摘要
推荐模型通常根据电子商务用户喜好和评价稀疏数据,预测客户兴趣并推荐相似结果。通过综述现有电子商务推荐算法,提出在SVM模型中,通过训练样本预测评分的方法填充用户-项目矩阵,根据各个项目的重要性对核函数相应的分量赋予相应的权重,以达到提高预测精度、产生理想的推荐效果目的。
Recommend models are usually sparse data, predict customer interest based on user preferences and evaluation of e-commerce and rec-ommend similar results. Through review of existing e-commerce recommendation algorithm, proposes in the SVM model, fills with the us-er through the training sample forecasting scoring method-project matrix, gives appropriate weight based on the kernel function corre-sponding importance of each component of the project, so as to improve the prediction accuracy, be recommended to produce the desired effect purposes.
出处
《现代计算机》
2014年第5期9-13,共5页
Modern Computer
基金
国家民委科研基金(No.12JSZ002)
湖南省教育厅科研项目(No.12C0290)