Time Series Analysis and Forecast of Sales of New Car and Used Car Using SARIMA Model

This paper conducts a comprehensive time series analysis of new and used car sales in the United States, focusing on intrinsic patterns captured by Seasonal AutoRegressive Integrated Moving Average (SARIMA) models. SARIMA models are applied to forecast sales over the next two years, and select the model based on standards such as Akaike Information Criterion with correction (AICc), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE). Notably, the ARIMA (4, 0, 3) (3, 1, 1) [12] model emerges as the optimal fit for new car sales, displaying superior time series fitting and lower errors. For used car sales, the ARIMA (2, 0, 3) (2, 1, 3) [12] model, although not the best-fitting, exhibits the lowest prediction errors. Consequently, these models are chosen for forecasting. The results suggest a continued upward trajectory in new and used car sales in the United States over the next two years, capturing the inherent cyclic and seasonal patterns inherent in the data.