Research Of Stock Classification Based On PCA-NBC Algorithm

Bayesian classification is an important application of Bayesian Statistical theory.It has been widely used,especially with the rise of Machine Learning and Data Mining.Currently,it has entered many application areas successfully,such as Industry,Computer etc.But it has few research in the Domestic economy.This paper uses the method called Naive Bayes Classifier (NBC) in the research of stock classification. It will be divided the stocks of Listed Companies into ST shares and non-ST shares.In this paper,the main work:1.Firstly, using Principal Component Analysis (PCA) on a large sample of high-dimensional stock data to reduce the dimension,to obtained a few main components that can comprehensive reflect the amount of information of raw data as much as possible.The obtained main components are also independent of each other,which is exactly in line with the prerequisites of Naive Bayesian Classification method.2.Secondly,using Naive Bayesian Classification on the processed data to classify.We will comparative study with Support Vector Machines. By comparison, Naive Bayesian Classification method has high classification accuracy and simple explanation in the practical problems during the classification process of high-dimensional data.However,we should notice that there have a sufficient number and representative number of training samples for training to make the model has a good classification capability in the learning stage.