forecasting daily closing values

Undergraduate research submission to the North Carolina Journal of Mathematics and Statistics.

abstract

Over the past decades, machine learning has become an essential area of research with relevant applications in classification and regression. Artificial intelligence techniques can be used for statistical analysis of stock markets which is one example of a time series. Within this research, Artificial Neural Network models were trained for the purpose of forecasting daily closing prices of five arbitrarily chosen stocks: Apple inc., Walmart, Bank of America Company, Ford Motor Company, and Coca-Cola. The accuracy of Back-Propagation models optimized using stochastic gradient descent were compared with an algorithm originating from Nanyang Technical University called Extreme Learning Machines. The results indicate that the Extreme Learning Machine model trained faster, classified the up and down of a stock more accurately, and had closer predictions when compared with the Back-Propagation Neural Network model.

install

sudo apt install texlive-full

usage

pdflatex main
bibtex main
pdflatex main
pdflatex main