Quinolone antibiotics (QNs) are extensively utilized in treating diseases and animal husbandry owing to their potent antimicrobial properties. However, the excessive utilization of QNs and their release into environmental water sources via wastewater results in their accumulation in natural aquatic ecosystems. The consequence of this phenomenon is the extensive propagation of bacteria that are resistant to antibiotics, as well as the proliferation of resistance genes in aquatic ecosystems. This poses a significant risk to both environmental ecology and human well-being. Traditional techniques employed in detecting QNs exhibit notable sensitivity and accuracy. However, these methods are characterized by their time-consuming nature, reliance on costly equipment, and the inherent difficulty of conducting on-site assessments. Fluorescence analysis technology necessitates a brief detection time, particularly in the case of three-dimensional fluorescence spectroscopy, which enables the acquisition of a substantial amount of target information within a limited timeframe. By integrating data statistics and machine learning models, mathematical methods can efficiently identify multiple QNs. In this study, the fluorescence spectroscopic information of QNs was extensively employed, and the support vector machine regression (SVMR) algorithm was utilized to develop prediction models for QNs, specifically ofloxacin (OFL) and norfloxacin (NOR). The fluorescence spectroscopic data of the unknown samples was subsequently fed into the developed models to obtain the determination outcomes efficiently. While constructing the model, a comparison was made between two supervised learning methods, i.e., PLS-DA and SVMR. It has been determined that SVMR exhibits a strong predictive capability. By manipulating parameters and kernel functions, it was possible to achieve a good linear range for determining OFL and NOR, spanning from 2 to 600 μg·L^-1. The resulting linear correlation coefficients exceeded 0.992 0, and the detection limits were 0.064～0.080 μg·L^-1. The validated method was proven applicable to real water samples, i.e., the recoveries were 98.62%～104.01% for seawater and 103.90%～105.89% for reservoir water. This method offers the advantage of rapid detection speed, allowing for the completion of quantitative analysis of an unknown sample within a mere 3 min. This rapid screening process identifies potential risk factors associated with QNs in the environment. This study employs a novel approach by integrating SVMR with fluorescence spectroscopy to develop a rapid detection method for QNs in real water samples. The proposed method offers a new and scientifically reliable solution for rapidly detecting QNs in environmental water.