HIS'2008: Artificial Data Sets based on Knowledge Generators: Analysis of Learning Algorithms Efficiency

The document proposes a framework for evaluating machine learning algorithms using artificially generated data sets based on predefined knowledge. It defines a methodology that involves sampling problems, generating data sets using different sampling methods and sizes, and running learning algorithms with varying parameters. Results are analyzed by plotting accuracy and optimal population over iterations to compare algorithm efficiency under different conditions. The methodology allows reducing the number of iterations needed for evaluation.