Monopulse signal recognition and application based on XGBoost

Pulsar search is an important research direction in the field of radio astronomy. With the continuous construction and development of large radio telescopes, the data collected is increasing exponentially. How to accurately identify pulsar signals from the massive data acquired quickly has become a great challenge. Taking the observation data of LOFAR Joint Array Sky Survey Project as an example, this paper designs 10 feature variables for monopulse signal recognition, and further studies the application of XGBoost combined with wrapper feature selection method in monopulse signal recognition. The experimental results show that, XGBoost combined with wrapper feature selection method has more comprehensive advantages in monopulse signal recognition, with the lowest classification error rate and the highest accuracy rate, recall rate and F1-score value of classification results, which are on average 1 to 2 percentage points higher than other models. The characteristic variables and recognition methods designed in this study can provide methods and technical references for pulsar search based on FAST detection signals in China.