A Comparative Investigation of Linear and Non-Linear Methods for Stock Price Prediction

Predicting stock prices is a common topic in finance and data science. Investors and companies are interested in possible future price movements, but making accurate predictions is difficult. Stock prices can be affected by trading activity, company information, economic conditions, investor decisions, news, and unexpected events. This paper reviews several approaches for stock price prediction, with a focus on both linear and non-linear methods. The methods discussed are Linear Regression, ARIMA, Support Vector Regression (SVR), Decision Trees, Random Forest, Artificial Neural Networks (ANN), and Long Short-Term Memory (LSTM). Previous studies are compared based on their datasets, prediction targets, evaluation measures, and main findings. In general, simpler models are easier to understand and use, while non-linear models can deal with more complicated patterns. However, complex models may require more data, computing power, and careful training. The paper also considers problems related to external information, model interpretability, overfitting, and changes in financial markets. Future studies could combine market data with news and sentiment, improve the explainability of AI models, and allow prediction systems to update as market conditions change.