摘要
Abstract
Massive data often contains complex user behavior patterns,item attributes,and their relationships,which often have non-linear characteristics.Traditional generative adversarial network(GAN)may face challenges in nonlinear modeling when processing sequence data.In order to effectively capture the long-short term interest changes of users,enrich the diversity of content,enhance the processing ability and stability in massive data scenarios,a generative adversarial network recommendation algorithm for massive data scenarios is proposed.In the long-short term memory network(LSTM),the user′s behavior patterns towards the data scene are used as input to output the dynamic sequence of long-short term data scenes of interest to the user.The LSTM is combined with GAN to form an L-GAN recommendation model.In this model,the long-short term dynamic sequences output by LSTM are input into the generator of GAN,and false samples similar to real data scenarios are generated by optimizing the loss function.The fake samples are input into the discriminator together with the real data scenes,and the authenticity is identified by means of its objective function.After repeated competition and training,the generator and discriminator can form an accurate recommendation network,so as to finally output a recommendation list of data scenes that meet the user′s interests.The experimental results show that the proposed algorithm can accurately capture the needs of users when processing massive data scenes,and make efficient and comprehensive personalized recommendations.关键词
海量数据场景/生成对抗网络/长短期记忆网络/推荐算法/动态序列/个性化推荐/目标函数Key words
massive data scenario/generative adversarial network/long-short term memory network/recommendation algorithm/dynamic sequence/personalized recommendation/objective function分类
信息技术与安全科学
