Search

This poster presents methodologies on training deep learning architectures to predict occurrences of Solar Flares(M1.0+) for the next 24 hours. We use magnetogram active region patches, automatically detected and provided by Joint Science Operations Center (JSOC), as inputs to our models.

Solar flares not only pose risks to outer space technologies and astronauts' well being, but also cause disruptions on earth to our hight-tech, interconnected infrastructure our lives highly depend on. While a number of machine-learning methods have been proposed to improve flare prediction, none of them, to the best of our knowledge, have investigated the impact of outliers on the reliability and those models' performance. In this study, we investigate the impact of outliers in a multivariate time series benchmark dataset, namely SWAN-SF, on flare prediction models, and test our hypothesis. That is, there exist outliers in SWAN-SF, removal of which enhances the performance of the prediction models on unseen datasets. We employ Isolation Forest to detect the outliers among the weaker flare instances. Several experiments are carried out using a large range of contamination rates which determine the percentage of present outliers. We asses the quality of each dataset in terms of its actual contamination using TimeSeriesSVC. In our best finding, we achieve a 279% increase in True Skill Statistic and 68% increase in Heidke Skill Score. The results show that overall a significant improvement can be achieved to flare prediction if outliers are detected and removed properly.

One of the limiting factors in training data-driven, rare-event prediction algorithms is the scarcity of the events of interest resulting in an extreme imbalance in the data. There have been many methods introduced in the literature for overcoming this issue; simple data manipulation through undersampling and oversampling, utilizing cost-sensitive learning algorithms, or by generating synthetic data points following the distribution of the existing data. While synthetic data generation has recently received a great deal of attention, there are real challenges involved in doing so for high-dimensional data such as multivariate time series. In this study, we explore the usefulness of the conditional generative adversarial network (CGAN) as a means to perform data-informed oversampling in order to balance a large dataset of multivariate time series. We utilize a flare forecasting benchmark dataset, named SWAN-SF, and design two verification methods to both quantitatively and qualitatively evaluate the similarity between the generated minority and the ground-truth samples. We further assess the quality of the generated samples by training a classical, supervised machine learning algorithm on synthetic data, and testing the trained model on the unseen, real data. The results show that the classifier trained on the data augmented with the synthetic multivariate time series achieves a significant improvement compared with the case where no augmentation is used. The popular flare forecasting evaluation metrics, TSS and HSS, report 20-fold and 5-fold improvements, respectively, indicating the remarkable statistical similarities, and the usefulness of CGAN-based data generation for complicated tasks such as flare forecasting., The 20th International Conference on Artificial Intelligence and Soft Computing (ICAISC02021). 13 pages, 4 figures

The authors of this white paper met on 16-17 January 2020 at the New Jersey Institute of Technology, Newark, NJ, for a 2-day workshop that brought together a group of heliophysicists, data providers, expert modelers, and computer/data scientists. Their objective was to discuss critical developments and prospects of the application of machine and/or deep learning techniques for data analysis, modeling and forecasting in Heliophysics, and to shape a strategy for further developments in the field. The workshop combined a set of plenary sessions featuring invited introductory talks interleaved with a set of open discussion sessions. The outcome of the discussion is encapsulated in this white paper that also features a top-level list of recommendations agreed by participants., Workshop Report

We report on progress made by our interdisciplinary Data Mining Lab at Georgia State University on this recently funded (October 1, 2019) project. We present brief overview of our project and focus on the first two phases of our research: (1) Data & Metadata Acquisition, and (2) Generation of Data Sets for Benchmarking.

We report on progress made by our interdisciplinary Data Mining Lab at Georgia State University on this recently funded (October 1, 2019) project. We present brief overview of our project and focus on the first two phases of our research: (1) Data & Metadata Acquisition, and (2) Generation of Data Sets for Benchmarking.

