Data Science
Hospital Drug Inventory Forecasting
Web app for hospital drug-stock management with a demand-forecasting module using Double Exponential Smoothing (Holt's method).
- Role
- Full-Stack Developer
- Year
- 2026
- Technologies
- PythonFlaskSQLiteChart.js
Overview
A drug-inventory information system whose forecasts exactly reproduce the research's Chapter IV results — from descriptive statistics to the period-106 prediction.
Approach
Grid search over 81 α × β combinations, choosing the parameters with the lowest test-set MAPE; MAD, MSE and MAPE are computed on test data only.
Solution
Flask + SQLAlchemy + SQLite (no database server), user login, local Chart.js charts that work offline, and a summary of the best α/β with MAPE category.
Model Details
- Problem
- Forecasting drug demand per period.
- Model
- Double Exponential Smoothing (Holt's linear method)
- Validation Strategy
- Train/test split; MAD, MSE, MAPE evaluated on the test set.
