Glossary of Frequently Used and Exotic Terms

advisor

: $\rightarrow$ data mining advisor

base database

: a database that is used for experimentation to obtain $\rightarrow$ metadata

data characteristics

: the collection of measurements obtained for a base database by the $\rightarrow$ DCT program and $\rightarrow$ landmarkers. A subset of these characteristics are used as meta-data.

database measurements

: $\rightarrow$ Data characteristics obtained by the $\rightarrow$ DCT program.

DCT program

: A program that calculates many different database measurements from a database. For more information on that program see [DCT doc].

data file

: A comma separated variables (CSV) file that contains the actual data for a database. Each line contains one database entry as a comma separated list of ASCII values. See Section 4 for details.

data mining advisor

: A web-based application that uses meta-data obtained with METAL-MLEE to build a model that will give algorithm ranking recommendations for new databases.

experiment

: The process of carrying out a complete run of evaluation steps for all learning algorithms for one base database.

filestem

: the common part of the the two files (the "data-" and the "names" - file) that together are used to describe a database. This is the filename without the file extension. See Section 4.

interface script

: A script that makes a learning algorithm program usable with the main METAL-MLEE experimentation program, run_exp.

landmark

: A $\rightarrow$ database characteristic calculated by running a fast learning algorithm on the database.

meta-data

: a collection of data describing $\rightarrow$ base databases and the performance of learning algorithms on these base databases.

meta-database

: a collection of meta-data that is used for $\rightarrow$ meta-learning

ranking

: A recommendation of the $\rightarrow$ data mining advisor is given as a ranking: a ranked list of algorithms - the most recommended first, the least recommended last.

results files

: The collection of all files that are generated as a result of an experiment: the .stats file, the .results file, the .dct file, and others (see Section 8).

.result file

: One of the files that gets generated during an experiment. The file name extension of that file is .results, hence the name.