Abstract

The Data Science team at the BELab Data Laboratory of Banco de España recently developed an interactive web application prototype to support data providers and lab technicians in the exploratory analysis and quality control of the micro datasets hosted by BELab. Despite the significant variations in size, nature, and level of confidentiality of these datasets, they share a common structure that facilitates the generalization and standardization of the quality control process to a large extent. The developed prototype provides a generic framework for building an integrated platform for quality assessment of complex statistical data. The primary advantage of the prototype lies in its generic nature, allowing for the standardized processing of a wide variety of datasets. Recently, the prototype has been extended to a broader collection of datasets from the Statistics Department of Banco de España, encompassing a wider variety of data types, including both categorical and numerical spatiotemporal variables describing multiple distinct entities. The initial prototype integrated basic traditional versions of each analysis module (exploratory data analysis, univariate and multivariate analysis and anomaly detection, etc.). The new version of the tool also incorporates more advanced versions of each module, focusing on state-of-the-art (SOTA) solutions for various analytical tasks, such as imputation, multivariate analysis handling both categorical and numerical variables, spatiotemporal analysis, and explainability. The tool was designed and implemented with modularity and ease of sharing as key requirements, to facilitate the addition of new capabilities and make it easier for potential users to adopt in various contexts.