Abstract: Software reliability assessment performance directly affects the workload of software testing. This paper proposes a neural network software testing method based on dynamic weighted combination for fault detection and fault introduction in software testing process. This method considers the diversity of software engineering. The neural network method is used to construct the dynamic weighted combination model, and the fault detection process is combined with the fault introduction process. Through the test of two sets of real failure data sets (DS1 and DS2), the proposed method is compared with the existing Software Reliability Growth Models (SRGMs), and the results show that the dynamic weighted combined neural network considering the fault introduction is considered. The network model has the best fitting effect, showing better software reliability evaluation performance and model versatility.