摘要
随着科学技术的进步,量子计算突破了传统的算力瓶颈,在各个领域发挥着越来越重要的作用。在金融领域中,信用评分场景是贷款行业的重中之重。特征选取是一种十分高效的数据预处理策略。特征选择可以构建更简单、更容易理解的模型,提高数据挖掘性能,从中提取有效的特征,降低数据维度,为金融业提供有效的贷款参考信息。主要讨论量子计算在信用评分场景下的应用,改进了金融数据预处理的方式,创新性地使用量子计算机来求解特征选择的QUBO模型,与one-hot转码相比较,所使用的WOE分箱处理策略可以直接解析出特征,筛选结果可以进行直接对比。基于量子计算的特征选取与传统的基于相关性的特征选取策略相比,差距很小,并且由于量子计算机的先天优势,此策略速度更快,更具前景。
With the progress of science and technology, quantum computing has broken through the traditional bottleneck of computing power and is playing an increasingly important role in various fields. In the financial field, the credit scor-ing scenario is the top priority of the loan industry. Feature selection can build simpler and easier-to-understand mod-els, improve data mining performance, prepare concise and understandable data to extract effective features and reduce data dimensions, and provide effective loan reference information for the financial industry. This paper mainly discusses the application of quantum computing in the credit scoring scenario, improves the financial data preprocessing method,and creatively uses quantum computer to solve the QUBO model of feature selection. Compared with one-hot encoding,the WOE strategy used in this paper can directly analyze the features, and the screening results can be directly com-pared. Compared with the traditional feature selection strategy based on correlation, the feature selection strategy based on quantum computing is little difference. Moreover, due to the inherent advantages of quantum computer, this strategy is faster and more promising.
作者
文凯
马寅
王鹏
朱德立
Wen Kai;Ma Yin;Wang Peng;Zhu Deli(Beijing Qboson Quantum Technology Co.,Ltd.,Beijing 100016,China;Everbright Technology Co.,Ltd.,Beijing 100040,China)
出处
《网络安全与数据治理》
2022年第9期13-18,共6页
CYBER SECURITY AND DATA GOVERNANCE
关键词
量子计算
特征选择
相关性
信用评分
quantum computing
feature selection
correlation
credit evaluation
