{"id":{"repo_id":"regina","oai_identifier":"oai:uregina.scholaris.ca:10294/8883"},"canonical_url":"https://search.dev.ndltd.org/etd/regina/oai:uregina.scholaris.ca:10294/8883","repository":{"repo_id":"regina","name":"University of Regina","base_url":"https://uregina.scholaris.ca/server/oai/request"},"display":{"title":"Supervised Classification of Imbalanced Bidding Fraud Data","abstract":"Online auctions have become one of the most convenient ways to commit fraud since a huge amount of money is being invested everyday by thousands of customers. Hence, online auctions are vulnerable to several types of fraud. Shill Bidding (SB), a predominant auction fraud, is the toughest to identify due to its resemblance to the normal bidding behavior. An innocent bidder can easily be cheated and lose money if shill bidders are not identified. Our goal is to devise a SB classification model, which is able to efficiently differentiate between legitimate bidders and shill bidders. For this thesis, we employ a real SB training dataset, which is unlabeled. First, we label the SB dataset by the help of hierarchical clustering with an optimal number of clusters, combining it with a semi-automated labeling approach. Second, we assess and compare several advanced over-sampling (SMOTE), under-sampling (NearMiss and ClusterCentroid) and hybrid sampling (SMOTE-ENN and SMOTE-TomekLink) methods to solve the imbalanced learning problem. We utilize the Randomized Search Cross Validation to tune the hyper-parameters for Support Vector Machine (SVM), Random Forest and default parameters for Artificial Neural Network (ANN) with MLP classifier, finally, to obtain the optimal classifier for our SB dataset. Therefore, we develop eighteen fraud classifiers, including fifteen classifiers using sampling techniques and three classifiers without sampling. We demonstrate that the optimal SB classifier exhibits very satisfactory testing performance for detecting and misclassifying shill bidders. The hybrid sampling method SMOTE-ENN combined with the SVM has turned out to be the most performing classifier. Also, every sampled SB dataset greatly improved the classification performance of all the fraud classifiers than the imbalanced SB dataset.","abstract_html":"Online auctions have become one of the most convenient ways to commit fraud since a huge amount of money is being invested everyday by thousands of customers. Hence, online auctions are vulnerable to several types of fraud. Shill Bidding (SB), a predominant auction fraud, is the toughest to identify due to its resemblance to the normal bidding behavior. An innocent bidder can easily be cheated and lose money if shill bidders are not identified. Our goal is to devise a SB classification model, which is able to efficiently differentiate between legitimate bidders and shill bidders. For this thesis, we employ a real SB training dataset, which is unlabeled. First, we label the SB dataset by the help of hierarchical clustering with an optimal number of clusters, combining it with a semi-automated labeling approach. Second, we assess and compare several advanced over-sampling (SMOTE), under-sampling (NearMiss and ClusterCentroid) and hybrid sampling (SMOTE-ENN and SMOTE-TomekLink) methods to solve the imbalanced learning problem. We utilize the Randomized Search Cross Validation to tune the hyper-parameters for Support Vector Machine (SVM), Random Forest and default parameters for Artificial Neural Network (ANN) with MLP classifier, finally, to obtain the optimal classifier for our SB dataset. Therefore, we develop eighteen fraud classifiers, including fifteen classifiers using sampling techniques and three classifiers without sampling. We demonstrate that the optimal SB classifier exhibits very satisfactory testing performance for detecting and misclassifying shill bidders. The hybrid sampling method SMOTE-ENN combined with the SVM has turned out to be the most performing classifier. Also, every sampled SB dataset greatly improved the classification performance of all the fraud classifiers than the imbalanced SB dataset.","abstract_has_math":false,"creators":["Anowar, Farzana"],"institution":"Faculty of Graduate Studies and Research, University of Regina","degree_name":"Master of Science (MSc)","degree_level":"Master's","degree_discipline":"Computer Science","degree_department":null,"school":null,"contributors":[],"advisors":["Sadaoui, Samira","Mouhoub, Malek"],"committee_chairs":[],"committee_members":["Louafi, Habib"],"year":2018,"date_issued":"2018-12","date_published":"2018-12","updated_at":"2026-07-24T04:03:34Z","subjects":[],"languages":["en"],"rights":[],"rights_urls":[],"identifier_entries":[{"key":"dc:identifier.doi","label":"DOI","values":["https://doi.org/10.82465/4216"],"render_values":[{"text":"https://doi.org/10.82465/4216","href":"https://doi.org/10.82465/4216","code":true}]}]},"links":{"outbound_url":"https://hdl.handle.net/10294/8883","outbound_label":"Handle","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Sadaoui, Samira","Mouhoub, Malek"]},{"key":"dc:contributor.committeemember","label":"Committee Member","values":["Louafi, Habib"]},{"key":"dc:creator","label":"Author","values":["Anowar, Farzana"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2019-06-21T19:58:53Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2019-06-21T19:58:53Z"]},{"key":"dc:date.issued","label":"Date","values":["2018-12"]},{"key":"dc:publisher","label":"Institution","values":["Faculty of Graduate Studies and Research, University of Regina"]},{"key":"dc:type","label":"Dc Type","values":["master thesis"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Computer Science"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Master's"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Master of Science (MSc)"]},{"key":"thesis:institution_name","label":"Thesis Institution Name","values":["Faculty of Graduate Studies and Research, University of Regina"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language.iso","label":"Language (ISO)","values":["en"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.doi","label":"DOI","values":["https://doi.org/10.82465/4216"]},{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://hdl.handle.net/10294/8883"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["A Thesis Submitted to the Faculty of Graduate Studies and Research In Partial Fulfillment of the Requirements for the Degree of Master of Science in Computer Science, University of Regina. viii, 80 p."]