Knowledge discovery for health risk prediction

Knowledge discovery for health risk prediction

Improving the accuracy of the diagnosis of disease can help to increase the quality of healthcare. Many researchers have developed classification models to support healthcare practitioners to make accurate diagnoses, avoiding the need to rely on their experience base diagnose diseases. However, these models are currently based on datasets collected from healthcare data including medical history. As a result, the reliability and accuracy of predicting results of the diagnosis, are limited.

Following the goal of improving the accuracy of health risk prediction, this thesis concentrates on the classification of tasks through mining healthcare data. The study suggests several frameworks and algorithms to develop classification models. In addition, challenges of extracting useful information and processing data noise from the real dataset are addressed as a way of learning models. Classification models are developed based on well-proven medical data sources. By using medical evidence, the study aims to improve the accuracy of classification for health risk prediction.

The first contribution of this thesis is an innovation of building a binary classification model to predict patients’ risks. The second contribution of this dissertation is to build a medical knowledge base to support classification models for improving the reliability and accuracy of the model. The third significant contribution of the thesis provides a framework for building a predictive model within multiple diseases.