| As a crucial part of consumer finance,credit card business has appeared in the early business of international banks.In recent years,with the rise of Internet consumer finance and online transactions,credit card fraud detection has become an important part of credit card risk management and has attracted much attention.Traditional credit card fraud detection methods in the past have some problems,such as being unable to deal with large amounts of data and adapting to real-time requirements.Therefore,the development of intelligent methods is urgently needed.In recent years,machine learning,as a hot technical means in the field of artificial intelligence,has been widely adopted and applied in the financial field,and credit card fraud detection and judgment models based on different machine learning methods have also been established.However,with the advent of the information age,the rapid growth of data scale leads to the increase of computing resources and low efficiency of classical machine learning algorithms.Quantum computing,with its potential for exponential acceleration,holds out the promise of breaking the performance bottleneck of classical computers in processing large-scale data.At the same time.Quantum machine learning comes into being,which is a new interdisciplinary theory and method combining machine learning and quantum computing,which can make the computational model more suitable for the processing of big data.This paper is based on this new theory to study the problem of credit card fraud detection in finance,the main research content is as follows:1.A credit card fraud detection model is proposed under the framework of hybrid(quantum + classical)quantum computing theory.This model uses quantum hidden Markov model to replace classical hidden Markov model to model the timing sequence of credit card transactions,and generates different quantum hidden Markov models on different data sets.The output value of the probability of different models on each piece of data is added into the feature data set as a new feature for fraud detection and judgment.It is found that although the quantum hidden Markov model has similar performance with the classical hidden Markov model,it uses less resources,which leads to the improvement of the efficiency of decision.2.Based on the established quantum credit card fraud detection,the influence of quantum classifiers on the accuracy and other performance of the model is further studied,that is,by using different quantum classifiers to replace the classical random forest classifier in the hybrid model,the performance difference of credit card fraud detection model algorithms under different quantum classifiers is studied.3.Based on Qiskit simulator,complete the algorithm code implementation of hybrid credit card fraud detection model and credit card fraud detection model based on quantum classifier.4.Aiming at the above two models,a series of training data and test data are used to optimize the parameters and perform performance tests on the models,and the accuracy and robustness characteristics of the constructed models are demonstrated.Finally,based on different test results,the advantages and disadvantages of different models are analyzed and compared. |
