Data Quality on KDD: a Real-life Scenario

Abstract. The growing di↵usion of IT-based services generates a lot of data useful for supporting the activities of firms, organisations, and state agencies. In such a context, data quality tasks are frequently addressed using cleansing routines, often framed in the wider context of ETL processes (Extraction, Transformation, and Loading). The design of these cleansing routines often relies on the experience of domain-experts, and this makes the evaluation of the quality level achieved a relevant concern to ensure the believability of the analysed results. In this paper we describe two model based techniques aimed at respectively evaluating the consistency of a dataset and at identifying the cleansing alternatives. The techniques have been applied on a real-world dataset derived from the Italian labour market domain, which we made publicly available to the community.