Deep Learning-Based Classification and Analysis of Pulsar Candidates in Fermi-LAT Unassociated Sources

This repository contains the machine learning classification results for unassociated sources in the Fermi-LAT catalog. These classifications are based on the methodology described in the associated research paper. Dataset Files The repository includes two primary CSV files: 1. fermi_unassociated_predictions_all.csv This file contains all the unassociated sources analyzed from the Fermi-LAT catalog. It includes all standard catalog columns alongside our model's predictions, with no probability thresholds applied. Model Output Columns: prob_AGN: Predicted probability of the source being an Active Galactic Nucleus (AGN). prob_PSR_MSP: Predicted probability of the source being a Pulsar (PSR) or Millisecond Pulsar (MSP). prob_MSP: Predicted probability of the source being a Millisecond Pulsar (MSP) (NaN if not classified as a pulsar). prob_YP: Predicted probability of the source being a Young Pulsar (YP / PSR) (NaN if not classified as a pulsar). predicted_class: The final label assigned (AGN, MSP, or PSR). confidence: The confidence score associated with the prediction. 2. fermi_unassociated_predictions_optimized_cuts.csv This file contains a high-confidence subset of the data. It is filtered based on optimized probability thresholds: PSR threshold: $\ge 0.98$ AGN threshold: $\ge 0.91$ The column structure is identical to the file above, containing only the sources that met these selection criteria. Data Usage The data is provided in CSV format and is compatible with standard data analysis tools: Python (Pandas): df = pd.read_csv('filename.csv') R: data <- read.csv('filename.csv') Spreadsheet Software: Excel, Google Sheets, LibreOffice Calc. Astrophysics Tools: Compatible with TOPCAT by loading the file directly. For details on the standard Fermi-LAT catalog parameters (e.g., RAJ2000, DEJ2000, GLON, GLAT, Signif_Avg, Flux1000), please refer to the official Fermi-LAT catalog documentation. Citation If you use this dataset in your research, please cite the following paper: