Finely tuned tool helps pick datasets for testing recommendation software
FINALLY: A Dataset Recommender System for Recommender-Systems Research
Information Retrieval
Summary
Choosing the right collections of data is important when testing software that suggests things like movies or products. The authors created FINALLY, an online tool that helps researchers pick groups of datasets matching specific rules and goals. They tested FINALLY with many different setups and showed it reliably makes selections that meet the desired criteria and follow certain patterns. However, the results don’t prove these choices are the absolute best or more useful in real-world situations.
recommender systemdataset selectionoffline evaluationmetadata filterdiversity strategyconfigurationrandom selectioneffective covarianceconvex hullreproducibility
Authors
Louis Owie
Abstract
Dataset selection shapes the empirical conditions under which recommender-system algorithms are evaluated, yet existing tools provide limited support for constructing complete dataset sets that jointly satisfy experimental constraints and set-level selection objectives. To address this problem, I developed FINALLY, a web-based dataset recommender for constructing configurable dataset sets for offline recommender-systems evaluations. FINALLY combines required datasets, candidate-pool restrictions, metadata filters, configurable target-set sizes, Random selection, and diverse and non-diverse strategies based on adapted Effective Covariance and Convex Hull objectives. I evaluated FINALLY through 420 recommendation runs across ten systematically varied configurations. All evaluated dataset sets satisfied the applicable target-size, duplicate-avoidance, snapshot-membership, required-dataset, and metadata-filter requirements. All 40 deterministic strategy--configuration combinations were reproducible. Both the Effective-Covariance-based and Convex-Hull-based strategies produced the expected diverse-versus-non-diverse score ordering in all ten configurations. Under their corresponding objectives, the diverse strategies produced scores above all 30 configuration-specific Random results, whereas the non-diverse strategies produced scores below all 30 Random results. These results establish technical consistency for the evaluated FINALLY workflow and show that the implemented strategies follow their intended optimization directions within the investigated configuration space. They do not establish the scientific suitability, global optimality, or practical superiority of the generated selections.