This invention discloses a ten-
stream integrated financial management method and
system based on a tree neural network, belonging to the field of
construction engineering data processing and
artificial intelligence technology. Addressing the shortcomings of independent collection and lack of linkage
verification of ten categories of multi-source heterogeneous data (contracts, invoices, materials, machinery, labor, expenses, construction,
documentation, funds, and maintenance) in construction projects, this invention encodes the feature vectors of the ten-
stream data into
decision tree nodes. A
decision tree algorithm is used for hierarchical judgment and anomaly localization. A
recurrent neural network is used to perform time-series modeling of the node feature sequences, outputting the matching degree status. An adaptive feedback module feeds the matching degree status back to the
decision tree, dynamically adjusting the warning threshold and model weights. This achieves multi-source cross-validation and automatic anomaly identification of the ten-
stream data, solving the problem of difficult linkage analysis of high-dimensional sparse data. Verified in actual projects, operating costs are reduced by more than 13%, and
management efficiency is improved by 88%. This invention is applicable to various construction companies. This technical solution has been evaluated and certified by the Ministry of Housing and Urban-
Rural Development.