Development of a New Machine Learning-Based Expert System for Prediction of Service Management of Infusomat/Infusion Pump

Today’s rapid technological development significantly impacts various industries, including healthcare. From routine checkups to intricate surgical procedures, medical devices play a crucial role at every stage of healthcare delivery. While advanced technology has streamlined doctors’ work and enhanced patient care standards, the inadequate and delayed maintenance of these devices may lead to fatal consequences. Despite legal restrictions on medical device maintenance, serious injuries due to malfunctions are still prevalent. The primary aim of our research is to take the initial steps toward automating medical device inspection and maintenance through innovative technological processes. This research resulted in an expert system capable of predicting when maintenance is needed and when a device may malfunction. We experimented with six machine learning algorithms combined with three feature selection algorithms to create the optimal machine learning model, serving as the core of the expert system. Each algorithm’s performance was evaluated across six different metrics, and the Support Vector Machine combined with the Filtered Attribute Evaluator emerged as the most efficient and accurate prediction model. Although all algorithms demonstrated satisfactory overall accuracy, the support vector machine outperformed others, achieving 100% classification accuracy with filtered attribute evaluator feature selection. This model will power the backend of a Java application, enabling predictive functionality for web and mobile applications. This marks the first step towards automating the management and inspection of medical devices.