The study explores the innovation of economic transaction model in the framework of blockchain technology and optimizes the key algorithms in the transaction. The study integrates the advantages of traditional sharing contracts and smart contracts to construct a blockchain model for sharing economic transaction innovation mode. Based on the feature selection method of chi-square test and feature correlation, the gated recurrent unit (GRU) is combined with the support vector machine (SVM) algorithm to form a detection model for abnormal blockchain transactions. For the Practical Byzantine Fault Tolerance (PBFT) consensus mechanism in the economic model, the study proposes a node pre-prepared layered consensus protocol to optimize it, and introduces a reputation model to rank nodes in terms of their reputation value, enhance the node’s ability to defend itself against Sybil’s witch attack, and improve node’s motivation. The research results show that the anomaly detection model based on feature selection can effectively realize the accurate detection of transaction anomalies, and the detection performance of the GRU-SVM model improves the F1 score by 1.11% in the feature subset than in the full feature dataset. The improved PBFT consensus algorithm has significant improvement over the original PBFT algorithm in terms of algorithmic complexity, communication complexity, and throughput, which effectively guarantees the security of sharing economy transactions.
