Personal Life Pattern Mining Research Based On Periodic Behavior

With the position acquiring technology development, personal position history data isknown as a new spatio-temporal data, and quickly become the research hotspot of experts andscholars. In the personal position history data, the periodicity is the phenomenon whichoccurs frequently. Finding periodic behaviors could better understand the moving behaviorsactivities of personal location history. Besides, personal location history data is generatedindividually and is closely related to person’s everyday life, so we believe that person’s lifestyle and regularity (life pattern) could be found out from his location history. This articleresearch the individual life pattern mining based on the periodic behavior of personal locationhistory by analyzing the personal position history data.For the problem that the behavior period parameter of the traditional life pattern miningalgorithm which only bases on frequent pattern mining algorithms is not adaptive and theproblem of the life patterns’ combinatorial explosion, this article proposes the personal lifepattern mining algorithm. For the complexity of periodic behavior: multiple crosscuttingperiodics, uncertainty partial period span, spatio-temporal noises and outliers etc., we dividethe periodic behavior mining of personal location history into four stages: Firstly, finding staypoints sequence from data sequence of the original location history; Secondly, mining the thesequence of stay point by the density-based clustering algorithm, find out the personalsignificant places(set of important places), such as company, supermarket, home, etc; Thirdly,take each place as a reference point, abstract the original location histority data into binarysequence by the location point in or out a place. Combinating two popular signal processingmethods periodogram and auto-correlation, finds the personal moving periods of every place;Fourthly, mining the periodic behavior of the places with the same personal moving periodsby using the method based on the hierarchical clustering to mining the periodic behaviorbetween different places.Based on the concept of life pattern, by analysising the logicalrelationship between the difference life patterns, we establish the framework of finding thelife patterns from the periodic behaviors of his personal location history.This article researchthe personal life pattern mining based on periodic behaviors of personal location history.The experiment results show that comparing with the traditional model only based on frequent pattern mining, the method this article proposed based on the periodic behavior hastwo advantages: The first is period parameter adaptive—the periodic behavior mining doesnot require additional input period parameter, it can fit the period parameter adaptivelyaccording to the actual data.The second is life pattern mining more accurate and the scalesmaller—by dynamic assessing the representation error of the periodic behavior, determiningthe number of periodic behavior, avoiding the problem of combinatorial explosion result fromthe frequent pattern fitting in high-dimensional space, so the life pattern mining based onperiodic behavior keeps in a precise and limited scope.