Machine Learning-Based Fundamental Stock Prediction Using Companies' Financial Reports

Portfolio management is a key goal of investment. Therefore, a key factor in extremely successful revenues is the ability to predict shares and securities with efficiency. The use of fundamental stock analysis to forecast the long-term performance of shares is effective and potent. However, further investigation in this field suffers from the lack of online fundamental features. This paper proposes two approaches of time-dependent and independent stock prediction utilizing fundamental analysis. To accomplish this, we first present a brand-new fundamental feature dataset from TSE. Its features are extracted from corporate financial records, and each share's return and risk values serve as its prediction targets. Additionally, the impact of using temporal dependency-which is lacking from the great majority of machine learning techniques-on prediction performance is examined in this paper. Finally, we suggest using machine learning techniques and assess their effectiveness of fundamental stock prediction.