Comparative Analysis of Traditional and LSTM Models for Netflix Stock Movement Classification

With the continuous development of machine learning methods, various types of models are increasingly used in stock prediction. However, the systematic comparison between traditional machine learning methods and deep learning methods is still relatively insufficient for classification prediction tasks such as upward and downward direction. In this study, a series of technical indicators were constructed as input features with the upward and downward direction of Netflix stock as the classification prediction target, and Logistic Regression, Decision Tree, Random Forest, eXtreme Gradient Boosting, Voting Classifier, and Long Short-Term Memory Network models were trained and evaluated, respectively. The results show that the logistic regression model and Extreme Gradient Boosting (XGBoost) perform well overall in all evaluation metrics, indicating their strong generalization ability in this classification task. And although Long Short Term Memory (LSTM) is good at dealing with time series, its performance on this dataset did not exceed that of traditional methods. After the study, it was found that traditional classification models have strong stability and practicality in certain situations. This paper provides empirical support to explore the applicability of traditional models in financial time series classification.