Project recommendation method based on multi-task training project attribute graph

• Main Authors: YU MENGNI, WANG XINYANG, WANG RENHUAN, PENG HAO, ZHANG JINGYUN, WANG SHUJUAN, WANG XIAOTONG, WANG XIAOCHEN
• Format: Patent
• Language: Chinese; English
• Published: 30.04.2024
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; DATA PROCESSING SYSTEMS OR METHODS, SPECIALLY ADAPTED FORADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORYOR FORECASTING PURPOSES; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE,COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTINGPURPOSES, NOT OTHERWISE PROVIDED FOR

The invention discloses a project recommendation method based on a multi-task training project attribute graph, which comprises the following steps: collecting data, and carrying out word sense disambiguation; constructing a project attribute graph, and performing multi-task pre-training to obtain a recommendation model; obtaining a user historical behavior sequence; and carrying out controllable diversified recommendation based on a time sequence by utilizing the obtained recommendation model and the user historical behavior sequence, including odd art spectrum smoothing and characterization degradation, so as to obtain a recommendation sequence. According to the invention, a more accurate and personalized recommendation result can be provided. 本发明公开一种基于多任务训练项目属性图的项目推荐方法，包括步骤：数据采集，并进行词义消歧；构建项目属性图，进行多任务预训练，获得推荐模型；获取用户历史行为序列；利用获得推荐模型和用户历史行为序列，进行基于时序的可控多样化推荐，包括奇艺谱平滑处理和表征退化，获得推荐序列。本发明能够提供更精确和个性化的推荐结果。