Sequential Recommendation Based on Attention Mechanism and Fourier Transform

A sequential recommendation algorithm named SAFTRec, based on the self-attention mechanism and Fourier transform, is proposed. using Fourier transform to filter the matrix output from the embedding layer. Effectively solves the noise problem that exists in sequence recommendation. Concurrently, the implementation of the self-attention mechanism aids in bolstering the sequence data's capacity for feature capture and in elevating the model's expressive capabilities. Rigorous experimentation on an authentic Amazon dataset has been executed, with outcomes indicating that our proposed approach excels over competing methods in terms of Hit Rate@N, NDCG@N, and MRR metrics.