Weighted neighborhood classifier based on fuzzy neighborhood rough sets and its application to evaluate quality of experience for online gaming

Abstract

To solve the problem that the current Quality of Experience assessment of online gaming cannot reflect the real Quality of Experience of online gaming, we choose multi-dimensional influencing factors to create a dataset of the Quality of Experience of Honor of Kings and propose a weighted neighborhood classifier based on fuzzy neighborhood rough sets for the existing neighborhood classifier shortcomings. First, the concept of fuzzy neighborhood similarity is proposed in this paper. Moreover, weighting the distances through the fuzzy neighborhood similarity can distinguish the differences between samples, which improves the decision-making of the neighborhood classifier. Then, combining the fuzzy neighborhood rough sets and neighborhood classifier can improve the anti-noise performance of the neighborhood classifier. Finally, the proposed weighted neighborhood classifier based on fuzzy neighborhood rough sets has fully experimented on several UCI datasets and Honor of Kings Quality of Experience datasets. Our method is compared with the state-of-the-art methods to demonstrate our method's superiority in classification accuracy under different class noise levels.