Predicting stroke using ant colony optimization algorithm / Azfaruddin Azri and Rizauddin Saian

Statistics from Malaysian government hospitals have revealed that there is an increase in stroke cases from year to year. Stroke illness detection requires additional work; however, it is not a simple process. Since the rule list for Ant-Miner is supposedly shorter than that of other rule induction techniques, this study employed it to predict stroke disease. Ant-Miner is an approach for ant colony optimization with data mining. The aim of this study is to develop a classification model for predicting stroke. Using WEKA as a tool, the data set is discretized by changing the numerical attributes to the nominal attributes. The dataset was then processed through the Gui Ant-Miner to discover the patterns and the degree of accuracy in predicting stroke condition. Later, J48 is used to compare the accuracy of Ant outputs Miner's in order to improve classification. To observe variations in accuracy, the number of rules, and the number of conditions, the dataset was run using a range of ants, from 50 to 250. When the minimal case per rule value was changed, rules and condition number were also observed. Due to the test's low bias and variance, the cross-validation number was set at k=10 times throughout. Other parameters, such as the maximum number of uncovered instances and the convergence rules were kept at 10 and 100 respectively.