Pharmaceutical Product Forecasting A Case Study of AAA Pharmacy in Nakhon Ratchasima Province, Thailand (Northeastern Region)

Abstract

This research aims to develop a forecasting model for pharmaceutical product demand at AAA Pharmacy in Nakhon Ratchasima Province by studying Machine Learning techniques. The study examines ten quantitative forecasting methods 1).Moving Average 3 months (MA3) 2).Moving Average 12 months (MA12) 3).Single Exponential (SE) 4).Double Exponential (DE) 5).Winters’ Method Multiplicative (WMM) 6).Winters’ Method Additive (WMA) 7).Trend Analysis (TA) 8).Time Series Decomposition Multiplicative (TSDM) 9).Time Series Decomposition Additive (TSDA) 10). Long Short-Term Memory (LSTM). Additionally, the research utilizes MINITAB software to reduce the issues of product shortages and excessive inventory. The study focuses on forecasting 29 hazardous product Stock Keeping Units (SKUs), categorized into 11 subgroups eye drops, antibiotics, anthelmintics, contraceptives, anxiolytics, cough medicine, anti-inflammatory drugs, cold medicine, antihistamines, pain relievers, and acid reflux medication. The data covers a three-year period from January 2018 to December 2020 to determine the most suitable forecasting method for each product category. The accuracy of the forecasts is measured using Mean Absolute Deviation (MAD).
From the research findings, appropriate demand forecasting methods were determined by categorizing products into four groups based on their characteristics. 1. The Moving Average method, which provides the highest forecasting accuracy and lowest error rate, is suitable for stable demand products such as pain relievers and anxiolytics. 2. The Trend Analysis method is appropriate for trending demand products like contraceptives, as it effectively captures growth patterns in the data. 3. Winters' Method Multiplicative, which offers the highest accuracy in forecasting seasonal demand products that experience high demand during specific seasons, including antibiotics, cough medicine, antihistamines, and cold medicine. 4. The Long Short-Term Memory (LSTM) machine learning model efficiently manages highly volatile data for irregular demand products with uncertain demand patterns, including eye drops, anthelmintics, anti-inflammatory drugs, and acid reflux medication. Pharmacies can select forecasting methods appropriate for their product categories. Overall research findings conclude that the LSTM method provides the best error reduction when implemented for forecasting, enabling companies minimize costs and maximize inventory management efficiency.