Comprehensive Evaluation of Time Series Forecasting Methods for Seasonal Data: A Statistical and Machine Learning Approach

Abstract- Forecasting sales data in industries characterized by seasonal demand, such as the wine industry, is essential forensuring optimal business operations and strategic planning. This study evaluates historical Rose wine sales data to developaccurate forecasting models that address both short-term fluctuations and long-term trends. The dataset, spanning severaldecades, exhibits clear seasonal patterns and an overall declining trend. Through comprehensive preprocessing and modelevaluation, several time series forecasting techniques were explored, including Exponential Smoothing, ARIMA, and SARIMA. Acomparative analysis of these models based on RMSE revealed that Holt-Winters Triple Exponential Smoothing achieved the bestperformance, effectively capturing both trends and seasonality. Additionally, advancements in time series forecasting, includingmachine learning techniques, are discussed to highlight future directions. This paper provides actionable recommendations foraddressing declining sales and improving operational efficiency through accurate forecasts.