A Kernel Score Perspective on Forecast Disagreement and the Linear Pool
The variance of a linearly combined forecast distribution (or linear pool) consists of two components: The average variance of the component distributions (average uncertainty'), and the average squared difference between the components' means and the pool's mean (disagreement'). This paper shows that similar decompositions hold for a class of uncertainty measures that can be constructed as entropy functions of kernel scores. The latter are a rich family of scoring rules that covers point and distribution forecasts for univariate and multivariate, discrete and continuous settings. We further show that the disagreement term is useful for understanding the ex-post performance of the linear pool (as compared to the component distributions), and motivates using the linear pool instead of other forecast combination techniques. From a practical perspective, the results in this paper suggest principled measures of forecast disagreement in a wide range of applied settings.
Code (0)
등록된 구현이 없습니다.
Similar Papers 제목 키워드 기반
Multi-view Kernel PCA for Time series Forecasting
In this paper, we propose a kernel principal component analysis model for multi-variate time series forecasting, where the training and prediction schemes are derived from the multi-view formulation of Restricted Kernel …
Time SeriesTime Series AnalysisTime Series ForecastingBeyond Top-Class Agreement: Using Divergences to Forecast Performance under Distribution Shift
Knowing if a model will generalize to data 'in the wild' is crucial for safe deployment. To this end, we study model disagreement notions that consider the full predictive distribution - specifically disagreement based o…
Generalized Gibbs Ensemble Weighting for Forecast Combination
Forecast combination is a reliable way to improve predictive performance when several forecasting models are available. Simple aggregation rules such as the mean, median, trimmed mean, inverse-loss weighting, and exponen…
The Measuring Hate Speech Corpus: Leveraging Rasch Measurement Theory for Data Perspectivism
We introduce the Measuring Hate Speech corpus, a dataset created to measure hate speech while adjusting for annotators’ perspectives. It consists of 50,070 social media comments spanning YouTube, Reddit, and Twitter, lab…
Experimental DesignCrop yield probability density forecasting via quantile random forest and Epanechnikov Kernel function
A reliable and accurate forecasting model for crop yields is of crucial importance for efficient decision-making process in the agricultural sector. However, due to weather extremes and uncertainties, most forecasting mo…
Decision MakingDensity EstimationFeature ImportancePrediction+3