The growing presence of microplastics in drinking water poses severe dangers to health and the environment, requiring enhanced detection methods. This work deals with the constraints of conventional detection methods, such as visual inspection and Raman spectroscopy, that are labour-intensive and unscalable. By contrasting CNN-SVM and CNN-RF, two highly efficient hybrid models of machine learning, the essential purpose is to improve the accuracy of detection of the microplastics. Convolutional neural networks (CNNs) extract features from images of water samples, and the method classifies them by using Support Vector Machine (SVM) and Random Forest (RF) algorithms. The study assesses the models’ precision, recall, f1-score, and overall accuracy. The findings show that these hybrid models greatly enhance detection abilities, resulting in a more effective and flexible solution. Practical uses involve real-time monitoring in water treatment plants, ecological evaluations of water bodies, as well as household water filtering systems, which provide vital information for compliance with regulations and public health safety. This study provides our knowledge on the problem of microplastic pollution and indicates possible future uses in monitoring the environment and policy-making, which will help attempts to reduce the harmful effects of microplastics in drinking water.