Exploring data elements pricing with market factors -machine learning models based on integrated algorithms

• Published in: Journal of data, information and management (Online) Vol. 6; no. 4; pp. 423 - 438
• Main Authors: Zhang, Chunling, Li, Anna
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 01.12.2024; Springer Nature B.V
• Subjects: Accounting; Accuracy; Algorithms; Artificial Intelligence; Big Data; Business and Management; Computational Intelligence; Data analysis; Data exchange; Decision trees; Deep learning; Empowerment; Engineering; Ensemble learning; Graphical representations; K-nearest neighbors algorithm; Machine learning; Original Article; Pricing; China; Stacking ensemble algorithm; Data element pricing; Data elements; Machine learning
• ISSN: 2524-6356; 2524-6364
• DOI: 10.1007/s42488-024-00133-0

Data have gradually become a core element for enterprises to acquire benefits from transformation and enhance their decision-making capabilities. The scientific pricing of data elements is a crucial link in unlocking their value of data elements. This study utilizes the transaction information of relevant data products on the Youyi Data Cloud data-trading platform. Firstly, it determines the scope of data research using word cloud analysis. Then, it employs machine learning to analyze and price the data elements using single-algorithm models. The accuracy and generalization ability of the models were evaluated by comparing them with actual market prices. Finally, based on the results, analysis and selection were conducted among random forest algorithm models, XGBoost algorithm models, KNN algorithm models, MLP algorithm models, gradient boosting tree algorithm models, and decision tree algorithm models. Appropriate algorithms were selected for stacking ensemble learning to further optimize the algorithm models. The results indicate that: (i) Using machine learning algorithms to price data elements can simulate prices under dynamic mechanisms and adjust to relevant fluctuations in a timely manner. (ii) Stacking ensemble algorithms can improve the fit of the data element pricing model and optimize learning effects. (iii) Data element pricing models based on machine learning algorithms can fully consider the influence of market factors, comprehensively explore the most real and accurate requirements for data element pricing, and provide a solid basis for pricing data products on data-trading platforms. (iv) Data element pricing models based on machine learning can capture dynamic changes in the market and make timely adjustments and optimizations to ensure the accuracy of pricing models.