CLARE: Classification-based Regression for Electron Temperature Prediction

2026-03-12

AI Summary

This study addresses the longstanding challenge of accurately predicting electron temperature (Te)—a critical parameter in upper atmospheric space weather—by introducing an innovative classification-based regression approach. Leveraging AKEBONO satellite observations and solar-geomagnetic indices, the method discretizes the continuous Te prediction task into 150 distinct intervals, thereby enabling both high predictive accuracy and robust uncertainty quantification. This strategy overcomes key limitations of conventional regression models. On the test set, the proposed model achieves a prediction accuracy of 69.67% within ±10% of the true Te values, with a storm-time accuracy of 46.17% during geomagnetic disturbances—representing a 6.46% improvement over traditional regression approaches.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Relational Probabilistic Models
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Electron temperature (Te) is an important parameter governing space weather in the upper atmosphere, but has historically been underexplored in the space weather machine learning literature. We present CLARE, a machine learning model for predicting electron temperature in the Earth's plasmasphere trained on AKEBONO (EXOS-D) satellite measurements as well as solar and geomagnetic indices. CLARE uses a classification-based regression architecture that transforms the continuous Te output space into 150 discrete classification intervals. Training the model on a classification task improves prediction accuracy by 6.46% relative compared to a traditional regression model while also outputting uncertainty estimation information on its predictions. On a held out test set from the AKEBONO data, the model's Te predictions achieve 69.67% accuracy within 10% of the ground truth and 46.17% on a known geomagnetic storm period from January 30th to February 7th, 1991. We show that machine learning can be used to produce high-accuracy Te models on publicly available data.

Problem

• electron temperature
• space weather
• plasmasphere
• machine learning
• prediction

Innovation

• classification-based regression
• electron temperature prediction
• uncertainty estimation
• space weather modeling
• machine learning