The rapid increase in Android malware presents a significant challenge to the security of mobile devices. To address this issue, this paper proposes a classification model that utilizes multi-view feature fusion to improve malware detection and classification performance. First, A two-layer graph convolutional network is employed to learn structured representations from the Android application’s call graph, effectively capturing its inherent topological characteristics. Secondly, A multi-scale convolutional neural network is leveraged to derive social network representations from the call graph, enhancing both local and global information expression. Subsequently, the two-view features are integrated via a cross-attention strategy, which facilitates more effective modeling of feature correlations and enhances classification accuracy. Experimental results on public datasets show that the proposed model outperforms existing methods in classification metrics, providing a new solution for the detection and classification of Android malware.