Influence on health behavior from peers is well known and it has been shown that participants in an online physical activity promotion program are generally more successful when they share their achievements through an online community. However, more detailed insights are needed into the mechanisms that explain the influence of a community on physical activity level (PAL). This paper discusses a detailed analysis of a data set of participants in an online physical activity promotion program. The analysis focuses on the comparison two groups of participants, namely participants who will join a community at some point in time and participants who will never join a community. A well-balanced selection is made to eliminate to a large extent factors that dilute the effect of the willingness to partake in a community. We create statistical models that describe the PAL increase at the end of the program. A comparison of these models shows that participants that will participate in a community not only have a higher PAL at the start of the program, but also that the PAL increase is significantly greater compared to participants that will not members. The results further support the hypothesis that the possibility to share achievements is an important feature of successful health promotion programs. At the same time, it raises the question whether part of the success is caused by a selection bias, as people that are willing to participate in a community are already more active at the start.

Solar energetic particles are a result of intense solar events such as solar flares and Coronal Mass Ejections (CMEs). These latter events all together can cause major disruptions to spacecraft that are in Earth's orbit and outside of the magnetosphere. In this work we are interested in establishing the necessary conditions for a major geo-effective solar particle storm immediately after a major flare, namely the existence of a direct magnetic connection. To our knowledge, this is the first work that explores not only the correlations of GOES X-ray and proton channels, but also the correlations that happen across all the proton channels. We found that proton channels auto-correlations and cross-correlations may also be precursors to the occurrence of an SEP event. In this paper, we tackle the problem of predicting >100 MeV SEP events from a multivariate time series perspective using easily interpretable decision tree models.

Solar energetic particles are a result of intense solar events such as solar flares and Coronal Mass Ejections (CMEs). These latter events all together can cause major disruptions to spacecraft that are in Earth's orbit and outside of the magnetosphere. In this work we are interested in establishing the necessary conditions for a major geo-effective solar particle storm immediately after a major flare, namely the existence of a direct magnetic connection. To our knowledge, this is the first work that explores not only the correlations of GOES X-ray and proton channels, but also the correlations that happen across all the proton channels. We found that proton channels auto-correlations and cross-correlations may also be precursors to the occurrence of an SEP event. In this paper, we tackle the problem of predicting >100 MeV SEP events from a multivariate time series perspective using easily interpretable decision tree models.

In this work we are analyzing scalability of the heuristic algorithm we used in the past to discover knowledge from multi-valued symbolic attributes in fuzzy databases. The non-atomic descriptors, characterizing a single attribute of a database record, are commonly used in fuzzy databases to reflect uncertainty about the recorded observation. In this paper, we present implementation details and scalability tests of the algorithm, which we developed to precisely interpret such non-atomic values and to transfer (i.e. defuzzify) the fuzzy tuples to the forms acceptable for many regular (i.e. atomic values based) data mining algorithms. Important advantages of our approach are: (1) its linear scalability, and (2) its unique capability of incorporating background knowledge, implicitly stored in the fuzzy database models in the form of fuzzy similarity hierarchy, into the interpretation/defuzzification process.

Solar filaments are cool, dark channels of partially-ionized plasma that lie above the chromosphere. Their structure follows the neutral line between local regions of opposite magnetic polarity. Previous research (e.g. Schmieder et al. 2013) has shown a positive correlation (80%) between the occurrence of filament eruptions and coronal mass ejections (CME’s). If certain filament properties, such as length, chirality, and tilt, indicate a tendency towards filament eruptions, one may be able to further predict an oncoming CME. Towards this end, we present a novel algorithm based on spatiotemporal analysis that systematically correlates filament eruptions documented in the Heliophysics Event Knowledgebase (HEK) with HEK filaments that have been grouped together using a tracking algorithm developed at Georgia State University (e.g. Kempton et al. 2014). We also find filament tracks that are not correlated with eruptions to form a null data set in a similar fashion. Finally, we compare the metadata from erupting and non-erupting filament tracks to discover which filament properties may present signs of an eruption onset. Through statistical methods such as the two-sample Kolmogorov-Smirnov test and Random Forest Classifier, we find that a filament that is increasing in length or changing in tilt with respect to the equator may be a useful gauge to predict a filament eruption. However, the average values of length and tilt for both datasets follow similar distributions, leading us to conclude that these parameters do not indicate an eruption event.