Machine learning predicts graph properties: Clique, girth, and independent numbers

Graph properties’ computation is widely used in science. Some algorithms for specific graphs help us compute properties like girth, clique number, and independent number. Nevertheless, processing time would be increased by the growth of the number of graph vertices. This paper suggests machine learning methods for computing a given graph’s girth, maximum clique number, and maximum independent set number. We leveraged random graph generation methods to collect many graph instances. Next, we present 14 features used for the training and testing while classifying or predicting those properties. The experimental results on the collected dataset offer accuracies of 90.51%, 91.61%, and 99.99% for binary classification and correlation coefficients of 0.9703, 0.8757, and 0.992 for value prediction for maximum clique number, girth, and maximum independent set number, respectively.