Feature Regularization and Deep Learning for Human Resource Recommendation
Haoxiang Wang, Guihuang Liang, Xingming Zhang · IEEE Access · 2018
A novel recommender system is proposed in this paper. It has been implemented for human resource recommendation and achieved improvement on different evaluation metrics. The algorithm leverages both gradient boosting tree model and a convolutional network-based deep learning model for feature regularization and recommendation. The optimizations of activation function and pooling strategy in the proposed network model have been investigated for mitigating the problems of the gradient disappearance and the feature loss in pooling and for the improvement of recommendation quality. Human resource datasets are fetched by using a cloud-based distributed data collecting framework. Using the datasets, experiments on the proposed recommender system have been done and analyzed. Our proposed algorithm shows better recall rate and F1-score than some other recommender algorithms.