/**
* Compute the best information gain for the specified antecedent
*
* @param instances the data based on which the infoGain is computed
* @param defAcRt the default accuracy rate of data
* @param antd the specific antecedent
* @param numConds the number of antecedents in the rule so far
* @return the data covered by the antecedent
*/
private Instances computeInfoGain(Instances instances, double defAcRt, Antd antd) {
Instances data = instances;
/* Split the data into bags.
The information gain of each bag is also calculated in this procedure */
Instances[] splitData = antd.splitData(data, defAcRt, m_Consequent);
/* Get the bag of data to be used for next antecedents */
if (splitData != null) return splitData[(int) antd.getAttrValue()];
else return null;
}
