As blockchain technology matures and becomes more prevalent, its applications are expanding into areas like finance and supply chains. Smart contracts, a core blockchain technology, are becoming increasingly important due to their decentralized and self-executing nature. However, once deployed, smart contracts are hard to modify, posing significant security risks. If malicious code such as Ponzi schemes is present, irreversible financial losses may occur. Effective detection of Ponzi schemes before deployment is thus critical. In this study, we introduce a new detection method using smart contract transaction graph representation. We analyze the control flow logic of contract opcodes and cross-contract data dependencies through static analysis. This approach builds a multi-dimensional transaction graph capturing behavior patterns, fund flows, and state changes. Additionally, we extract transaction semantic features from the transfer application programming interface, to generate high-quality graph data highlighting key Ponzi scheme characteristics. Experimental results show our method improves detection accuracy for complex Ponzi schemes while maintaining high efficiency. This makes it highly valuable for enhancing blockchain security.