Comparative analysis of missing data imputation methods for CSST survey: Impact on photometric redshift estimation performance

¹ Shanghai Key Lab for Astrophysics, Shanghai Normal University, Shanghai 200234, China
² Center for Astronomy and Space Sciences, China Three Gorges University, Yichang 443000, PR China
³ South-Western Institute for Astronomy Research, Yunnan University, Kunming 650500, China
⁴ Department of Astronomy, School of Physics,Peking University, Beijing 100871, China
⁵ Changchun Institute of Optics, Fine Mechanics and Physics, Chinese Academy of Sciences, Changchun 130033, China
⁶ University Observatory, Faculty of Physics, Ludwig-Maximilians-Universität, Scheinerstr. 1, 81679 Munich, Germany
⁷ National Astronomical Observatories, Chinese Academy of Sciences, 20A Datun Road, Chaoyang District, Beijing 100101, PR China
⁸ Purple Mountain Observatory, Chinese Academy of Sciences, Nanjing 210023, China
⁹ School of Physics and Astronomy, Sun Yat-sen University, Zhuhai 519082, PR China
¹⁰ Shanghai Astronomical Observatory, Chinese Academy of Sciences, 80 Nandan Road, Shanghai 200030, PR China
¹¹ CSST Science Center for the Guangdong-Hong Kong-Macau Greater Bay Area, Zhuhai 519082, PR China
¹² School of Physics and Electronics, Hunan Normal University, 36 Lushan Road, Changsha 410081, China
¹³ Computer Network Information Center, Chinese Academy of Sciences, 2 East Kexueyuan South Road, Haidian District, Beijing 100083, PR China
¹⁴ Center for Astrophysics and Great Bay Center of National Astronomical Data Center, Guangzhou University, Guangzhou, Guangdong 510006, PR China
¹⁵ Faculty of Information Engineering and Automation, Kunming University of Science and Technology, No.727 Jingming South Road, Kunming 650500, PR China
¹⁶ School of Astronomy and Space Science, University of Chinese Academy of Sciences, Beijing 101408, PR China
¹⁷ Key Laboratory of Space Astronomy and Technology, National Astronomical Observatories, Chinese Academy of Sciences, Beijing 100101, China

^★ Corresponding author: This email address is being protected from spambots. You need JavaScript enabled to view it.

Received: 25 January 2026
Accepted: 10 April 2026

Abstract

Improving the accuracy of photometric redshifts (photo-z) is essential for reliable statistical studies of cosmology and galaxy evolution. However, missing photometric bands are a common observational challenge that can significantly degrade photo-z estimation accuracy. In this work, we present a systematic evaluation of data imputation methods aimed at improving photo-z performance. We benchmark a range of representative machine learning and deep learning architectures, identifying k-nearest neighbors (KNN) and the attentionbased SAITS model as the leading performers. These models are then applied to China Space Station Survey Telescope mock data to assess their performance under realistic observational conditions. Our results show that KNN yields the highest accuracy under idealized missing completely at random (MCAR) conditions with complete training sets, whereas robustness tests reveal that SAITS significantly outperforms KNN when training data are incomplete or when applied to realistic mixed-mechanism scenarios. We find that domain consistency between training and testing missingness patterns is a prerequisite for optimal performance, highlighting the risks of domain shift in supervised regression tasks. Furthermore, our analysis demonstrates that while general imputation models are highly effective for MCAR and missing at random data, they are detrimental when applied to missing not at random data arising from flux limits, as statistical models fail to capture the physical information inherent in these nondetection. Consequently, we advocate for more sophisticated architectures capable of disentangling stochastic missingness from physical nondetection to address these distinct mechanisms individually.