چكيده به لاتين
The nullity of a graph is the existence of zeros as eigenvalues in its spectrum. Nullity calculations are widely used to determine the stability of a chemical molecule. To calculate the nullity, formulas have been presented for specific graphs, but memorizing the formula of each specific graph is challenging and there is no formula for all types of graphs. But by using the rank of the graph, one can calculate the nullity. However, processing time increases with the growth of the number of vertices of graphs. Most graph properties, including girth, clique number and independence set number, have the same problem. Therefore, in this thesis, by examining machine learning methods for calculating graph property values, including nullity, girth, clique number and independence set number, a solution is proposed. In this regard, random graph generation methods are used to collect training samples. The accuracy of the experimental results obtained on the collected data sets is 91.33, 90.51, 91.61 and 99.98 percentage for binary classification of nullity, clique number, girth and independence set number, respectively. Also, the correlation coefficient of the best experimental results obtained on the collected data sets is 0.95, 0.9703, 0.9393 and 0.992 for predicting the value of nullity, clique number, girth and independence set number, respectively.