As the number of investors in the blockchain investment field increases, the impact of ponzi schemes hidden in smart contracts becomes worse. Although some researchers have begun to pay attention to the Ponzi scheme in the blockchain, most of them remain at the level of detection. This paper will conduct further research on the basis of the existing Ethereum Ponzi scheme detection method, and propose a novel Ethereum Ponzi scheme type identification method. The method is based on the source code and transaction record of the smart contracby analyzing the extracted keywords, matching the keywords with the source code of the contract to be tested, and then combining the logic of the transaction record, performing a secondary analysis to determine which type of scam the contract belongs to. Experiments on the real dataset of Ethereum show that the classification accuracy of the method can reach 80% compared with the results of manual classification. This study will help researchers and investors better understand the nature of ethereum smart contract ponzi scheme.