Machine learning algorithms for forecasting and backcasting blood demand data with missing values and outliers: A study of Tema General Hospital of Ghana

The major challenge in managing blood products lies the uncertainty of demand and supply, with a trade-off between shortage wastage, especially most developing countries. Thus, reliable predictions can be imperative planning voluntary donation campaigns improving availability within Ghana hospitals. However, historical datasets on are predominantly contaminated missing values outliers due to improper database management systems. Consequently, time-series prediction challenging since data cleaning affect models’ predictive power. Also, machine learning (ML) power for backcasting past years’ lost is understudied compared their forecasting abilities. This study thus aims compare K-Nearest Neighbour regression (KNN), Generalised Regression Neural Network (GRNN), Auto-regressive (NNAR), Multi-Layer Perceptron (MLP), Extreme Learning Machine (ELM) Long Short-Term Memory (LSTM) models via rolling-origin strategy, from government hospital Ghana. KNN performed well (12.55% error); whereas, ELM achieved highest (19.36% error). Future studies also employ ML algorithms as good alternative that time-reversible.