Forecasting India S&P BSE SENSEX and USA S&P-500 Benchmark Indices Using SARIMAX and Facebook Prophet Library

The study forecasts the equity market benchmark index’s daily close values and captures volatility dynamics from assorted trends of India and USA stock data. To accomplish the objectives, the use of descriptive statistics and plots related to lag, ACF, and PACF were developed to understand the dataset with unit-root tests such as Augmented Dickey-Fuller for checking the stationarity and to achieve reliable performance, models such as SARIMAX and the Prophet (open-sourced by Facebook) are trained and validated. The prophet-based model has significantly outperformed the corresponding SARIMAX models, with MAPE values of 1.06% for S&P BSE SENSEX of India and 0.62% for S&P-500 of the USA. The models developed can capture the assorted trends with the help of exogenous variables explicitly introduced in the index’s data, which usually presents a challenge when dealing with financial time series datasets. These models can help develop the investment strategy for short-term or performing portfolio optimizations to maximize profits. Although there exist various studies [7] about the use of classical ARIMA and Prophet-based models for forecasting, it can be argued that the novel approach presented in the study to collectively forecast and compare the performance of models using exogenous variables to capture the volatility dynamics from assorted trends present in the India and USA benchmark time series data.