Small sample data pricing research based on Reptile algorithm

Reasonably pricing data resources contributes to the development of the digital economy. However, in the early stage of the development of the data market, the data transaction volume is small, and the resulting small sample problem makes it difficult to accurately model and forecast the price of data. To address this issue, this paper constructs a meta-learning-based data price prediction model: MLP-Reptile. The model introduces a meta-learning tuning module to optimize the weight parameters of the base model, facilitating effective knowledge transfer learned from multiple tasks to enhance prediction accuracy in new tasks. Experimental results demonstrate that the proposed MLP-Reptile model excels in small sample data pricing tasks, outperforming other models. Additionally, the paper analyzes the different primary factors influencing data prices in various industries. The methods proposed in this research are universally applicable for addressing the small sample problem in data pricing, providing a reference for solving similar issues.