The Ionosphere Data Set Sample Clauses
The Ionosphere Data Set. On the Ionosphere data set (Table 4.6) we see again that fuzzification has a negative effect on classification performance when the performance of the original Boolean gp algorithms is quite good. Only in the case of our clus- tering gp algorithm using k = 2 does fuzzification significantly improve the performance. In the case of the gain gp and gain ratio gp algorithms fuzzi- fication significantly decreases the classification performance.
The Ionosphere Data Set. When we consider the results of our gp algorithms on the Ionosphere data set in Table 3.9 we see that our gain ratio gp algorithms perform the best. Together with the gain gp algorithm with k = 2 they are significantly better than our other gp algorithms. If we look at the results of our new gp algo- rithms with respect to k we see that the misclassification rate increases with k, except for our clustering gp with k = 2. Our three best gain ratio gp algorithms are also significantly better than OC1. Compared to the results of Ltree and C4.5 the differences in perfor- mance with our new gp algorithms are not statistically significant, except for the worst performing gp algorithm, clustering gp with k = 5. The differ- ences between cefr-miner and our new algorithms are also no statistically significant.
The Ionosphere Data Set. The Ionosphere data set contains information of radar returns from the iono- sphere. According to [7] the data was collected by a phased array of 16 high-frequency antennas with a total transmitted power in the order of 6.4 kilowatts. The target class consists of the type of radar return. A “good” radar return shows evidence of some type of structure of electrons in the ≈ × ionosphere while a “bad” return does not. Although the number of records is quite small (351) the number of attributes is the largest (34) of the data sets on which we have tested our algorithms. All attributes are continuous valued. Because our simple gp constructs a node for each possible value of continuous valued attributes it constructs no less than 8147 possible internal nodes as well as 2 terminal nodes for the target classes. This results in a search space of size 1.1 10147. One fold consists of 36 records while the other 9 folds consist of 35 records each.