Predicting Pulsars from Imbalanced Dataset with Hybrid Resampling Approach

Table 1

Summary of related work.

Ref	Name	Year	Dataset	Accuracy
[6]	Lyon et al.	2016	HTRU-1 and LOTAAS	Recall = 92.8% and FPS = 50%
[7]	Eatough et al.	2010	650 survey beams of the PMPS	92% of total pulsars are present in dataset of 2.5 million observations
[11]	Wagstaff et al.	2016	ATNF Pulsar Catalogue (not sure)	98.6% accuracy on historical data and a 99%-100% accuracy on (labeled) newly collected data
[12]	Devine et al.	2016	Prestos single_pulse_search on data from the GBT drift-scan (Green Bank Telescope)
[13]	Bethapudi and Desai	2018	HTRU-S survey data	99.1%
[15]	Wang et al.	2019	PALFA, HTRU, GBNCC, and FAST datasets	96%.
[16]	Bates et al.	2012	HTRU mid-latitude survey	85%
[17]	Li et al.	2018	PMPS-26 k and HTRU dataset	PMPS = 87.50%; HTRU = 97.74%
[18]	Yao et al.	2016	HTRU, HTRU-1, and LOTAAS	Recall: HTRU = 97.49%, HTRU-1 = 84.52%, LOTAAS = 100%
[19]	Mohammed	2018	HTRU-1 dataset	97.8%
[20]	Mohammed and Guo	2020	HTRU, MNIST, and CIFAR-10 datasets	MINIST = 97.50%, HTRU = 100%, CIFAR-10 = 100%.
[21]	Chen et al.	2020	Dataset of pulsar candidate samples collected during high time resolution cosmometry	99%
[14]	Rustum et al.	2020	HTRU2	98.3%