Modelling on gross domestic product annual growth rate data by using time series, machine learning, and probability models

Background: The gross domestic product (GDP), which displays linear and non-linear patterns, is an essential economic indicator for gauging a nation's economic growth. To model and forecast Switzerland's GDP annual growth rate percentage, a crucial financial indicator of the country, this study will investigate and suggest an effective and precise time series approach for modeling and forecasting. Methodology: The data consists of Switzerland's GDP growth rate (%) data from 1961 to 2023, sourced from the World Bank's official website. The Time series approaches such as the Auto-Regressive Integrated Moving Average (ARIMA) model, an Artificial Neural Network (ANN), Error Trend Seasonal (ETS) model, and Exponential Smoothing Models (Brown, Holt, and Winters) have been applied. Furthermore, deep learning techniques, including Support Vector Machine (SVM), Random Forest (RF), Decision Tree Model (DTM), and Artificial Neural Networks (ANN) are studied. Moreover, probability distributions such as Exponentiated Normal (EN), Exponentiated Logistic (EL), Exponentiated Gumbel (EG), Normal (N), Logistic (L), and Gumbel (G) to analyze Switzerland's GDP annual growth rate percentage. We use the goodness of fit criteria such as Cramer-von Mises (W), Anderson-Darling (A), and Kolmogorov Smirnov (KS) for comparison of EL, EG, N, L, and G models. Results: ANN is the best model for predicting Switzerland's GDP growth rate in time series analysis based on the overall lower error values (RMSE, MAE, and MASE). Even when deep learning techniques and probability distributions are considered, ANN performed very well and is the most accurate model for this specific application.