Research On Human Behavior With Temporal Preference

Human behavior involves many social,technological and economic phenomena,it’s necessary to analyse human behavior.Because of the rapid advance of information technology,more and more behavior on the internet is recorded and scholars try to analyze the behavior patterns of users.Many cases of user behavior show the long tail phenomenon: the behavior impact of individual or overall appear power law distribution as time goes on.In this paper,the temporal preference model of human behavior can explain the the long tail phenomenon accurately,short-term memory has great effect on human behavior as well.Temporal preference model is based on the theoretical of "the more we do recently,the more likely we will do it next" which can corresponds to human behavior in real life appropriately,including online shopping,travel,e-mail,online games etc.While the time series data meets requirements to the time preference model,in terms of theory,author come up with temporal preference feature and sliding window sample,which provides guidance for the feature and sample when modeling human behavior according to prediction;In terms of algorithm,inspired by the stack model and its application in the deep neural network,author come up with diff-feature stack ensemble algorithm,which features are separated by business dimensional and stacked ensemble each feature dimension’s submodel.Experimental analysis of the shopping and travel behavior data.In the analyse of shopping data,building e-commerce recommendation model according to the big data and algorithms,mining rich content behind data,recommending suitable products content for mobile users at the right time and right place.In the analyse of travel data,exploring the user habits while travel by public transport,mining trip rule at different time periods,different weather conditions and different user types,providing guidance on travel control,line selection and vehicle scheduling,constructing safe and comfortable travel environment.Research on human behavior with temporal preference,author ranked 2nd(out of 7186 teams in total)on Ali mobile recommended competition and 5th(out of 1881 teams in total)on Guangdong traffic prediction competition.