The improvement of model forecasting performance for long- standing multivariate time series is the primary focus of academic research on time series prediction that encompasses a range of methodologies. This paper aims to examine the performance of Autoregressive Integrated Moving Average model (ARIMA) and Long Short-Term Memory (LSTM) models in predicting timeseries data. Through the process of historical data, the author evaluated the precision and efficiency of these two models in forecasting Apple Inc.’s closing stock prices. The previous articles suggest that while the ARIMA model performs well with linear and stable datasets, the LSTM model, due to its deep learning architecture, can capture more complex patterns and interactions, leading to more accurate predictions in nonlinear and volatile markets. The study also discusses the trade-offs between the two models in terms of data requirements, computational resources, and model interpretability. These two models are pivotal in time series forecasting, providing robust frameworks for predicting future data points by analyzing historical patterns, with ARIMA focusing on linear dependencies through autoregressive and moving average components, and LSTM leveraging deep learning to capture complex, non-linear relationships.
