面向在线产品评论数据的有效性建模与测度研究

Modeling and measure research for online product review data

为提高对虚假评论的识别精度并对评论数据的有效性进行准确预测,提出一种面向在线产品数据的有效性建模与测量方法。通过结合基于核主成分的特征提取方法和最小二乘支持向量机对在线产品的虚假评论进行识别,基于排序Logit构建回归模型对量化的评论数据进行有效性判别预测。实验结果表明,该方法在虚假评论识别和数据有效性分析方面效果良好,可以为消费者提供更为精确的消费参考、为商业机构提供更具辨识意义的评论数据,具有良好的应用价值。

To analyze online reviews effectively and provide valuable information to both consumers and companies,this paper proposed data modeling and measure system for online product reviews. Firstly,this paper proposed the identifying method based on the KPCA-LS-SVM（ kernel principal component analysis least squares support vector machine） model for fake reviews problem. Meanwhile,the paper solved the problem of review data validation analysis by ordinal logistic probability model for the problem of review data validation analysis. At last,experiments were conducted on the real dataset. The results show that it not only can effectively classify fake online reviews,but also improve discriminant validity of the data efficiently.