A novel adaboost based algorithm for processing defect big data

Abstract

In the practice applications of defect detecting, large amounts of data need to be analyzed. In this paper, a new analysis method is developed based on adaboost algorithm. By using neural networks with a fixed structure, a series of models are built which may be not accurate. Error rates of the models are computed to gain and adjust the weights of every model. A higher accurate model is built by the models and weights. Compared with traditional neural network method, this adaboost based method does not need to adjust the node numbers of neural networks. In addition, it remains accuracy and reduces complexity. Finally, an example is given to demonstrate the effectiveness and advantages of the methods.