Talent Evaluation Model Of Colleges And Universities Based On Optimized Deep Belief Network

Talent evaluation is important information for colleges and universities to select talents and post adjustment.It is great significance to integrate resources and construct talent teams.At present,the shallow machine learning model,such as BP neural network and support vector machine,fails to fully exploit the essential rules of talent evaluation indicators and evaluation results and the evaluation accuracy is not good enough to meet the actual work requirement of colleges and universities.In order to improve the accuracy of evaluation,this paper proposes a talent evaluation model based on deep confidence network and the harmony search algorithm is introduced to optimize the DBN model to solve the problem that the deep confidence network(DBN)is easy to fall into the local optimum due to the random initialization weight and bias during the training process.The main work of this paper is as follows:(1)In order to solve the problem that the basic harmony search algorithm(HS)has slow convergence speed and insufficient local search ability in later phase,this paper proposes a global adaptive adjustment harmony search algorithm,called GAAHS.Learning from the elite learning strategy of particle swarm optimization algorithm,this algorithm can make full use of the best optimal harmony information in the harmony memory to improve the algorithm's learning efficiency.Besides,we also propose a new method to adjust harmony automatedly.In this method,the new harmony adjustment step is determined by the difference between the best optimal harmony and the worst optimal harmony in harmony memory,which can improve the local search ability in later phase and enable the algorithm to converge at the most optimal results more closely.In order to verify the effectiveness of our algorithm,four standard test functions are used to do simulating experiment.The experimental results show that,comparing with the traditional harmony search algorithm and two classical improved harmony search algorithms(GHS and SGHS),the results of GAAHS is closer to the most optimal results and the convergence speed of GAAHS is faster.(2)In order to solve the randomness in initializing weights of DBN,this paper combines DBN with GAAHS algorithm and proposes GAAHS-DBN model.In this model,the reconstruction error function is used as the fitness function of GAAHS algorithm to find a set of better initial weights and offsets for the DBN.Then,combine with the Contrastive Divergence algorithm for training.To verify the effectiveness of optimizing the traditional DBN by GAAHS algorithm,we compare with and analyze the traditional DBN on the MNIST dataset.The experiment results show that our method improve to a certain extent in the convergence speed and accuracy.(3)We establish a talent evaluation model for colleges and universities using GAAHSDBN.In this paper,we first analyze the work of college talents and then combine the literature research method and expert interview method to construct the evaluation criteria system.Then the criteria data is collected and preprocessed and the parameters of the network model are determined experimentally.Finally,a talent evaluation model based on GAAHSDBN is established.In order to verify the validity and effectiveness of the GAAHS-DBN evaluation model,it is compared with the widely used BP neural network,support vector machine(SVM)and the traditional DBN model.The experimental results show that the accuracy of GAAHS-DBN model proposed in this paper improves by 16.4%,7.3% and 3.6% respectively,comparing with the other three models.Besides,the evaluation results are similar to the real world,which can provide a more fundamental basis for the personnel department in universities to make decisions,such as talent introduction,performance appraisal and resource allocation.