Research On Projects Quality Early Warning Based On Improved Support Vector Machine

With the rapid growth of China’s national economy and urbanization process, construction shows the tend of huge investment, complex technology and using new materials. The complexity of project quality brings difficulties and challenges for the construction quality supervision.However, the current construction project quality supervision means are simple and lack informationization and intelligence. There are still phenomena of ex post regulation during quality management. This situation can not solve the complex and non-linear quality management.This study argues that solving complex and non-linear engineering quality supervision requires more advanced scientific methods and models for quality early warning, so as to reduce unnecessary losses. After studying research status of construction project quality supervison,control and quality warning, this study proposes to build construction project quality early warning mode based on improved SVM model, hightlighting the quality of preventive supervison, science, timeliness and comprehensiveness, when constructing the quality early warning, this study improve SVM model from two aspects of tecnology and application improvement, in order to allow SVM to own better early warning capabilities, to meet the actual regulation and to reflect preferences engineering. In the technology improvement aspect, this study first improve genetic algorithm optimization to Adaptive Genetic Algorithm. Then adaptive genetic algorithm is used to find out the optimal SVM penalty factor C and kernel parameter σ, which can avoid the consequences brought about by selected parameters according to experience, and make up for deficiencies of previous theory of intelligent optimization. In the application improvement aspect, this study introduces sample imbalance coefficient μ, based on considering sample imbalace, infrastructure projects and the construction unit risk warning preferences.it can enables quality early warning model eliminates the effects of the sample imbalance and take into account the quality of construction projects and quality regulators risk warning preferences.Finally, based on the early warning model buiding, this study designs quality early warning response mechanism. Designing a brief warning response process based on construction unit can help construction unit build proper quality early warning response and realize the ultimate quality early warning, supervision and correction closed-loop operation.There is a shortage in the traditional method of construction project quality management. This study increases precautionary, science, intelligence and information for construction project quality management through building quality early warning pattern and response mechanisms from the perspective of the construction unit. This study provides valuable ideas and direction for solving complex, non-linear construction project quality mangement.