Flare Prediction Modeling based on the Time Series of SHARP Parameters along the Polarity Inversion Line of Active Regions

The polarity inversion line (PIL) in active regions (AR) is considered being closely associated with solar flare eruptions. In this study, we rigorously constructed standardized datasets based on time series of different lengths using the SHARP parameters along the PIL. We compared the actual performance of the traditional logistic regression model (non-sequential) and time-series models in solar flare prediction tasks, as well as the predictive performance differences between time series models of different lengths within the CNN-BiLSTM-AT framework. The following conclusions are drawn: 1. It is verified that the prediction performance of the new SHARP parameters determined along the PIL is better. 2. In the actual prediction task, the time-series model is better than the non-sequential model, and the F1 score is almost doubled, reaching 0.59. 3. The most suitable hyperparameters of the model are estimated and the importance of the input parameters is evaluated based on the experimental results. This study provides further references and suggestions for data/model selection for flare prediction.