A Novel Algorithm for Criminal Statistical Processing

Abstract

Social order tendency analysis is studied based on intensive probe to present criminal information research jobs. The phenomenon of “routine fluctuation” and “warning fluctuation” in the field of social public order is extensively studied. An urban crime distribution vector generation algorithm, which based on exponential attenuation, is generated by combining the historic statistic data and with the present one. a Composite Statistic Profiling-Vector Hypothesis-test Algorithm, consist of a system anomaly testing expression and a corresponding referenced anomaly testing expression, is presented to exam the present statistics value set, thus social system anomaly index is obtain. Prototype system and experiment results show its precision and credibility. Introduction As the developments and changes of politics, economics and culture in a certain human society, social order tendency fluctuates correspondingly. To be objectively, change is eternal; however the changes of social order surely affect the social stability and social life. Therefore, how to distinguish the warning changes from routine changes is becoming event important for us. Generally, crime tendency can be abstracted from the comprehensive processing toward many relevant factors, which involve social status, population fluctuation, political reformation, crime quantity, crime type, crime structure, etc. In historical researches, time series analysis algorithm, regression analysis, SVM, Bayes analysis and Markov chain algorithm are used to calculate or forecast the trend of social crime. Though many researches on criminal trend have been achieved, we still find that rare algorithm can effectively process the social order trend for decision makers to formulate the comprehensive method to decrease the criminal rate. In this paper we mainly deal with a novel algorithm which analyze social order trend from criminal statistical data. Attenuation Based Profile Vector In terms of individual, each occurrence of a crime is certainly a (p, 1-p) binary event, which is a common probability event in reality. In addition, generally if a certain event possesses a characteristic that its occurrence is “rare”, a Poisson stream may be used to approximate it. Thus, we can verify that general crime events observe Poisson distribution and urban criminal event observes same distribution, then we may use some values derived from distribution property to depict the urban public order status. Composite distribution. Considering the great diversity of crime types, we have to aggregate some crime types in order to obtain a better statistics distribution. Since that the sum of a series Poisson distribution also observes a Poisson distribution, we attempt to use Poisson distribution means to represent the urban crime occurrence property. Nevertheless, as a profile value which can indicate the implicit rule of urban crime, it must represent the fluctuation rule of present status as well as the affection of historic statistic data.Hence, for each component of profile vector, we process it as follows. 2nd International Conference on Electrical, Computer Engineering and Electronics (ICECEE 2015) © 2015. The authors Published by Atlantis Press 1115 At the first time, we collect the mean value of a certain crime. We note it as w and draw a directed segment to represent it. After a specified time span which is defined in advance, we whirl the previous segment counterclockwise with angle a, the project of it on its original direction is: ) cos( ' ∂ ⋅ = w w . When second measure comes, rotates the historic vector and projects it, combine with the present statistics data to generate new profile vector. ∑ = −