Granularity-driven sequential three-way decisions: A cost-sensitive approach to classification.

•We propose a granularity-driven sequential three-way decision (S3WD) model for classification problem.•A sequence of multi-level granularity is constructed for S3WD.•A cost structure as evaluation metric for S3WD is built to interpret the granularity-driven S3WD.•12 UCI data sets are used to validate the model in two application scenarios.•Comparison study shows that granularity-driven sequential three-way decision model is a reasonable and meaningful approach for practical applications where relevant information is not immediately available and is costly to acquire.