PEACE: Pulsar Evaluation Algorithm for Candidate Extraction -- A software package for post-analysis processing of pulsar survey candidates
Modern radio pulsar surveys produce a large volume of prospective candidates, the majority of which are polluted by human-created radio frequency interference or other forms of noise. Typically, large numbers of candidates need to be visually inspected in order to determine if they are real pulsars. This process can be labor intensive. In this paper, we introduce an algorithm called PEACE (Pulsar Evaluation Algorithm for Candidate Extraction) which improves the efficiency of identifying pulsar signals. The algorithm ranks the candidates based on a score function. Unlike popular machine-learning based algorithms, no prior training data sets are required. This algorithm has been applied to data from several large-scale radio pulsar surveys. Using the human-based ranking results generated by students in the Arecibo Remote Command enter programme, the statistical performance of PEACE was evaluated. It was found that PEACE ranked 68% of the student-identified pulsars within the top 0.17% of sorted candidates, 95% within the top 0.34%, and 100% within the top 3.7%. This clearly demonstrates that PEACE significantly increases the pulsar identification rate by a factor of about 50 to 1000. To date, PEACE has been directly responsible for the discovery of 47 new pulsars, 5 of which are millisecond pulsars that may be useful for pulsar timing based gravitational-wave detection projects.