Research on data mining and machine learning tasks are commonly developed under assumptions of uniform preferences, where cases are equally important, and issues such as data acquisition costs are not considered. However, many real-world data-mining applications involve complex settings where such assumptions do not apply. Frequently, predictive analytics involve settings where the consideration of costs is unavoidable. Such costs can appear at all stages of the data mining process, e.g. data acquisition, modelling or model application. In this workshop we will target tasks involving the consideration of costs and/or benefits which may arise from different sources.
The most studied setting regards binary classification tasks with costs applied at the evaluation level. In this case, different penalizations and/or benefits are assigned to different mistakes and/or accurate predictions, and a cost matrix is used to assess the performance of model. However, other settings may also be cost dependent such as regression and time series or data streams forecasting tasks. Moreover, there are other issues which, although relevant, are still unsolved or need improved solutions, such as performance evaluation and applications involving unsupervised and semi-supervised tasks.
Tackling the issues raised sby cost-sensitive learning problems is crucial to both academia and industry, as it allows the development of more suitable and robust systems for complex settings. For industry partners, this presents the opportunity to develop frameworks targeting specific contexts, embedding in the solutions the necessary domain knowledge. Examples include dealing with budgeted resources, limited space or computational time, prediction of rare events and anomaly detection.
This workshop will bring together practitioners and researchers from both academia and industry that are linked to all levels of cost-sensitive learning. This will promote a wider knowledge exchange as well as the interaction between different agents. Our workshop invites inter-disciplinary contributions to tackle the problems that many real-world domains face nowadays, in order to promote significant developments in this field.
The research topics of interest to COST'2018 workshop include (but are not limited to) the following:
Foundations of cost- and utility-based learning
Probabilistic and statistical models
New knowledge discovery theories and models
Deep learning in the context of cost-sensitive learning
Handling cost-sensitive big data
Learning with non i.i.d. data
Relations between cost/utility-based learning and data pre-processing/post-processing
Feature selection and feature transformation
Evaluation in cost-sensitive learning
Knowledge discovery and machine learning in cost and utility-based tasks
Classification, ordinal classification
Data streams and time series forecasting
Adaptive learning and algorithm-level approaches
Multi-label, multi-instance, sequence and association rules mining
Spatial and spatio-temporal learning
Applications of cost and utility-based learning
Fraud detection (e.g. finance, credit and online banking)
Anomaly detection (e.g. industry, intrusion detection)
Environmental applications (e.g. meteorology, biology)
Social media applications (e.g. popularity prediction, recommender systems)
Real world applications (e.g. oil spill detection)
Naoki Abe, IBM
Roberto Alejo, Tecnológico de Estudios Superiores de Jocotitlán
Colin Bellinger, University of Alberta
Seppe Vanden Broucke, Katholieke Universiteit Leuven
Nitesh Chawla, University of Notre Dame
Christopher Drummond, National Research Council Canada
Ines Dutra, DCC - FCUP
Tom Fawcett, Apple
Mikel Galar, Universidad Pública de Navarra
Nathalie Japkowicz, American University
Charles Ling, Western University
Dragos Margineantu, Boeing Research and Technology
Ronaldo Prati, Universidade Federal do ABC
Foster Provost, NYU Stern
Jose Hernandez-Orallo, Universitat Politècnica de València
Rita Ribeiro, LIAAD / INESC Tec
Shengli Victor Sheng, University of Central Arkansas
Marina Sokolova, University of Ottawa
All accepted papers will be included in the workshop proceedings, published as a volume in Proceedings of Machine Learning Research (PMLR).
Additionally, based on the success of the workshop, authors of selected papers may be invited to submit extended versions of their manuscripts to a premier journal concerning the topics of this workshop.