Electromagnetic sensors have seen substantial growth in various practical applications, particularly in integration with machine learning techniques. This combination facilitates an opportunity for analysis and classification of materials, especially in liquid-based applications. This project proposed a sensor design based on a coaxial probe. The developed sensor is highly sensitive and non-invasive. A model of the sensor was constructed in CST Studio Suite and a substantial amount of data was simulated to build a database for training three machine algorithms, i.e., the k-nearest neighbors, the random forest, and the support vector machine. The results showed that descent classification accuracies were obtained by using the designed sensor.