Human Resource Recommendation Algorithm Based On Boosting Tree And Neural Network

In recent years,the employment problem has always been a social problem that needs to be solved urgently.With the rapid development of network technology,human resource has further shifted to a network-driven trend,and human resource recommendation faces the problem of information overload.The traditional recommendation algorithms can't adapt well to the phenomenon of information expansion in the era of big data.Numerous job information brings about information review fatigue easily for job applicants.In this way,they can't clarify their job-hunting requirements,spending more energy finding posts that are truly suitable for them.Therefore,in order to improve human resource recommendation,this dissertation proposed an improved human resource recommendation algorithm with the gradient boosting tree and convolutional neural network based on the application scenario of human resource recommendation.The main work of this dissertation is as follows:(1)This dissertation studied and implemented a streaming distributed data collection method for collecting job seekers information,job information and user behavior information.Then,this dissertation combined the characteristics of human resource to perform data preprocessing for the collected data,such as data cleaning,data extracting and data transforming.(2)This dissertation proposed a human resource recommendation algorithm based on the gradient boosting tree and convolutional neural network.Firstly,the gradient boosting tree was used for feature transformation process to complete selection and encoding of the input features.Then,the transformed features were input into the hybrid convolutional neural network designed in this dissertation.The hybrid convolutional neural network used the convolutional operation to perform the high-level feature learning,in order to implement the personalized human resource recommendation.(3)This dissertation proposed some strategies to optimize hybrid convolutional neural network training.The Exponential Linear Unit(ELU)activation function was used to solve the problem of neuron dead-zone;a hybrid pooling strategy was proposed to improve the loss of feature information in max-pooling process;the improved cross-entropy loss function was used to solve the problem of insufficiency in learning samples which were difficult to be classified.The strategies were used to complete the optimization of the model training process and finally improve the recommendation.This dissertation used the actual human resource data as the training set and test set in order to verify the proposed algorithm.The experimental results showed that the proposed algorithm had better recall rate and F1-Score than other recommendation algorithms,further proving the effectiveness of the proposed algorithm.