The detection of powerful 22 GHz emission created by Microwave Amplified Stimulated Emission Radiation (maser) from water molecules proved to be crucial for obtaining direct measurements of distances to galaxies, as well as for the most accurate measurements of masses of supermassive black holes residing at the center of these systems. The water maser systems are, however, extremely rare; only approximately 3-4% of all galaxies surveyed to date reveal their presence. We present here a machine learning approach to improving the efficiency in detecting water megamaser disks based on empirical features discovered to correlate best with the water masing phenomenon and predict the highest maser detection rates. This work explores the use of the K-Nearest Neighbor model to predict the probability of classifying galaxies as maser hosts for given observed galaxy traits, along with the development of a web tool that produces such classifications for any input feature set. The results of this study are essential for the greater efficiency of telescope time usage for the future maser surveys.