Modeling multi-aspect preferences and intents for multi-behavioral sequential recommendation

Multi-behavioral sequential recommendation has recently attracted increasing attention. However, existing methods suffer from two major limitations. Firstly, user preferences and intents can be described in fine-grained detail multiple perspectives; yet, these fail to capture their multi-aspect nature. Secondly, behaviors may contain noises, most could not effectively deal with noises. In this paper, we present an attentive recurrent model projections Multi-Aspect INTents (MAINT short). To extract target behaviors, propose a projection mechanism for generating preference representations aspects. multi-typed behavior-enhanced LSTM refinement attention mechanism. The filter out noises generate intent different adaptively fuse intents, gated fusion Extensive experiments conducted on real-world datasets have demonstrated the effectiveness of our model.