Stock price forecasting with a financial ratio based neural network algorithm

Forecasting stock prices is quite difficult due to uncertainty, chaotic nature and noise in financial markets. The aggregated impact of factors such as political instabilities, financial fragility, international financial integrity, technological developments and change in investor risk preferences make the estimation of stock prices harder. However, the challenge of developing a good estimation model in such an environment, the positive contribution of a successful model to the return of investment make the problem attractive for researchers. It is known that machine learning algorithms are useful in generating predictions in such chaotic environments as stock market, which have multiple sources of data flow. In this study, three stocks traded in Borsa İstanbul are selected according to different criteria and price estimation performances of proposed Artificial Neural Network model together with known Support Vector Machines, Logistic Regression, Random Forest and Naive Bayes Classifier machine learning algorithms are compared. 18 financial ratios frequently used in evaluation of company performances with 102 other independent variables are used as inputs and monthly rate of return of stocks in 2009-2018 period are classified and estimated. Analyses on given period have shown that the proposed artificial neural network algorithm is a classifier that can be used as an alternative to other algorithms for stock market forecasting.