NOVEL MACHINE LEARNING ALGORITHMS FOR PREDICTING RISKS OF CHRONIC DISEASES

Recent years have seen a strong interest in analyzing data with electronic patient health records (EHR). Early detection of preventable diseases is essential for better disease management, improved interventions, and effective allocation of health care resources. AAR is one of the key carriers for the success of the healthcare revolution with this data. There are many challenges in working directly with EHR, such as physicality, noise, bias, etc. Many machine learning methods have been developed to use information in electronic health records (EHR) for this job. Most of the previous attempts focused on structured fields and lost a large amount of information in unstructured records. To overcome these drawbacks, this paper uses novel machine learning approach with a general multi-task framework for disease onset prediction that combines both unstructured and structured information. To understand the predictive performance of this approach several performance metrics are used. The experimental result shows that the proposed method provides a superior predictive effect than other classifier.