The traditional ship repair settlement process often involves tedious manual operations, including referring to price books, manually entering base prices, coefficients, etc., and using tools like calculators and Excel for calculations. To improve efficiency, some companies have established settlement systems and electronic price libraries, simplifying the settlement process to where the settlement personnel manually select the appropriate settlement codes based on the project content, and the system then automatically calculates based on the quantities. However, the step of manually matching settlement codes is prone to errors and time-consuming, directly affecting the efficiency of the settlement. To address this issue, this paper proposes an intelligent matching method for ship repair settlement codes based on the fusion of multiple features. It uses BERT to represent the project content text as word vectors, extracts contextual features of the text through BiLSTM, and uses CNN to extract local features of the text. The outputs are subjected to average pooling, and features from different sources are fused into a single feature vector, which is then used to obtain the corresponding settlement code through the output layer. The results show that the overall accuracy of the multi-feature fusion model is 85.48%.