6. Experimental results
We have implemented our algorithm and used standard datasets to obtain experimental results. We have collected
the usual and known datasets from Holte [37]. The 16 problems that Holte used to compare his 1R method with the
C4.5 method of Quinlan [22] were used in our work. This set of problems aims to represent all real-life problems.
Holte claimed that the experimental results obtained for these problems by learning systems could shed light on
the performance of each learning system in a real situation. All the selected problems comply with two conditions:
representing a real-life problem that has not been constructed artificially; and the examples are described by means
of attributes used naturally in real life. The only dataset that complies with the second condition is CH. This dataset
represents the endgames in chess. According to Holte, this dataset is designed to fit well in Quinlan’s ID3.