Figure 6.

Joint application of ME and ML methods. We can obtain 80% correct classification on 32% of the data by using both methods together. Based on a mixture set of queries of all lengths from the S500R data set, EPA got 41% correct and Phyclass got 40% correct. They were both in agreement, correctly or incorrectly, on 32% of queries. Among the agreeing queries, the ratio of correct to incorrect queries is 4 to 1, that is, among these 32%, 80% are correct. In this way, by running the two classification methods in parallel, it is possible to roughly double the accuracy for about a third of the queries by running both methods in parallel.