API Completion Recommendation Algorithm Based on Programming Site Context

In the process of software development, developers often face a large number of APIs. However, due to the numerous APIs and issues such as missing or incorrect documentation, developers find it difficult to learn and organize various APIs. Nowadays, development tasks cannot be separated from feature-rich and extensive third-party libraries. To address these problems, a feasible solution is to utilize deep learning to extract features of the API's associated object types from the programming context and make API completion recommendations. Based on this, the top N recommendations can be provided to programmers. This paper proposes an API completion method that takes the object types of APIs in the programming context as features. This method first extracts the API call sequence from the programming context and groups them according to object types. Then, it uses a deep learning model to encode them individually, generates the context state of the programming scene through the object state, calculates and analyzes the APIs contained in the object type of the API to be completed, and thus obtains the predicted top N API recommendation output. Experimental results show that this method can recommend relevant APIs to developers based on the analysis of object types in the programming context, with a top-10 recommendation accuracy of up to 95.72%, which is significantly better than several compared API completion models.