https://doi.org/10.1140/epjc/s10052-026-15968-7
Regular Article - Computing, Software and Data Science
A statistical investigation of radio-loudness in SDSS DR16Q quasars using ML-based classification and physical feature analysis
Indian Institute of Technology Hyderabad, Sangareddy, 502284, Telangana, India
a
This email address is being protected from spambots. You need JavaScript enabled to view it.
Received:
3
December
2025
Accepted:
10
June
2026
Published online:
5
July
2026
Abstract
The problem of quasar classification comes in the class of highly imbalanced classification problems since Radio-loud quasars are rare and make up very small portion of total quasar population. In this work,We perform classification of Radio-Loud quasars (RLQs) and Radio-Quiet quasars (RQQs) on a sample of 84,326 QSOs (19,058 RL + 65,268 RQ) with 
from the SDSS-DR16 dataset with FIRST geometric cut, using Random Forest (RF), PCA based regression classifier (PRC), and XGBOOST (XGB) classifiers with balanced weights. Using original feature set consisting of PSF u,g,r,i,z, XMM & ROSAT x-ray, derived colors, UKIDSS, GALEX and WISE features,we consider first ten PC loadings accounting for 
explained data variance. We present systematic analysis for top performing PC planes for deg.1 to 3 decision boundaries with regression classifiers. We assess the success of each model based upon correctly classified RLQs and physical interpretability. PRC achieves a recall comparable to XGB and RF. Top performing plane (
) for each PRC degree correctly classifies on an average, 3900/5717 RLQs. Balanced XGB (
) and RF (
) correctly classify 4261/5717 and 4181/5717 RLQs respectively. Kernel density maps and Markov Chain Monte Carlo (MCMC) posterior distribution on PC planes suggest partial separability between RLQs and RQQs in PC planes with loadings consisting of PSF and Gaia magnitudes, 
, optical colors, WISE and X ray flux. Some PC components exhibit dominance in overall classification workflow with noticeable linear components. We introduce RLQ Separation Rating (RLQSR) metric to quantify geometric separability in PC planes and prescribe a routine for model optimization effects too. Regression classifier analysis on PCA reduced data holds potential for effectively tackling the classification problem with a physically interpretable and transparent pipeline achieving a good recall on minority class. We conclude that the separability in PCA planes is only partial and is dominated by certain physical features. RLQs and RQQs significantly overlap in high density zones of PC planes and the separability arises only due to extended (lobe like)/sparse distributions of RLQ population in PC planes. We find that pairing optical passband with IR and X-ray passbands show better separability.
R. Joshi and V. Shinde contributed equally.
© The Author(s) 2026
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Funded by SCOAP3.
