Enhancing Revenue Forecasting for Pharmaceutical Companies with a Hybrid FFNN-Holt-Winters Model

Accurate revenue prediction is a critical aspect of financial planning and decision-making in the pharmaceutical industry Traditional forecasting methods often struggle to capture the complex patterns and dynamics inherent in revenue data PT Kalbe Farma Tbk a leading Indonesian pharmaceutical company, faces challenges in achieving precise revenue forecasts, hindering optimal resource allocation and strategic planning. This study addresses this challenge by proposing a novel hybrid revenue prediction model that leverages the strengths of both a Feedforward Neural Network (FFNN) and the Holt-Winters algorithm The FFNN captures intricate non-linear relationships within 15 years historical revenue data, while the Holt-Winters algorithm effectively models the underlying trend and seasonality inherent in this extensive dataset. By integrating these two complementary techniques, the hybrid model aims to provide more accurate and reliable revenue forecasts than either method alone, particularly over longer time horizons. The model is trained and evaluated using the 15-year revenue dataset and its performance is benchmarked against standalone FFNN and Holt-Winters models. The results demonstrate a significant improvement in prediction accuracy, as measured by Mean Absolute Error (MAE) and Mean Squared Error (MSE), highlighting the potential of this hybrid approach to enhance revenue forecasting capabilities in the pharmaceutical industry, especially when utilizing extensive historical data. This research contributes a practical and effective solution for revenue forecasting in the pharmaceutical industry, ultimately benefiting and potentially other companies in similar sectors by enabling more informed decision-making and improved financial performance.