Knowledge discovery of decision table based on support vector machine

We describe how the support vector machine (SVM) technique can be applied to the knowledge discovery of decision table. In the case of decision rules acquisition, attribute set is reduced and characteristic samples are extracted by using SVM, and decision rules are then acquired based on a smaller number of support samples which could represent all samples. As a result, rules acquisition becomes faster and easier. In the case of class forecast, the samples of decision table are classified by using SVM, and a simple decision function is obtained. This decision function could forecast the sample´s class and act as decision rules. It is another kind of knowledge expression. Experiments indicate that our method is simple and feasible, while it performs faster. Results also show that it has better performance for large decision table.