The Prediction of Offender Identity Using Decision-Making Tree Algorithm

Theft of electric bicycle cases is a new type of property crime since 2008 in China. It is difficult to arrest thieves, because it is very hard to find the information about the identity of suspects due to fewer clues on the spot. In this paper, we develop a criminals’ native places prediction model that can assist police offices in mining investigative intelligence in order to identify and catch the offenders based on historical broken cases data about criminals. The sample broken cases data we use has five important attributes, to be more specific, method, time, district, object and sites, which are believed by police to be related with thieves’ native places. Decision Tree algorithm is utilized to create a decision tree in the modeling and prediction of criminals’ native places that corresponds to the sample data. At each level of the tree, Information Gain Ratio is computed for each attribute in order to be chosen as a splitting attribute if it gives us the highest Information Gain Ratio. We also perform different algorithms such as BayesNet, Logistic and NaïveBayes to analyse their performance. Obtained results show that Decision Tree can be deployed with a high degree of accuracy in the prediction of criminals’ native places.