@InProceedings{10.1007/978-3-031-87719-3_3,
author="McTeer, Matthew
and Missier, Paolo",
editor="Ordonez, Carlos
and Sperl{\`i}, Giancarlo
and Masciari, Elio
and Bellatreche, Ladjel",
title="Stacked Generalization for Overlapping Asymmetric Datasets",
booktitle="Model and Data Engineering",
year="2025",
publisher="Springer Nature Switzerland",
address="Cham",
pages="38--52",
abstract="In the context of training sets for Machine Learning, we use the term Overlapping Asymmetric Datasets (OADs) to refer to a combination of data shapes where a large number of observations (Vertical data, {\$}{\$}{\backslash}mathcal {\{}V{\}}{\$}{\$}V) are described using only few features (x), and a small subset of the observations (Horizontal data, {\$}{\$}{\backslash}mathcal {\{}H{\}}{\$}{\$}H) are described by a larger number of features (x plus some new z). A common example of such a combination is a healthcare dataset where the majority of patients are described using a baseline set of clinical and socio-demographic features, and a handful of those patients have a richer characterisation, having undergone further testing . Given a classification task, a model trained solely on {\$}{\$}{\backslash}mathcal {\{}H{\}}{\$}{\$}Hwill benefit from the many features, but its performance will be limited by a small training set size . In this paper we study the problem of maximising model performance on {\$}{\$}{\backslash}mathcal {\{}H{\}}{\$}{\$}H, by leveraging the additional information available from {\$}{\$}{\backslash}mathcal {\{}V{\}}{\$}{\$}V. Our approach is based on the notions of stacked generalization and meta-learning, where the predictions generated by an ensemble of weak classifiers for {\$}{\$}{\backslash}mathcal {\{}V{\}}{\$}{\$}Vare fed into a second-tier meta-learner, where the z features are also used. We conduct extensive experiments to explore the benefits of this approach over a range of dataset configurations. The results suggest that stacking improves model performance, while using z features only provides modest improvements. This may have practical implications as it suggests that in some settings, the effort involved in acquiring the additional z features is not always justified.",
isbn="978-3-031-87719-3"
}

{"_id":"BJBr5BCoffrNWr9X4","bibbaseid":"mcteer-missier-stackedgeneralizationforoverlappingasymmetricdatasets-2025","author_short":["McTeer, M.","Missier, P."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","author":[{"propositions":[],"lastnames":["McTeer"],"firstnames":["Matthew"],"suffixes":[]},{"propositions":[],"lastnames":["Missier"],"firstnames":["Paolo"],"suffixes":[]}],"editor":[{"propositions":[],"lastnames":["Ordonez"],"firstnames":["Carlos"],"suffixes":[]},{"firstnames":[],"propositions":[],"lastnames":["Sperlì, Giancarlo"],"suffixes":[]},{"propositions":[],"lastnames":["Masciari"],"firstnames":["Elio"],"suffixes":[]},{"propositions":[],"lastnames":["Bellatreche"],"firstnames":["Ladjel"],"suffixes":[]}],"title":"Stacked Generalization for Overlapping Asymmetric Datasets","booktitle":"Model and Data Engineering","year":"2025","publisher":"Springer Nature Switzerland","address":"Cham","pages":"38–52","abstract":"In the context of training sets for Machine Learning, we use the term Overlapping Asymmetric Datasets (OADs) to refer to a combination of data shapes where a large number of observations (Vertical data, $}{$\\mathcal \\V\\$}{$V) are described using only few features (x), and a small subset of the observations (Horizontal data, $}{$\\mathcal \\H\\$}{$H) are described by a larger number of features (x plus some new z). A common example of such a combination is a healthcare dataset where the majority of patients are described using a baseline set of clinical and socio-demographic features, and a handful of those patients have a richer characterisation, having undergone further testing . Given a classification task, a model trained solely on $}{$\\mathcal \\H\\$}{$Hwill benefit from