Demand Forecasting in the Tunisian Pharmaceutical Industry: A Comparative Study.

Area of Science:

• Pharmaceutical industry
• Data science
• Machine learning

Background:

• Accurate demand forecasting is vital for efficient pharmaceutical operations, inventory management, and distribution.
• The Tunisian pharmaceutical sector faces challenges in achieving precise demand prediction.
• Dynamic markets necessitate advanced forecasting strategies.

Purpose of the Study:

• To compare the forecasting accuracy of Holt-Winters (HW) and Multilayer Perceptron (MLP) neural networks.
• To evaluate forecasting performance across Antiviral, Antibiotic, and Pain Relief drug categories.
• To provide recommendations for enhancing pharmaceutical demand forecasting strategies in Tunisia.

Main Methods:

• Analysis of a 24-month historical sales dataset (October 2020 - September 2022).
• Application of the Holt-Winters model for seasonal adjustments.
• Utilization of a Multilayer Perceptron neural network to capture non-linear sales patterns.

Main Results:

• The Multilayer Perceptron (MLP) neural network demonstrated significantly superior forecasting accuracy compared to the Holt-Winters (HW) method.
• MLP achieved markedly lower Mean Squared Error (MSE) values across all analyzed drug categories (e.g., 0.0206 for Antivirals vs. 30.06 for HW).
• MLP exhibited greater adaptability to demand variability and complex sales patterns.

Conclusions:

• Multilayer Perceptron (MLP) neural networks are superior to Holt-Winters (HW) for pharmaceutical demand forecasting due to their accuracy and adaptability to non-linear data.
• Findings support Tunisian pharmaceutical companies in adopting advanced machine learning for improved planning.
• Future research should investigate hybrid models and incorporate external market dynamics for enhanced forecasting.