论文部分内容阅读
针对众筹项目由于社会影响力不足而成功率较低的问题,提出了面向众筹平台的社交网络影响力预测方法。该方法基于众筹网站和社交网络的实时观测数据,分别提取累积和增量等多类别预测特征,并随着社会推广的进行而渐进地预测项目的社交网络影响力增益,最后采用带L1一范数约束惩罚的逻辑回归等方法进行预测特征分析。实验结果表明:在整个推广过程中,众筹项目的社交网络影响力可以被精确预测,准确率最高达88.31%;分类器在采用累积特征时具有比采用增量特征更好的预测效果;项目统计特征和推广者的社会影响力等特征具有更高的重要性和更稳定的显著性。该方法成功地解决了众筹项目的社会影响力预测问题,并为设计更好的社交网络推广策略提供了依据。
In view of the success rate of crowdfunding projects being low due to lack of social influence, a social network impact forecasting method for crowdfunding platform is proposed. Based on real-time observational data of crowdfunding websites and social networks, this method extracts multi-category forecasting features such as accumulation and increment, respectively. As the social promotion progresses, the social network influence gain of the project is gradually predicted. Finally, Norm constraint punishment logic regression and other methods to predict the characteristics of the analysis. The experimental results show that the social network influence of crowdfunding projects can be accurately predicted with the accuracy up to 88.31% throughout the promotion process; the classifier has a better prediction effect than the incremental features when using cumulative features; The features of statistics and the social influence of promoters are of higher importance and more stable significance. The method successfully solves the social impact prediction of crowdfunding projects and provides the basis for designing better social network promotion strategies.