Knowledge Discovery from Hospital Outpatient Medical Database

Abstract

Knowledge discovery from hospital outpatient medical databases is a health information analytics research area where data MEKA and WEKA are employed as important tools to analyze medical databases. Recently, researchers have been interested in applying data mining tools in the field of medical sciences for decision support to discover new sets of information. As clinical medical records become more popular, the quantity of data collected increases, with most of it unanalyzed. A reasonable 300 outpatient medical records were collected manually from the hospital with eighteen (18) attributes. The processed data was fed into MEKA/WEKA for data analysis using Microsoft Excel. The classification techniques used are J48, Random Tree, and Random Forest, while the clustering technique uses a simple k-mean algorithm. The performance evaluations for classification were hamming score, exact match, hamming loss, zero-on loss, and accuracy, while the clustering technique used confidence, lift, leverage, and conviction as metrics.The experiments conducted showed that the random tree performed better with a hamming score of 0.108, an accuracy of 0.144, and a total time taken to run the classifier was 0.05 seconds. This work establishes decision rules that would be helpful for medical practitioners in predicting prescriptions and categorizing patient medical databases for management planning. These research findings prove that the random tree algorithm has the potential to be the most effective and efficient technique for medical data mining. It will be of great benefit to the medical sector and further research.