Abstract
Stock is part of a company's principal. A person who buys stock of a company shares the profit or loss of this company. Large volume transactions are made on stock exchanges where stocks are traded. Stock prices are difficult to predict because they are affected by many variables, but when they can be predicted, great benefits are provided. Prediction of stock prices is possible with today's computers using machine learning algorithms. Machine learning provides more successful results than fundamental and technical analysis in stock price prediction. In our study, daily closing price predictions were made by collecting approximately 5-years data of the top 5 stocks with the highest market value traded in BIST 100 between 2016 and 2020. Multiple linear regression, Bayesian regression, random forest, decision trees, support vector machines, artificial neural networks algorithms were applied to include maximum 22 features and the results were compared. The most successful result was obtained in the artificial neural networks algorithm. To achieve the highest success, data pre-processing, normalization, cross-validation, parameter optimization and feature selection were applied. It has been observed that using these methods together increases the success.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call