Ionosphere: Johns Hopkins University Ionosphere database

IonosphereR Documentation

Johns Hopkins University Ionosphere database


This radar data was collected by a system in Goose Bay, Labrador. This system consists of a phased array of 16 high-frequency antennas with a total transmitted power on the order of 6.4 kilowatts. See the paper for more details. The targets were free electrons in the ionosphere. "good" radar returns are those showing evidence of some type of structure in the ionosphere. "bad" returns are those that do not; their signals pass through the ionosphere.

Received signals were processed using an autocorrelation function whose arguments are the time of a pulse and the pulse number. There were 17 pulse numbers for the Goose Bay system. Instances in this databse are described by 2 attributes per pulse number, corresponding to the complex values returned by the function resulting from the complex electromagnetic signal. See cited below for more details.


data("Ionosphere", package = "mlbench")


A data frame with 351 observations on 35 independent variables, some numerical and 2 nominal, and one last defining the class.


  • Source: Space Physics Group; Applied Physics Laboratory; Johns Hopkins University; Johns Hopkins Road; Laurel; MD 20723

  • Donor: Vince Sigillito (

These data have been taken from the UCI Repository Of Machine Learning Databases (Blake & Merz 1998) and were converted to R format by Evgenia Dimitriadou in the late 1990s.

The current version of the UC Irvine Machine Learning Repository Ionosphere data set is available from \Sexpr[results=rd]{tools:::Rd_expr_doi("10.24432/C5W01B")}.


Sigillito, V. G., Wing, S. P., Hutton, L. V., and Baker, K. B. (1989). Classification of radar returns from the ionosphere using neural networks. Johns Hopkins APL Technical Digest, 10, 262-266.

They investigated using backprop and the perceptron training algorithm on this database. Using the first 200 instances for training, which were carefully split almost 50% positive and 50% negative, they found that a "linear" perceptron attained 90.7%, a "non-linear" perceptron attained 92%, and backprop an average of over 96% accuracy on the remaining 150 test instances, consisting of 123 "good" and only 24 "bad" instances. (There was a counting error or some mistake somewhere; there are a total of 351 rather than 350 instances in this domain.) Accuracy on "good" instances was much higher than for "bad" instances. Backprop was tested with several different numbers of hidden units (in [0,15]) and incremental results were also reported (corresponding to how well the different variants of backprop did after a periodic number of epochs).

David Aha ( briefly investigated this database. He found that nearest neighbor attains an accuracy of 92.1%, that Ross Quinlan's C4 algorithm attains 94.0% (no windowing), and that IB3 (Aha and Kibler, IJCAI-1989) attained 96.7% (parameter settings: 70% and 80% for acceptance and dropping respectively).

Blake, C.L. & Merz, C.J. (1998). UCI Repository of Machine Learning Databases. Irvine, CA: University of California, Irvine, Department of Information and Computer Science. Formerly available from ‘⁠⁠’.


data("Ionosphere", package = "mlbench")

mlbench documentation built on May 29, 2024, 4:49 a.m.

Related to Ionosphere in mlbench...