基于受限玻尔兹曼机的个体行为预测模型的研究  

Research on prediction model of individual behavior based on restricted Boltzmann machine

作　　者：任春霞 李金宝[1,2,3] REN Chunxia;LI Jinbao(School of Computer Science and Technology,Heilongjiang University,Harbin 150080,China;School of Software,Heilongjiang University,Harbin 150080,China;Key Laboratory of Database and Parallel Computing of Heilongjiang Province,Heilongjiang University,Harbin 150080,China)

机构地区：[1]黑龙江大学计算机科学技术学院,哈尔滨150080 [2]黑龙江大学软件学院,哈尔滨150080 [3]黑龙江大学黑龙江省数据库与并行计算重点实验室,哈尔滨150080

出　　处：《黑龙江大学自然科学学报》2018年第6期743-749,共7页Journal of Natural Science of Heilongjiang University

基　　金：国家自然科学基金资助项目(61370222)

摘　　要：个体行为预测能够有效地帮助用户预测最适合自己的运动行为,但在早期的研究中,往往只考虑了个人的历史行为因素和社会相关因素,忽略了用户多样性、动态行为以及隐藏的社会影响,这使目前的个体行为预测问题更加具有挑战性。本文提出了社会受限玻尔兹曼机(Social restricted Boltzmann machine,StRBM)作为一种新的预测模型,该方法将社会影响区分为显性社会影响和隐性社会影响的同时,将时间影响加到了显性社会影响权重上。使用YesiWell数据集以及合成数据集进行了对比实验,验证所提出方法的准确性,证明了提出的StRBM模型比其他基本模型具有更高的预测精度。Individual behavior prediction can effectively help users predict the most suitable sports behavior. However,in the early study,individual behavior prediction only took the individual historical behavior factors and social related factors into account. The user diversity,dynamic behavior and hidden social impact make accurate individual behavior prediction difficult. A new prediction model is proposed based on the social restricted Boltzmann machine( StRBM),which will not only divide the social influence into dominant social influence and recessive social influence,but will consider the influence of time in dominant social influence. In order to verify the accuracy of the proposed method,the YesiWell data set and the synthetic data set are adopted to carry out a comparative experiment. It is proved that the proposed StRBM model has a higher prediction accuracy than other basic models.

关 键 词：行为预测 社会网络 受限玻尔兹曼机 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]