Abstract: Web API is a lightweight solution based on Web services, representing the smallest reusable component. By combining a variety of Web APIs, it is possible to create a combined application Mashup with high-level functions to achieve business value-added requirements. The rapid growth of the number and types of Web APIs has made it more difficult to discover suitable Web APIs for Mashup creation. To solve this problem, a feature representation reinforcement Web API recommendation method is proposed to efficiently perform Web API recommendation and improve the efficiency of Mashup creation. First, it map the functional text of Mashup to the vector space for feature comparison to obtain a neighbor Mashup similar to the target Mashup. Then, it use the neural network-based feature extraction model to learn the text features of the target Mashup and neighbor Mashup. The extracted features are combined with category features for representation reinforcement. Finally, Web API recommendation is performed based on the reinforced semantic features. Experimental results show that this method can effectively recommend Web API, and has achieved significant improvement in multiple indicators.