Click prediction in mobile display advertising based on HTML5 features

The paper describes an approach to click prediction in mobile display advertising based primarily on ad features. The increasing number of mobile display ads is based on HTML5 web standard. The openness of this standard makes it easier to extract some additional ad features, such as animation, video, sound and other components. We suggest using these features in conjunction with global image features, user features and publisher features to improve click prediction. We have applied a click prediction algorithm on real world data, collected during mobile advertising campaigns, which included more than 1000 ads. We present the results of our experiments and suggest further work in dynamic optimization of mobile ads, based on the findings.