WSEAS Transactions on Computers
Print ISSN: 1109-2750, E-ISSN: 2224-2872
Volume 18, 2019
Comparison of Classification Algorithms on Financial data
Author:
Abstract: Today’s life, big data can be seen in many fields. There are many computer-based methods developed and continuing to be developed to assess the big data more efficiently. Data mining is one of them. In this paper, two Canadian banks’ daily stock market price changes are examined by ten data mining algorithms to see which algorithm or algorithms classify the financial data well. For this purpose, thirty-seven years of daily stock price changes for two Canadian banks with 21 independent variables and one dependent variable, price, were obtained from NASDAQ. Ten data mining algorithms were applied to two datasets separately and the performances of the algorithms were compared and tested based on accuracy, kappa statistic, process time and confusion matrix. It was observed that tree algorithm, J48, and meta-analysis algorithms, Meta-Attribute Selected Classifier, Meta-Classification via Regression and Meta-Logitboost, classified the financial data with high accuracy. The results show that tree algorithm, J48, and the meta-analysis algorithms, Meta-Attribute Selected Classifier, Meta-Classification via Regression and Meta-Logitboost, are promising alternative to the conventional methods for financial prediction.
Search Articles
Keywords: Classification, Logistic Regression, Fuzzyrough-NN, Genetic Programming, J48, Random Forest, Navie Bayes, Navie Net, Meta-Analysis, Weka, Data mining
Pages: 256-263
WSEAS Transactions on Computers, ISSN / E-ISSN: 1109-2750 / 2224-2872, Volume 18, 2019, Art. #33