The combined approach proposes the use of PBFT and Raft to ensure data consistency and fault tolerance in the system, and also integrates recurrent neural networks to analyze and predict the behavior of nodes in the network. RNNs can be used to detect anomalies, predict system load, and analyze time series data related to node operation. The proposed combined approach opens up new prospects for the development of distributed systems, increasing their reliability, fault tolerance and adaptability to changing conditions. Further research in this direction could lead to more efficient and secure distributed systems that can efficiently handle complex real-world scenarios.