the many features, but its performance will be limited by a small training set size . In this paper we study the problem of maximising model performance on $}{$\\mathcal \\H\\$}{$H, by leveraging the additional information available from $}{$\\mathcal \\V\\$}{$V. Our approach is based on the notions of stacked generalization and meta-learning, where the predictions generated by an ensemble of weak classifiers for $}{$\\mathcal \\V\\$}{$Vare fed into a second-tier meta-learner, where the z features are also used. We conduct extensive experiments to explore the benefits of this approach over a range of dataset configurations. The results suggest that stacking improves model performance, while using z features only provides modest improvements. This may have practical implications as it suggests that in some settings, the effort involved in acquiring the additional z features is not always justified.","isbn":"978-3-031-87719-3","bibtex":"@InProceedings{10.1007/978-3-031-87719-3_3,\nauthor=\"McTeer, Matthew\nand Missier, Paolo\",\neditor=\"Ordonez, Carlos\nand Sperl{\\`i}, Giancarlo\nand Masciari, Elio\nand Bellatreche, Ladjel\",\ntitle=\"Stacked Generalization for Overlapping Asymmetric Datasets\",\nbooktitle=\"Model and Data Engineering\",\nyear=\"2025\",\npublisher=\"Springer Nature Switzerland\",\naddress=\"Cham\",\npages=\"38--52\",\nabstract=\"In the context of training sets for Machine Learning, we use the term Overlapping Asymmetric Datasets (OADs) to refer to a combination of data shapes where a large number of observations (Vertical data, {\\$}{\\$}{\\backslash}mathcal {\\{}V{\\}}{\\$}{\\$}V) are described using only few features (x), and a small subset of the observations (Horizontal data, {\\$}{\\$}{\\backslash}mathcal {\\{}H{\\}}{\\$}{\\$}H) are described by a larger number of features (x plus some new z). A common example of such a combination is a healthcare dataset where the majority of patients are described using a baseline set of clinical and socio-demographic features, and a handful of those patients have a richer characterisation, having undergone further testing . Given a classification task, a model trained solely on {\\$}{\\$}{\\backslash}mathcal {\\{}H{\\}}{\\$}{\\$}Hwill benefit from the many features, but its performance will be limited by a small training set size . In this paper we study the problem of maximising model performance on {\\$}{\\$}{\\backslash}mathcal {\\{}H{\\}}{\\$}{\\$}H, by leveraging the additional information available from {\\$}{\\$}{\\backslash}mathcal {\\{}V{\\}}{\\$}{\\$}V. Our approach is based on the notions of stacked generalization and meta-learning, where the predictions generated by an ensemble of weak classifiers for {\\$}{\\$}{\\backslash}mathcal {\\{}V{\\}}{\\$}{\\$}Vare fed into a second-tier meta-learner, where the z features are also used. We conduct extensive experiments to explore the benefits of this approach over a range of dataset configurations. The results suggest that stacking improves model performance, while using z features only provides modest improvements. This may have practical implications as it suggests that in some settings, the effort involved in acquiring the additional z features is not always justified.\",\nisbn=\"978-3-031-87719-3\"\n}\n\n\n\n","author_short":["McTeer, M.","Missier, P."],"editor_short":["Ordonez, C.","Sperlì, Giancarlo","Masciari, E.","Bellatreche, L."],"key":"10.1007/978-3-031-87719-3_3","id":"10.1007/978-3-031-87719-3_3","bibbaseid":"mcteer-missier-stackedgeneralizationforoverlappingasymmetricdatasets-2025","role":"author","urls":{},"metadata":{"authorlinks":{}}},"bibtype":"inproceedings","biburl":"https://bibbase.org/f/MTSG9SdhWPisKNpZX/MyPublications-bibbase.bib","dataSources":["ze2X9uz8Dcv2oGipf","afppXLgSuddAzAL9e"],"keywords":[],"search_terms":["stacked","generalization","overlapping","asymmetric","datasets","mcteer","missier"],"title":"Stacked Generalization for Overlapping Asymmetric Datasets","year":2025}