Intervening or associated? Machine learning classification of redshifted H <scp>i</scp> 21-cm absorption

ABSTRACT In a previous paper, we presented the results of applying machine learning to classify whether an H i 21-cm absorption spectrum arises in source intervening sightline more distant radio or within host itself. This is usually determined from optical giving redshift. However, not only will this be impractical for large number sources expected detected with Square Kilometre Array, but bright are most ultraviolet luminous at high redshift and so bias against detection cool, neutral gas. Adding another 44, mostly newly absorbers, sample 92, test four different algorithms, again using line properties (width, depth, Gaussian fits) as features. Of these three gave some improvement over sample, logistic regression model best results. suggests that inclusion further training data, new absorbers detected, increase prediction accuracy above current ?80 per cent. We use zabs = 0.42 towards PKS 1657?298 find associated, which consistent study zem ? K-band magnitude–redshift relation.