Abstract
The present disclosure focuses on mitigating underfitting issue in time series model using regressors. The present disclosure replaces the large parameter with a new structure. The new structure may include a few layers of smaller models. Specifically, the “smaller” model may be defined as the number of parameters for the model may be smaller but can handle all the features. Thus, each smaller model is trained with all the data points. The smaller models may generate intermediate predictions which may be fed as the input to the next smaller model present in the next layer. As a result, the DPP value of each model in the structure is substantially higher and hence the underfitting issue may be efficiently resolved.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Li, Pengfei; Yang, Pei; Wang, Shuo; Wang, Zhuoyi; and Wang, Dan, "A METHOD FOR MITIGATING UNDERFITTING ISSUE IN TIME SERIES MODEL USING REGRESSORS", Technical Disclosure Commons, (April 19, 2023)
https://www.tdcommons.org/dpubs_series/5816