Evaluating Machine Learning Techniques for Stock Loss Forecasting: Hypothesis Testing and Model Selection

Comparative Study is a technique that provides deep insights into the context via provisioning various methodologies and optimizing to get the final conclusion. In this paper, a hypothesis is proposed “Different firms, at different locations influence the stock prices of one another”. In this paper, a comparative analysis is performed to ensure the reliability of the model and prove the hypothesis. Various classical machine learning models like Linear Regression, Lasso and Ridge Regression, Support Vector machines (SVMs), Decision Trees, Random Forests and XGBoost are used to interpret the agenda of the paper uniquely. Prior studies include proposing different models for different firms, but in this paper instead of relying on intuition, a scientific method of testing and documentation in order to arrive at a relationship between independent features. The relationship is proposed in the results section, along with a prediction accuracy of $\mathbf{9 9. 9 8 7 \%}$.