With the explosive use of the information technology and the administrative reform progressively intensified in the recent years, the information technology has been playing an important role in different fields, especially in the public administration area in China. Information and Communication Technology (ICT) has been applied in security, Social Networking Services (SNS), E-commerce, E-government, and etc. E-government integrates various applications and technologies into governance. Egovernment websites are rapidly popularized based on the different needs, such as information disclosure, management and service, interactive communication, public opinion, website functions and management. Using E-government to improve the access and the quality of public services is a political, social and economic imperative for all developing countries. In order to evaluate E-government websites’ construction and popularity rates, this paper presents a novel method to evaluate the performance of the E-government websites using neural network technology. The method is based on Principal Component Analysis (PCA) and the main factors are picked up. Back Propagation (BP) neural network is applied in this field which can weaken the randomness and appraiser’s subjectivity. In order to compare and demonstrate which one is most appropriate, we study the other three classifiers for comparison which are the Support Vector Machine (LIBSVM), Decision Tree J48, and the rule algorithm of JRIP. Through comparing the experimental results that based on the training models and datasets, the BP classifier rates up to 97.14% and it is better than the other three classifiers. It is effective to address large number of evaluation objects and proved to be superior to other methods. The study highlights the effectiveness and efficiency of the application of BP algorithm