Comparing feature sets and machine-learning models for prediction of solar flares Topology, physics, and model complexity
Comparing feature sets and machine-learning models for prediction of solar flares Topology, physics, and model complexity
复制标题
DOI：
10.1051/0004-6361/202245742
复制
发表时间：
2023-06-19
影响因子：
6.5
通讯作者：
Meiss, J. D.
中科院分区：
文献类型：
--
作者：
Deshmukh, V.;Baskar, S.;Meiss, J. D.
上下文。预测太阳耀斑的机器学习方法通常采用基于物理的特征，这些特征是由专家精心挑选的，目的是捕捉太阳光球磁场的显著特征。尽管这些模型的复杂性和复杂性随着时间的推移而增长，但在特征集的选择上却几乎没有什么进展，也没有任何关于额外的模型复杂性是否会导致更高的预测技能的系统研究。该研究比较了四种不同的基于机器学习的耀斑预测模型的相对预测性能，并增加了复杂性。它评估了三个不同的特征集作为每个模型的输入：一个“传统的”基于物理的特征集，一个来自太阳磁场拓扑数据分析（TDA）的新颖的“基于形状的”特征集，以及这两个集的组合。为了保证模型在不同特征集之间的公平比较，采用了系统的超参数调整框架。最后，利用主成分分析方法研究了降维对特征集的影响。结果表明，在典型的24小时耀斑预报问题上，自由参数较少的简单模型比复杂模型的预报效果更好。换句话说，更复杂的机器学习架构并不一定保证更好的预测性能。此外，我们发现基于形状的特征集包含的有用信息与基于物理的特征集一样多，并且这些特征集的维度-特别是基于形状的特征集-可以在不影响预测精度的情况下大大降低。
Context. Machine-learning methods for predicting solar flares typically employ physics-based features that have been carefully chosen by experts in order to capture the salient features of the photospheric magnetic fields of the Sun.Aims. Though the sophistication and complexity of these models have grown over time, there has been little evolution in the choice of feature sets, or any systematic study of whether the additional model complexity leads to higher predictive skill.Methods. This study compares the relative prediction performance of four different machine-learning based flare prediction models with increasing degrees of complexity. It evaluates three different feature sets as input to each model: a "traditional" physics-based feature set, a novel "shape-based" feature set derived from topological data analysis (TDA) of the solar magnetic field, and a combination of these two sets. A systematic hyperparameter tuning framework is employed in order to assure fair comparisons of the models across different feature sets. Finally, principal component analysis is used to study the effects of dimensionality reduction on these feature sets.Results. It is shown that simpler models with fewer free parameters perform better than the more complicated models on the canonical 24-h flare forecasting problem. In other words, more complex machine-learning architectures do not necessarily guarantee better prediction performance. In addition, it is found that shape-based feature sets contain just as much useful information as physics-based feature sets for the purpose of flare prediction, and that the dimension of these feature sets - particularly the shape-based one - can be greatly reduced without impacting predictive accuracy.