Analysis Of Trajectory Of Individual Behavior Based On Threshold Limit Method

There are repeating and identifiable routines in every person’s life。We describe thebehavior of a person’s characteristics and predict his behavior in the near future byresearching these patterns of behavior. This study plays a vital role in evaluating the linkbetween social network of a population. Predicting human behavior and the application arewidely used in many fields, especially in the all courier and some other crowdsourcingsystems. This paper will convert predicting human behavior into classifying human behavior,that is to determine whether the user will go to a fixed place by knowing the user’s data of aday.Classic behavioral classification algorithms, such as SVM, decision tree, neural networkalgorithms, are lack of data filtering and screening, so that during the parameters inversionalgorithm, these algorithms are accustomed to using all the data as input data. However,through the analysis of the data, we know that not all data are suitable as input data sample, sowe are trying to evaluate the trainability of behavioral data and use two evaluation measures.Based on these two measures,we screen and filter the data of an individual’s behavior, extractthe samples which can represent the whole data set as training data set for the followingalgorithm. In this context, this paper proposes a method of threshold limit for the analysis ofindividual behavioral trajectory.The data set we used that is composed of more than100participants’ data, eachparticipant’s data also includes more than50attributes, such data is very huge, so it isnecessary to preprocess the data set. In this paper we use the attribute subset selection methodof data reduction to filter the data’s attributes. We filter the incomplete and unreliable data,delete the attributes that are irrelevant or weak related to the purpose of this paper, then wecan get the data that is meaningful and related to the purpose of this paper. After datareduction, the attributes of the data are still needed further analysis and process. We analysisand integrate the attributes that are directly related to the location information by setting theactive value, so we can get the data that the algorithm can directly process. Thus when we runalgorithm to process the data, we can get results more quickly and at the same time we can getmore accurate results.The threshold limits method processes and filters the data twice, the threshold value isadded each time during the filtering process, so we can get high-quality sample nodes, therebywe can provide high-quality training data samples for the next prediction model. After we getthe high-quality sample nodes, we use artificial neural network to predict the node samples which have screened. When we set the reasonable threshold, we may find that thepredictability of the data upgrade from71.2%to95.9%although the reasonable data down toabout20%of the original. We can set threshold as needed to meet the actual needs in theactual scenes. The experimental results show that the effect of setting the threshold CloseTimeand PositionChange to improve the predictive accuracy of the algorithm is very obvious.