Wellbore construction engineering is an important intermediate project connecting reservoir and surface. In order to ensure the analysis effect and calculation efficiency of wellbore condition diagnosis, lift design, injection and production optimization, dynamic production data analysis, oil recovery engineering theory and mechanics theory should be comprehensively adopted. Traditional methods of wellbore lifting and injection-production analysis include calculation of pipe flow equation, theory of material mechanics and empirical formulas, diagrams and other methods. These methods are more dependent on the experience of technical personnel and have limited expansibility in the face of complex problems. Combining business knowledge with machine learning theory is an important means to break through the upper limit of existing theoretical methods. Compared with traditional methods, machine learning has stronger adaptability and the ability to excavate complex rules. The research group proposed to combine machine learning and deep learning with traditional oil production engineering, and combined injection and production system through connectivity analysis, focusing on solving key problems such as integration, comprehensive utilization of complex data, small samples and method expansibility in the process of diagnosis optimization, and forming a set of methods with high theoretical level and practical value. Furthermore, an intelligent wellbore diagnosis and injection and production optimization analysis platform based on micro-service was built.