Methods for node attribute prediction tasks utilize graph embedding approaches for graph data processing. However, many of these embedding strategies are developed leveraging specifically graphs’ structural information. However, the graph offers diverse information of various types, and in case of scarce input information, careful incorporation of available data types can help achieve the expected performance. Incorporating available network information types, such as node attributes, inherent properties, and structures, into the graph embedding strategy for node attribute prediction requires understanding their effectiveness. To this end, our study thoroughly analyzes the relationship between network information types and their impact on attribute prediction performance within graphs. We evaluate the performance of the attribute prediction method with diverse information types and explore the benefits of different data combinations. Our results prove the contribution of feature information alongside network topology for node attribute prediction.