An intelligent predictive model for knowledge discovery from clinical datasets

Abstract

An Intelligent prediction model for the discovery of knowledge from clinical datasets is proposed and evaluated. The model has been specifically designed to be used by physicians as an aid for clinical decision making. The features supported by the model include extracting knowledge from the clinical datasets and representing the extracted knowledge in the form of rules or network parameters. The first component this research work focuses on techniques for extracting the knowledge from hepatitis data. This model has been tailored to include normalization, principal component analysis and fuzzy c-means clustering in the pre-mining subsystem. The second component extracted knowledge from time-series hepatitis data. The pre-mining subsystem processed the data to represent the variations of the data values with respect to the date of examination. The system extracted knowledge from hepatitis data using the techniques association rule mining algorithm, decision tree algorithm and neural network. The third component extracted knowledge from heart disease data. The pre-mining subsystem discretized the continuous valued attributes using entropy based discretization. The fourth component was also used to extract knowledge from heart disease data. The pre-mining subsystem was implemented to fuzzify the continuous valued attributes. A survey had been carried out on the studies carried out to identify correlation among cardiovascular, diabetes, hepatitis and anemia (renal). Some of the studies reveal that hepatitis is more prevalent among hemodialysis patients. Studies reveal that hepatitis is associated with diabetes; some of the studies identify positive correlation between them whereas others a negative correlation. Patient with hepatitis infection was found to have different cardiovascular risks when compared to a non infected patient. Studies show that diabetic patients are more prone to have cardiovascular risks.