An improved combined small sample time series data prediction model
Yong Shuai, Chuan Yang, Biao Li, Xiaoping Huang, Xinyi Su · 2023
Aiming at the problems that the current artificial intelligence (AI) large model can not accurately carry out small-sample time series data prediction and the sample incremental method was not applicable to small-sample time series data prediction, we proposed an improved combination prediction model for small-sample time series data. Firstly, we used multidimensional data incremental collection method based on the assumption of homogeneous distribution, incremental dataset slicing method, and parallel data preprocessing method to carry out incremental modeling for the sample set. Then, we carried out model selection and hyper-parameter optimization modeling, customized the selection method of optimal model and hyper-parameter combination and the definition of predictive data. Finally, the case study showed that the model proposed in this paper had higher prediction accuracy and broader applicability.