Consider a scenario where a dataset (D) has 10 instances. Each of these instances is represented by 4 descriptive feature (F1, F2, F3 and F4) and has 1 target feature that takes binary values. Because the target feature is binary, only 1 bit is required to store the target value of each instance. Features F1, F2, F3 and F4 are all categorical, with 5, 4, 3, and 2 possible discrete values, respectively. The decision trees to be trained on D will be deployed and utilized to classify previously unseen instances on a high-performance computer, hence algorithm efficiency is not a concern. Given this context, which measure is the most appropriate for the ID3 algorithm to induce a decision tree using D?