Research On Quantum Machine Learning In The Field Of Sequential Finance

With the continuous development of artificial intelligence,various fields have recognized artificial intelligence as a crucial technology for future development.Machine learning algorithms,as an important branch of artificial intelligence,have demonstrated significant model advantages in various practical scenarios,especially in the financial field that involves massive amounts of data.The stock data,being a temporal problem,has gained attention from many researchers in the financial field because of its large data volume and complex influencing factors.Machine learning models have been successful in solving such problems due to their algorithmic advantages.However,traditional machine learning algorithms and classical computers face challenges like increasingly complex data types,growing data volume,and model parameters.In recent years,the continuous development of quantum computers and quantum machine learning has provided a new solution to these problems.Quantum machine learning can use a reduced number of quantum bits compared to traditional computers to code massive financial data.Moreover,the number of parameters in the machine learning algorithm can be significantly reduced.Additionally,the unique calculation method of quantum computers makes operations more efficient in both time and space dimensions,making it possible for quantum machine learning algorithms to be applied in the field of temporal finance.In thesis,we applied quantum machine learning algorithms to the field of temporal finance and completed the following work:1.To solve the problem of existing quantum hidden Markov models being limited to dealing with discrete data,we propose a method of discretization and feature extraction for stock price trend data.This method includes uniform discretization,as well as discretization and feature extraction using Gaussian mixture models and the MVTF(Mean and Variance To Feature)model.The proposed method uses Gaussian mixture models to discretize continuous sequential financial stock data based on specific methods,enabling them to be better applied to quantum machine learning models.2.Based on the data characteristics of stock prices,we compared and operated on three different models to predict different stock price trends using historical data,and proposed the final funds Fc’ after buying and selling operations according to the model prediction as an important indicator to measure the predictive effect of the model.3.Based on the methods proposed in parts I and II,and the framework of quantum machine learning models,we designed a complete set of quantum hidden Markov models that comply with quantum computing rules.We applied the algorithm to actual temporal financial stock markets,and it demonstrated the feasibility of the algorithm model in the field of temporal finance.4.Based on the price trend data of a total of 15 stocks from the Shanghai Stock Exchange and Shenzhen Stock Exchange over a period of time,the quantum hidden Markov model proposed by III was used to predict stock prices.The prediction of different stock trends under different algorithm models was compared.The final indicators Fc’ of each stock were calculated using historical data,and the prediction accuracy of quantum machine learning and classical machine learning was compared.Additionally,the differences in model structure,parameter types,and resource consumption between quantum machine learning and classical machine learning were compared to demonstrate the advantages of quantum machine learning in the field of temporal finance.