Q-Rung Orthopair Fuzzy Twin Support Vector Machine And Its Application In Stock Prediction

Twin support vector machine,as an extension of support vector machine,has not only fast computing speed but also strong generalization ability,and it has achieved a good application effect in stock prediction.However,owing to the complexity and randomness of the stock market,there are a lot of noise points and outliers in the stock data,which affect the prediction effect of TSVM to some extent.Based on this problem and twin support vector machine,q-rung orthopair fuzzy numbers are applied to twin support vector machines and proposed the q-rung orthopair fuzzy twin support vector machine(q-ROFTSVM).And it is applied to the practical problem of stock prediction to verify its validity and practicability.The main contents of this thesis are as follows:(1)The construction of q-rung orthopair fuzzy twin support vector machine.Based on the kernel function,the independent calculation methods of membership and non-membership of training data on q-rung orthopair fuzzy numbers are given in the high-dimensional feature space,the score function of q-rung orthopair fuzzy numbers is used to measure the taxonomic contribution of each training point,the score function of the orthopair fuzzy number of each training data is applied to the construction of the twin support vector machine,and establishs the q-rung orthopair fuzzy twin support vector machine(q-ROFTSVM),which improves the sensitivity of the algorithm to noise and outliers and he generalization ability of the algorithm to classification problem.In the simulation experiments of UCI data sets show that the process has better conversion performance and noise resistance.(2)The multi-class problem of q-rung orthopair fuzzy twin support vector machine.Based on the thought of "one-versus-one TSVM" and "one-versus-rest TSVM",this thesis proposes multi-classification problem’s algorithms of q-rung orthopair fuzzy twin support vector machine,extends q-rung orthopair fuzzy twin support vector machine to multi-classification,and analyzes and verifies it on the UCI data sets.The experiments show that the proposed "one-versus-one q-ROFTSVM" and "one-versus-rest q-ROFTSVM" algorithms in this thesis is more efficient than "one-versus-one SVM","one-versus-one TSVM" and "one-versus-rest TSVM" on some of the data.(3)The application of q-rung orthopair fuzzy twin support vector machine in stock prediction.The q-rung orthopair fuzzy twin support vector machine is applied tothe stock price trend prediction,through stock selection of relevant indicators,the pretreatment of the raw data and the prediction of stock price trend,it further proves the effectiveness and practicability of this algorithm.