Evaluation and Extension of Inductive Venn-Abers Predictive Distribution

Ilia Nouretdinov, James Gammerman, Daljit Rehal

Research output: Other contribution

Abstract

In this work the Inductive Venn-Abers Predictive Distribution (IVAPD) regression al- gorithm was reviewed and developed.
The method was first outlined in the conference proceeding paper (Nouretdinov et al., 2018) where classical Venn Prediction was extended to the regression problems.
The contribution of this work is the algorithm that allows combination of more than one underlying method. It has been shown with the Conformal Predictions method that way allows to obtain a better accuracy of prediction (Toccaceli and Gammerman (2019)), and another aggregation approach was developed for Venn-Abers prediction.
In addition, more attention was paid to an evaluation, considering Continuous Ranked Probability Score (CRPS) as the standard scoring rule.
In the application, the data sets were related to the areas of predictive maintenance and energy consumption.
Original languageEnglish
TypePoster presented at conference
Media of outputposter
PublisherProceedings of the Ninth Symposium on Conformal and Probabilistic Prediction and Applications.
Number of pages1
Publication statusPublished - 11 Sept 2020

Cite this