Solar flare is a type of solar activity that occurs at active regions at the surface of the sun. The emission of solar flares has numerous consequences, including the dis- turbances of magnetic fields, disruptions from energetic particles, and geomagnetic explosions. All those consequences have numerous impacts on human civilization, including the degradation of communication systems, power grids, space navigation, and even natural disasters. Thus, those minor or catastrophic consequences are al- ways threatening to the normal operation of society and decision-makers of those systems always seek a precise and accurate prediction of hazardous solar flares. This paper aims to develop a forecast model that can accurately decide whether solar flares would happen in the future. The data is extracted from the NOAA (National Oceanic and Atmospheric Administration) GOES-16 X-Ray Sensor that monitors solar activity by measuring the flux intensity of X-Ray. The original data is in the form of time series. Markov Transition Field is applied to the time series data, transforming the data into the form of 3-dimensional images. Therefore, the data undergone pre-processing could be applied to computer vision models. The aim of these models is to accurately recognize the Markov Transition Field images that symbolize there would be solar flare emission one hour later through a binary classification. Deep learning architects are the major components to accomplish this forecast task. Convolutional Neural Network (CNN) is a common approach in doing clas- sification tasks, which is also frequently used in recent studies that aim to predict flare emission through X-Ray images of the active regions. There are several classic CNN undergone training and testing, including LeNet-5, AlexNet, VGGNet 16 and 19, and ResNet-18, that utilizes the residue block structure. These CNN architects provide fascinating reliability and accuracy in this prediction task of solar flares, with multiple structures providing accuracy greater than 80%. Furthermore, Vision Transformer, a deep learning architect also used in classification based on trans- former structure is applied to the flare task. It is comprised of the core structure of multiple-head self-attention, residue blocks, layer normalization, and multilayer perceptron. Vision Transformer has shown outstanding accuracy (89.89%) while making predictions of solar flare emissions.

Mingjie Wei. Prediction Based on Convolutional Neural Networks and Vision Transformer for GOES-XRS Solar Flare Time Series. Academic Journal of Computing & Information Science (2023), Vol. 6, Issue 8: 158-164. https://doi.org/10.25236/AJCIS.2023.060819.

[1] A. Stanislavsky, M. M. K. B. J. J., W. Nitka. (2020). Prediction performance of hidden markov modelling for solar flares.

[2] Bo Wang, J. D., & Jiang, H. (2022). Markov transition field combined with convolu- tional neural network improved the predictive performance of near-infrared spectroscopy models for determination of aflatoxin b1 in maize.

[3] J. P. Mason, J. T. H. (2010). Testing automated solar flare forecasting with 13 years of michelson doppler imager magnetograms.

[4] N Nishizuka, Y. K. M. D. M. I., K Sugiura. (2018). Deep flare net (defn) model for solar flare prediction.

[5] Patricia L. Bornmann, J. H. L. M. R. G. H. A. G. R. V., David Speich. (1996). Goes x-ray sensor and its use in predicting solar-terrestrial disturbances.

[6] Wheatland, M. S. (2000). The origin of the solar flare waiting-time distribution. Wheatland, M. S. (2004). A bayesian approach to solar flare prediction.

[7] X Huang, L. X. J. L. R. L. X. D., H Wang. (2018). Deep learning based solar flare forecasting model. i. results for line-of-sight magnetograms.

[8] X Li, X. W. L. W., Y Zheng. (2020). Predicting solar flares using a novel deep convo- lutional neural network.

[9] Yanfang Zheng, X. W., Xuebao Li, & Zhou, T. (2019). Modified convolutional neural network with transfer learning for solar flare prediction.

[10] Yang Chen, A. O. H. G. T. B. D. T. Z. X. W. H. Z. Z. S. T. I. G., Ward B. Manchester. (2000). Identifying solar flare precursors using time series of sdo/hmi images and sharp parameters.

[11] ZhiguangWang, T. O. (2015). Imaging time-series to improve classification and impu- tation.

[12] Z Sun, X. W. Y. W. H. S. T. G. Y. C. A. H., MG Bobra. (2022). Predicting solar flares using cnn and lstm on two solar cycles of active region data.