REAL TIME ANOMALY DETECTION IN ELECTRONIC HEALTH RECORDS

Electronic Health Records (EHRs) are invaluable in healthcare, offering a wealth of patient data. Timely anomaly detection in EHRs is critical for identifying emerging health risks, enabling early intervention, and enhancing patient care. The project addresses the need for effective anomaly detection methods in the context of coronary heart diseases using machine learning algorithms. The aim of this project is to evaluate and compare the performance of several machine learning algorithms, including Random Forest, Decision Tree, Naive Bayes, Artificial Neural Networks (ANN). Analyzing the best algorithm based on their suitability, accuracy, and adaptability to evolving data patterns. The project meticulously preprocesses EHR data, identifies relevant features, and the algorithm is used to improve anomaly detection. Each algorithm is trained and optimized, considering performance metrics such as precision, recall, and F1-score. Real-time inference capabilities are assessed, emphasizing the need for models to adapt to changing data patterns. Continuous monitoring and model updating are emphasized to minimize false alarms. The integration of an alerting mechanism facilitates timely healthcare professional intervention upon anomaly detection. In the comparative analysis, Random Forest and ANN demonstrate superior performance, capturing intricate data relationships effectively. Decision Tree and Naive Bayes show moderate performance. The project underscores the significance of real-time anomaly detection in EHRs for coronary heart diseases. It highlights the strengths and weaknesses of various machine learning algorithms in this context. Random Forest and ANN emerge as promising choices, balancing accuracy and interpretability. Ultimately, this research contributes to the advancement of healthcare systems by enhancing their ability to proactively detect anomalies, thereby improving patient outcomes. To address this, data augmentation techniques, such as random flip, random rotation, random translation, random zoom, random contrast, random hue, random brightness, and random saturation, were employed to reduce perspective variability. The study evaluated the performance of different design strategies to identify the approach that achieves the highest accuracy in monument recognition.