Application Of PRESTO-based Multiple Single-pulse Candidate Identification Methods To CRAFTS Data

Single-pulse searches are a powerful tool for pulsar searches and play an important role in detecting rotating radio transient sources as well as fast radio bursts.However,the biggest challenge in conducting single-pulse search experiments is the generation of huge amounts of spurious candidates caused by RFI or noise.Especially for observations of the size of FAST,hundreds of millions of candidates can be generated.This requires single-pulse candidate identification methods to be fast and effective enough,otherwise it will delay the discovery of science targets.Based on this,we evaluate the performance of four single-pulse candidate identification methods applied to CRAFTS ultra-wideband data to explore how to perform efficient single-pulse candidate identification.First,a benchmark dataset is constructed.Then,we evaluate the performance of four single-pulse candidate identification methods applied to the benchmark dataset.Finally,the performance data of the screening strategy or machine learning classifier with the highest F1-score among the four recognition methods are compared and analyzed,and the results show that SPEGID achieves the best performance（high recall rate and low false positive rate）.The analysis of misclassified candidates shows that the interference of strongly mimicking pulsars（RFI1,RFI2,RFI3）causes multiple recognition methods to generate a large number of false candidates.In addition,The speed of screening single pulsar candidates by the four identification methods is also compared and analyzed,and the results show that SPS has the fastest screening speed.Based on the analysis results,we propose a feasible strategy to carry out efficient single-pulse candidate identification in the future,and point out the direction for the future development of single-pulse candidate identification methods based on FAST data.To address the problem of high false positive rate of RRATtrap pulsar candidates in CRAFTS ultra-wideband data,we propose the method of using three features to distinguish pulsar signals from RFI,and by limiting the threshold range of the three features,the problem of high false positive rate of single pulsar candidates is reduced to a certain extent.