GTRM: A Top-N Recommendation Model for Smartphone Applications
Nengjun Zhu, Jian Cao · 2017
With the number of smartphone applications (apps) growing explosively, it has a practical significance to provide personalized app recommendations. In this paper, we propose the GTRM, a new recommendation model which builds the top-N app list by optimizing the metric Group-oriented Mean Average Precision (GMAP). GMAP is an extension of the traditional metric Mean Average Precision (MAP) and it measures the precisions of top-N list in terms of the collective positions of related items rather than the position of individual item. Therefore, GTRM can recommend a more reasonable top-N app list by avoiding overfitting problem. The details of GMAP and GTRM are described. Extensive experiments on a real-world app dataset demonstrate the effectiveness of GTRM, and show that GTRM significantly outperforms the compared methods.