Prethodno priopćenje

https://doi.org/10.17559/TV-20220628151845

LSTM Deep Neural Network Based Power Data Credit Tagging Technology

Ding Li ; State Grid Xiongan Financial Technology Group Co., Ltd., China
Jiayi Chen ; State Grid Xiongan Financial Technology Group Co., Ltd., China
Zhuo Wang ; State Grid Xiongan Financial Technology Group Co., Ltd., China
Yuehui Song ; State Grid Xiongan Financial Technology Group Co., Ltd., China

Sažetak

The value of power data credit reporting in the social credit system continues to increase, and the government, users and the whole society have deep expectations and support for power data credit reporting. This paper will combine the data labeling theory as the support, define the power data label and explain its labeling implementation. Based on the construction of knowledge graph, the method of labeling power data is introduced in detail: demand analysis method, index selection method, data cleaning method and data desensitization method. Use the sorted data labels to establish a label system for power data, and through its system, visualize the comprehensive situation of enterprise power data credit information to meet the development of power data credit business. This paper takes shell enterprises as the main representatives of credit risk enterprises, analyzes the power data in the three stages before and after loans, and builds a value mining model for power credit data. In the future, the data labeling technology and value mining model of the power data credit business will be comprehensively applied, and the power data label library and credit model library will be established and continuously improved, so as to facilitate the evaluation of the operation of the enterprise at different stages.

Ključne riječi

labeling; logistic regression model; power credit data; value mining

Hrčak ID:

288433

URI

https://hrcak.srce.hr/288433

Datum izdavanja:

15.12.2022.

Posjeta: 452 *