| The field of recidivism reform knowledgement are not generally included in the management which of the research is few.Therefore,it has become an urgently needed that how to incorporate the recidivism reform knowledgement into management and form a complete knowledge management system,accurately predict the probability of recidivism and reflect the quality of reform so as to formulate a set of appropriate assistance plan and improve the quality of reform.For this purpose,this thesis based on the expert experience and the questionaire of a prison,combined with ontology technology,machine learning models,fuzzy comprehensive evaluation and others,and the research is carried out from the following three aspects: The construction of knowledge base of recidivism reform field,recidivism reform prediction models and design and implementation of knowledge management system of recidivism reform field.Firstly,this thesis combines the experience of domain experts,the characteristics of the field of recidivism reform and ontology technology to build a knowledge base of recidivism reform.Domain classes,attributes and their relations are defined,domain ontology is visualized,domain knowledge base is stored by relational database which improves retrieval efficiency and reduces maintenance cost.Secondly,based on fuzzy analytic hierarchy process and fuzzy comprehensive evaluation method,the fuzzy comprehensive evaluation model of recidivism is studied deeply,and the weight system of the influencing factors of recidivism is constructed.By comparing the prediction effect of three machine learning algorithms on the ability of recidivism,through experiments,logistic regression is selected as the prediction model,and the results of fuzzy comprehensive evaluation and logistic regression are weighted to get the final more accurate prediction results,which can provide a reference for experts in this field to formulate targeted assistance plans.Finally,the prototype of knowledge management system in the field of recidivism transformation is developed,including knowledge browsing,knowledge retrieval,knowledge management,knowledge maintenance,recidivism prediction,user management and other functional modules.The test results show that each functional module of the system can reach the expected target. |