it, it’s likely an accidental click.
The proposed method uses a dwell-time threshold of three seconds to classify clicks as either intentional or accidental. Any click event with a dwell-time below this threshold is labeled as accidental, while those above are considered intentional. This approach allows us to identify and filter out accidental clicks, which can improve the overall performance of web ads.
The article references several relevant works in the field of web advertising, including a study on factorizing machines for click prediction [13]. These models provide a good guess for the probability of a click given an impression, which is crucial for accurately predicting accidental clicks. The authors also discuss the importance of dwell-time feature in classifying clicks, as it provides a more accurate indication of user intent than other features such as CTR.
In summary, the proposed method offers a novel approach to predicting accidental clicks on web ads based on dwell-time feature. By leveraging machine learning algorithms and analyzing the dwell-time of each click event, we can accurately classify clicks into two categories: intentional and accidental. This approach can help improve the performance of web ads by filtering out accidental clicks and increasing the accuracy of CTR predictions.