| The arrival of the big data era,puts more challenges for the investigation,trial and qualitative cases,using the existing machine learning method,accurately and reasonable prediction of the new criminal case results,and reduce the occurrence of miscrosis cases Come and more important.The article is based on how the two criminal types(intentional murder and overpaged death sin)in criminal cases are used as an example.Select American data samples as experimental data.It has established criminal types to Murder or Manslaughter.The predictive model and decision system of Manslaughter can provide the theoretical basis and system realization of the exploitation of my country’s criminal case auxiliary decision-making.The main work made by this article has the following three points:(1)The preprocessing of 638454 sample data was completed,14467 training set data and4823 test set data were obtained.Combined with the actual situation of the criminal record data,the objective requirements of the decision-making model for the data were analyzed,and corresponding strategies were adopted to process the data.Firstly eliminate outdated data to achieve data cleaning;then according to the actual situation,use mode or mean to fill or group to fill missing values;use label coding to achieve feature conversion;resample data;use Featexp module to achieve data dimensionality reduction and final Data splitting work.(2)The design and creation of the decision-making model have been completed,and the prediction accuracy rate has reached 96.10%.The article uses the CART algorithm,RandomForest algorithm,and Ada Boost algorithm to create decision-making models,and achieves the goal of assisting decision-making by case types.The accuracy of each model has reached 89.07%,90.77%,and 96.06% respectively;the influence of each feature on the target is obtained.The situation facilitates the analysis and application of biased features in the future;finally,the above three algorithms are integrated based on the Stacking strategy,once again increasing the accuracy of prediction to 96.10%.(3)There have completed the design and implementation of the auxiliary decision-making system for criminal cases.Based on the RandomForest decision-making algorithm,using the B/S system architecture,and using technologies such as Java+SpringBoot+BootStrap+MySQL+IDEA,an auxiliary decision-making system with criminal record management,crime type decision-making and authority log management functions is realized.Finally,according to the experimental results and the test and use of the decision-making system,it can provide reference suggestions for the research in the field of case-type assistance decision-making in our country,and summarize the shortcomings in the research process.It is necessary to assist decision-making research in criminal cases Make further improvements. |
