Under the special circumstances of military college,the original intention of military postgraduate education is to cultivate high-quality blood to be transported to various joints of our army,to improve the ability and quality of the reserve of talent.At present,colleges,and universities train what kinds of talents are required by the army.However,in the process of connecting postgraduates and posts,the allocation cannot be accurately matched for the academic expertise of the postgraduates is not clear.The interview and selection work has no way of starting.This is not only a waste of educational resources,but also reduces the image of military graduate education.Among the recommendation technologies carried by existing recommendation systems,collaborative filtering(CF)is currently the most used and the most effective recommendation method.However,CF technology predicts the user’s future project selection based on many past behaviors or records.The postgraduate graduation allocation is one-time,and there is no corresponding historical allocation record.Therefore,the CF method cannot be directly used for postgraduate recommendation.In addition,the selection preferences of the employing positions play a key role in the selection of postgraduates to be allocated.Therefore,how to analyze and determine the employing units’ preference for postgraduates is also one of the challenges faced by this research.Moreover,due to the problem of information asymmetry,personas suitable for the recommendation algorithm are needed to help employing positions visualize and comprehensively know the postgraduates to be allocated.To iron out the above problems,this paper develops three parts of the research work as response.First,this paper constructs a student allocation data set,based on student training data and student evaluation data.Combined with the results of the AHP questionnaire for postgraduate evaluation,a five-dimensional model of postgraduate evaluation is proposed.The five-dimensional model can describe the comprehensive quality of students in an allround way from five dimensions: ideological and political quality,military quality,professional quality,research innovation and practical ability,and physical and mental quality.In addition,based on the basic data of postgraduates and the five-dimensional model of postgraduate evaluation,a postgraduate persona model is constructed,which comprehensively displays the postgraduates’ situation from the three dimensions of the five-dimensional evaluation model,basic description information,and individual characteristics.The student persona model is the core technology for generating the holographic student persona of each postgraduate.Secondly,this paper proposes a Feature Recognition Algorithm via Dynamic Preference Learning(FRDPL),which aims to establish a user preference model by learning the characteristics of postgraduates in the user’s historical behavior data.FRDPL can effectively identify the dynamic and static characteristics of postgraduates,learn a large amount of user behavior data through machine learning methods,obtain the feature preferences of positions to postgraduates,and continuously update the preference weights in iterations to improve the accuracy of postgraduate feature recognition.Experiments on the postgraduate assignment data set show that the algorithm can use the existing data to learn user preferences,realize the effective identification of postgraduates’ characteristics,and improve the "qualified" rate of postgraduate allocation.Finally,this paper proposes a Knowledge-enhanced Collaborative Filtering Algorithm(Ke CF),which aims to enhance the information in the user preference matrix through the existing knowledge graph,solve the problems of cold start and data sparsity in the collaborative filtering algorithm,to achieve the precise matching of majors and positions.Ke CF can update the preference matrix information in combination with the feedback data of postgraduate allocation,and identify changes of the corresponding relation between majors and positions in the time dimension.In order to obtain a better recommendation effect,this paper adopts the hybrid recommendation method combining Ke CF and FRDPL,proposing a knowledge-enhanced dynamic hybrid recommendation algorithm(Ke DHR),which can realize accurate recommendation from "postgraduates characteristics + majors" to "troop + position + employment preference".The experimental results on the postgraduate allocation dataset verify the effectiveness and superiority of Ke DHR.In addition,this paper designs a postgraduate allocation recommendation system for military academies.The system is equipped with the postgraduate persona model and Ke DHR algorithm,which can provide three core functions of postgraduate statistics,postgraduate evaluation,and postgraduate recommendation,and is highly suitable for the special task about postgraduate recommendation in military academies. |