Fiche du document
7 septembre 2021
- handle: 10670/1.ecanhw
- isbn: 978-3-030-84788-3
- hal: hal-04175112
- doi: 10.1007/978-3-030-84789-0_18
Ce document est lié à :
info:eu-repo/semantics/altIdentifier/doi/10.1007/978-3-030-84789-0_18
http://creativecommons.org/licenses/by/
Mots-clés
Linked data Open Government Data Machine learning Artificial intelligence XAI SHAP XGBoostSujets proches
Intellectronics AI (Artificial intelligence) Artificial thinking Electronic brains Intelligent machines Intelligence, Artificial Machine intelligence Thinking, Artificial IACiter ce document
Evangelos Kalampokis et al., « Applying Explainable Artificial Intelligence Techniques on Linked Open Government Data », HAL-SHS : sciences de l'information, de la communication et des bibliothèques, ID : 10.1007/978-3-030-84789-0_18
Métriques
Partage / Export
Résumé
Machine learning and artificial intelligence models have the potential to streamline public services and policy making. Frequently, however, the patterns a model uncovers can be more important than the model’s performance. Explainable Artificial Intelligence (XAI) have been recently introduced as a set of techniques that enable explaining individual decisions made by a model. Although XAI has been proved important in various domains, the need of using relevant techniques in public administration has only recently emerged. The objective of this paper is to explore the value and the feasibility of creating XAI models using high quality open government data that are provided in the form of linked open statistical data. Towards this end, a process for exploiting linked open statistical data in the creation of explainable models is presented. Moreover, a case study where linked data from the Scottish open statistics portal is exploited in order to predict and interpret the probability the mean house price of a data zone to be higher than the average price in Scotland is described. The XGBoost algorithm is used to create the predictive model and the SHAP framework to explain it.