Sales Forecasting: A Comparison of Traditional and Modern Times-Series Forecasting Models on Sales Data with Seasonality

The adoption of advanced technologies is crucial for the efficient operations of modern businesses. Sales prediction, demand forecasting, production planning, and inventory forecasting are essential elements of business planning. This study compares classic and modern sales forecasting models to create a reliable and generic prediction tool for businesses and aid CEO decision-making. On sales transactional data, Exploratory Data Analysis was conducted to understand the features in the dataset. After data wrangling, the processed data was used to train the various forecasting models. To help in business strategy decision-making, different forecasting models were applied to the given dataset and the behavior and accuracy of each model were studied. At first, the traditional models based on moving average, Autoregressive Integrated Moving Average (ARIMA), and Seasonal Autoregressive Integrated Moving Average (SARIMA) were utilized, followed by the use of modern sales forecasting models including Long Short-Term Memory (LSTM) and its variants, and Fb-Prophet. The Root Mean Square Error was utilized as the performance measure to compare the various models. The results of the study showed that SARIMA models are better for organizations with fast-moving sales and seasonal patterns. While SARIMA outperforms the basic LSTM model, it is possible to enhance LSTM models to surpass current sales forecasting methods in the future.