Stock Price Forecasting for Nvidia Corporation Based on a Hybrid LSTM-ARIMA Model

Investors study forecasting models for future stock values to make assertive decisions, as each trend movement offers different investment opportunities. The forecasting models described in this research are well suited to provide a realistic view of the future behaviour of stock prices. The autoregressive integrated moving average (ARIMA) model is one of the most relevant linear models for time series forecasting. Recurrent neural networks (RNN) are a class of neural networks that allow using previous outputs as inputs, while having hidden states and capturing the robust and non-linear relationships of the sequence. This paper proposes a hybrid methodology that takes advantage of the strengths of RNNs, linear and ARIMA models in value forecasting problems. Real datasets of Nvidia Corporation (NVDA) stock price on NASDAQ were used to analyse the forecasting accuracy of the proposed model. The main objective is to compare the performance of the combined model compared to each of them separately when it comes to stock price forecasting. By hybridising these models, the methodology is able to correctly predict the NVDA share price. The root mean square error (RMSE), mean absolute percentage error (MAPE) and mean absolute error (MAE) metrics were used to assess accuracy while coefficient of determination (R²) was used to measure goodness of fit.