Shigeru Nemoto | Searcharxiv

arXiv2016

A Deep-Learning Approach for Operation of an Automated Realtime Flare Forecast

Automated forecasts serve important role in space weather science, by providing statistical insights to flare-trigger mechanisms, and by enabling tailor-made forecasts and high-frequency forecasts. Only by realtime forecast we can experimentally measure the performance of flare-forecasting methods while confidently avoiding overlearning. We have been operating unmanned flare forecast service since August, 2015 that provides 24-hour-ahead forecast of solar flares, every 12 minutes. We report the method and prediction results of the system.

astro-ph.SR↗

arXiv2015

UFCORIN: A Fully Automated Predictor of Solar Flares in GOES X-Ray Flux

We have developed UFCORIN, a platform for studying and automating space weather prediction. Using our system we have tested 6,160 different combinations of SDO/HMI data as input data, and simulated the prediction of GOES X-ray flux for 2 years (2011-2012) with one-hour cadence. We have found that direct comparison of the true skill statistics (TSS) from small cross-validation sets is ill-posed, and used the standard scores ($z$) of the TSS to compare the performance of the various prediction strategies. The $z$ of a strategy is a stochastic variable of the stochastically-chosen cross-validation dataset, and the $z$ for the three strategies best at predicting X, $\geq$M and $\geq$C class flares are better than the average $z$ of the 6,160 strategies by 2.3$σ$, 2.1$σ$, 3.8$σ$ confidence levels, respectively. The best three TSS values were $0.75\pm0.07$, $0.48\pm0.02$, and $0.56\pm0.04$, respectively.

astro-ph.SR↗