Stock prices forecasting using machine learning techniques

Abstract

Predicting stock behavior is an issue that has been studied by many scientists and engineers, and it forms today's cutting-edge research. Great interest can be found in the combination of the financial science and the IT science, as the correct prediction of stock shares is of great economic importance for the apparent reason that an application could be profitable for either the creator of the customer. Using computing power today, which has reached unprecedented levels of power, we can build more complex and computationally powerful applications than ever before. A typical example for approximately the past ten years is the exponential growth in the development and use of "Artificial Intelligence" applications. More and more people see the prospects and the countless applications that such models in the real world could give, as countless companies are already providing these services, such as the colossal companies Google and Amazon. Therefore, using Python programming language and its libraries like SciKit Learn and Keras, we can create stock behavior, prediction models. More specifically, we will study linear models of the SciKit Learn library such as Linear Regression, Tree Decision, and Deep Learning models of the Keras library, such as Long Short-Term Memory, which is a more accurate model than the previous models in terms of stock forecasting. The purpose of the current thesis is to analyze these forecasting techniques and learn how they can be applied in the financial data and become familiar with the Python programming language and its capabilities in data analysis and processing.