A Wide & Deep Learning Sharing Input Data for Regression Analysis
Minkyu Kim, Suan Lee, Jinho Kim
Abstract
The Wide& Deep learning which combinesthe wide component and deep component shown good performance in recommendation systems. However, the Wide & Deep learning lack of a research and experimental result on regression analysis. In this paper, we experiment the Wide & Deep learning on regression analysis and we also present new Wide & Deep structure named WDSI which have better performance than the existing Wide & Deep learning in regression analysis. The wide component and deep component of the WDSI sharing a same input data. This strategy can reduce burden on a hand-crafted variables and it have better performance than the Wide & Deep learning in regression analysis. We also show that the WDSI outperforms a traditional machine-learning and deep-learning models in regression analysis.