<![CDATA[Rapid technological developments have changed people's lifestyles, marked by an increase in online activity. APJII states that by 2025, 229 million Indonesians will be internet users, while BPS recorded 3,816,750 digital businesses in 2023. This growth has encouraged the use of websites as the primary medium for business and information. Therefore, understanding visitor trends and seasonal patterns is crucial for effective and efficient management. This study offers a new contribution by predicting the number of visitors to the website of PT. XYZ, a tea company in Indonesia. It uses three time series models: Seasonal Auto Regressive Integrated Moving Average (SARIMA), Long Short-Term Memory (LSTM), and Holt-Winters Triple Exponential Smoothing (Holt-Winters TES) with a prediction period of 71 days prior and to predict 14 days ahead. The dataset used consists of 32,518 visitor data entries. Model performance is evaluated using Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE). The results show that LSTM achieved the lowest error with an MSE of 8257.23, an RMSE of 90.87, and a MAPE of 12.73%. Therefore, the LSTM model achieved the highest accuracy, while Holt-Winters TES performed better than SARIMA in certain aspects. Visitor predictions can support strategic decision-making and content management]]>
Copyrights © 2026
