{"doi":"10.1145/3712255.3726556","title":"Evaluating the Generalizability of Machine Learning Pipelines When Using Lexicase or Tournament Selection","abstract":"Evolutionary Algorithms have been successfully applied in Automated Machine Learning (AutoML) to evolve effective machine learning (ML) pipelines. Here, we use 12 OpenML classification tasks and the AutoML tool TPOT2 to assess the impact of lexicase and tournament selection on the generalizability of pipelines. We use one of five stratified sampling splits to generate training and validation sets; pipelines are trained on the training set, and predictions are made on the validation set. Lexicase and tournament selection use these predictions to identify parents. At the end of a run, TPOT2 returns the pipeline that achieved the best validation accuracy while maintaining the lowest complexity. The generalizability of this pipeline is assessed using the test set provided for an OpenML task. We found that lexicase produced pipelines with higher validation accuracy than tournament selection in all tasks for at least one split. In contrast, tournament selection produced pipelines with greater generalizability for 10 of the 12 tasks on at least one split. For most cases where tournament selection outperformed lexicase on test accuracy, we detected differences in validation accuracy and pipeline complexity.","journal":"Proceedings of the Genetic and Evolutionary Computation Conference Companion","year":2025,"id":558779,"datarank":0.0,"base_score":0.0,"endowment":0.0,"self_citation_contribution":0.0,"citation_network_contribution":0.0,"self_endowment_contribution":0.0,"citer_contribution":0.0,"corpus_percentile":null,"corpus_rank":null,"citation_count":1,"citer_count":0,"citers_with_citation_signal":0,"citers_with_endowment":0,"datacite_reuse_total":0,"is_dataset":false,"is_dataset_confidence":0.9446,"is_data_producer":false,"deposit_databanks":null,"is_oa":true,"file_count":0,"downloads":0,"has_version_chain":false,"published_date":"2025-01-01","fair_score":null,"fair_percentile":null,"algorithm_id":"datarank_citation_only_1hop_v6","ranking_scope":"data_only","authors":[{"id":1333365,"name":"Anil Kumar Saini","orcid":"0000-0002-9211-1079","position":1,"is_corresponding":false},{"id":1459585,"name":"Ankit Gupta","orcid":"0009-0001-0950-5983","position":2,"is_corresponding":false},{"id":14812,"name":"Jason H. Moore","orcid":"0000-0002-5015-1099","position":3,"is_corresponding":false},{"id":1333364,"name":"Jose Guadalupe Hernandez","orcid":"0000-0002-1298-5551","position":0,"is_corresponding":true}],"reference_count":9,"raw_metadata":null,"created_at":"2026-07-19T02:55:30.312295Z","pmid":null,"pmcid":null,"fwci":null,"citation_percentile":null,"influential_citations":0,"oa_status":null,"license":null,"views":0,"total_file_size_bytes":0,"version_count":0,"fair_f":null,"fair_a":null,"fair_i":null,"fair_r":null,"fair_zscore":null,"fair_rationale":null,"fair_model":null,"fair_agent_version":null,"fair_fulltext_source":null,"fair_has_llm":null,"fair_computed_at":null,"clinical_trials":[],"software_tools":[],"db_accessions":[],"linked_datasets":[],"topics":[]}