The pollution of water environments is a worldwide concern. Not only marine pollution by plastic litter, including microplastics, but also the spillage of water-soluble synthetic polymers in wastewater have recently gained increasing attention due to their potential risks to soil and water environments. However, conventional methods to identify polymers dissolved in water are laborious and time-consuming. Here, we propose a simple approach to identify synthetic polymers dissolved in water using a peptide-based molecular sensor with a fluorophore unit. Supervised machine learning of multiple fluorescence signals from the sensor, which specifically or nonspecifically interacted with the polymers, was applied for polymer classification as a proof of principle demonstration. Aqueous solutions containing different polymers or multiple polymer species with different mixture ratios were identified successfully. We found that fluorophore-introduced biomolecular sensors have great potential to provide discriminative information regarding water-soluble polymers. Our approach based on the discrimination of multiple optical signals of water-soluble polymers from peptide-based molecular sensors through machine learning will be applicable to next-generation sensing systems for polymers in wastewater or natural environments.