The study of carbon-enhanced metal-poor (CEMP) stars is of great significance for understanding the chemical evolution of the early universe and stellar formation. CEMP stars are characterized by carbon overabundance and are classified into several subclasses based on the abundance patterns of neutron-capture elements,including CEMP-s,CEMP-no,CEMP-r,and CEMP-r/s. These subclasses provide important insights into the formation of the first stars,early stellar nucleosynthesis,and supernova explosions. However, one of the major challenges in CEMP star research is the relatively small sample size of identified stars, which limits statistical analyses and hinders a comprehensive understanding of their properties. Fortunately, a series of large-scale spectroscopic survey projects have been launched and developed in recent years, providing unprecedented opportunities and technical challenges for the search and study of CEMP stars. To this end, this paper draws on the progress and future prospects of existing methods in constructing large CEMP data sets and offers an in-depth discussion from a technical standpoint, focusing on the strengths and limitations. In addition,we review recent advancements in the identification of CEMP stars,emphasizing the growing role of machine learning in processing and analyzing the increasingly large data sets generated by modern astronomical surveys. Compared to traditional spectral analysis methods,machine learning offers greater efficiency in handling complex data,automatic extraction of stellar parameters,and improved prediction accuracy. Despite these advancements,the research faces persistent challenges,including the scarcity of labeled samples,limitations imposed by low-resolution spectra,and the lack of interpretability in machine learning models. To address these issues,the paper proposes potential solutions and future research directions aimed at advancing the study of CEMP stars and enhancing our understanding of their role in the chemical evolution of the universe.