The Study On Classification Model And Fitting Model Of Industrial Bigdata Based On Neural Network

With much more development of the industrial big data during these years,the framework and processing methods of big data analysis becomes even more various.Big data analysis takes increasingly important places in all walks of life,and the data means assets.Big data analysis methods also bring significant changes into industry because industry creates data no less than any other vocation.Machine learning and neural network produce more chances to big data analysis as they grew up so rapidly recent years.Neural network is different from traditional data analysis methods,it don’t need to design all the processing steps one by one,one of it’s training mode is supervised learning,which just need some data with labels,and the model "learn" to analyze data automatically.These processing methods greatly enhanced the efficiency of data analysis.This paper started from examples,discussed the classification model of industrial big data,and the fitting model for Fourier transformation served for ultrasonic detection.This paper discussed the advantages and disadvantages of several optimization methods,and to achieve higher accuracy rate,greater analysis speed,I used the result to design the best neural network model for data analysis.This paper used the most popular framework of deep learning-TensorFlow to design,train and apply neural network,this can be more convenient compared to the traditional methods,and make establishing a neural network more efficient.This paper focuses on big data processing from the following aspects:1)Big data modeling technology;different big data analysis models,analysis and optimization according to principles.2)Neural network implementation technology based on TensorFlow;TensorFlow’s frame structure;calling and implementing neural networks using TensorFlow.3)The application of industrial big data classification models;different optimization methods;compares the differences in different optimization methods according to principles.4)Ultrasonic detection and processing ultrasonic detection signals via Fourier transform;sample acquisition and training of fuzzy Fourier transform;analysis of results.