Machine Learning Algorithms and Applications in Investment Analysis

Abstract

We can simplify investment analysis as filtering out speculative stocks, bonds, derivatives and other financial products. This area is very challenging yet extremely critical since individual investors’, large institutions’ and the public’s understanding of investment and investing behaviors determine the long-term stability of the U.S. and the interconnected global economy. This dissertation focuses on how to utilize Machine Learning (ML) techniques to facilitate investment analysis, what are the challenges in practice, and how to bridge the gap by choosing appropriate algorithms and modify them to mitigate the risk of significant financial losses. A general work path and investigated topics are as follows: 1. Demonstrate a comprehensive understanding of the U.S. economy, its key industries, and the traditional investment analysis principles. 2. Develop a thorough knowledge of several widely applied ML algorithms and gain hands-on experience through applications. We simulate these algorithms and its derivations in different scenarios and test it with correspondent accuracy and efficiency measures. 3. Present cases to explain why and how irregularities or uncertainties affect algorithms performance, propose and implement solutions to improve classification results. This dissertation uses the Mergent or Securities Exchange Commission (SEC) published datasets. By connecting algorithms with real word datasets, this dissertation successfully demonstrates how crucial it is to understand your data and how ML algorithms can facilitate similar decision-making processes.