},{"key":"dc:description.abstract","label":"Abstract","values":["Online auctions have become one of the most convenient ways to commit fraud since a huge amount of money is being invested everyday by thousands of customers. Hence, online auctions are vulnerable to several types of fraud. Shill Bidding (SB), a predominant auction fraud, is the toughest to identify due to its resemblance to the normal bidding behavior. An innocent bidder can easily be cheated and lose money if shill bidders are not identified. Our goal is to devise a SB classification model, which is able to efficiently differentiate between legitimate bidders and shill bidders. For this thesis, we employ a real SB training dataset, which is unlabeled. First, we label the SB dataset by the help of hierarchical clustering with an optimal number of clusters, combining it with a semi-automated labeling approach. Second, we assess and compare several advanced over-sampling (SMOTE), under-sampling (NearMiss and ClusterCentroid) and hybrid sampling (SMOTE-ENN and SMOTE-TomekLink) methods to solve the imbalanced learning problem. We utilize the Randomized Search Cross Validation to tune the hyper-parameters for Support Vector Machine (SVM), Random Forest and default parameters for Artificial Neural Network (ANN) with MLP classifier, finally, to obtain the optimal classifier for our SB dataset. Therefore, we develop eighteen fraud classifiers, including fifteen classifiers using sampling techniques and three classifiers without sampling. We demonstrate that the optimal SB classifier exhibits very satisfactory testing performance for detecting and misclassifying shill bidders. The hybrid sampling method SMOTE-ENN combined with the SVM has turned out to be the most performing classifier. Also, every sampled SB dataset greatly improved the classification performance of all the fraud classifiers than the imbalanced SB dataset."]},{"key":"dc:title","label":"Title","values":["Supervised Classification of Imbalanced Bidding Fraud Data"]}]}],"canonical_facts":{"dc:contributor.advisor":["Sadaoui, Samira","Mouhoub, Malek"],"dc:contributor.committeemember":["Louafi, Habib"],"dc:creator":["Anowar, Farzana"],"dc:date.accessioned":["2019-06-21T19:58:53Z"],"dc:date.available":["2019-06-21T19:58:53Z"],"dc:date.issued":["2018-12"],"dc:description":["A Thesis Submitted to the Faculty of Graduate Studies and Research In Partial Fulfillment of the Requirements for the Degree of Master of Science in Computer Science, University of Regina. viii, 80 p."],"dc:description.abstract":["Online auctions have become one of the most convenient ways to commit fraud since a huge amount of money is being invested everyday by thousands of customers. Hence, online auctions are vulnerable to several types of fraud. Shill Bidding (SB), a predominant auction fraud, is the toughest to identify due to its resemblance to the normal bidding behavior. An innocent bidder can easily be cheated and lose money if shill bidders are not identified. Our goal is to devise a SB classification model, which is able to efficiently differentiate between legitimate bidders and shill bidders. For this thesis, we employ a real SB training dataset, which is unlabeled. First, we label the SB dataset by the help of hierarchical clustering with an optimal number of clusters, combining it with a semi-automated labeling approach. Second, we assess and compare several advanced over-sampling (SMOTE), under-sampling (NearMiss and ClusterCentroid) and hybrid sampling (SMOTE-ENN and SMOTE-TomekLink) methods to solve the imbalanced learning problem. We utilize the Randomized Search Cross Validation to tune the hyper-parameters for Support Vector Machine (SVM), Random Forest and default parameters for Artificial Neural Network (ANN) with MLP classifier, finally, to obtain the optimal classifier for our SB dataset. Therefore, we develop eighteen fraud classifiers, including fifteen classifiers using sampling techniques and three classifiers without sampling. We demonstrate that the optimal SB classifier exhibits very satisfactory testing performance for detecting and misclassifying shill bidders. The hybrid sampling method SMOTE-ENN combined with the SVM has turned out to be the most performing classifier. Also, every sampled SB dataset greatly improved the classification performance of all the fraud classifiers than the imbalanced SB dataset."],"dc:identifier.doi":["https://doi.org/10.82465/4216"],"dc:identifier.uri":["https://hdl.handle.net/10294/8883"],"dc:language.iso":["en"],"dc:publisher":["Faculty of Graduate Studies and Research, University of Regina"],"dc:title":["Supervised Classification of Imbalanced Bidding Fraud Data"],"dc:type":["master thesis"],"thesis:degree_discipline":["Computer Science"],"thesis:degree_level":["Master's"],"thesis:degree_name":["Master of Science (MSc)"],"thesis:institution_name":["Faculty of Graduate Studies and Research, University of Regina"]},"updated_at":"2026-07-24T04:03:34Z"}