Big Data Classification Research Based On Granular Computing And Deep Learning

With the era of artificial intelligence,data becomes more and more indispensable,but because of the large-scale,multi-modality and growth characteristics of big data,traditional data analysis theories,methods and technologies are faced.Serious challenges such as computability,effectiveness and timeliness.At the same time,big data itself contains infinite value.Mining potential value from data is the meaning of big data.The variety and variety of big data makes it difficult for us to measure the value of it.The emergence of neural networks allows us to understand the internal structure of data and the correlation between data.The more data,the more hidden value,and the complexity of the model of neural network is increasing.high.The rise of deep learning has improved the understanding of neural networks,but the demand for time and space has also increased.Granular computing is a discipline that specializes in the theory,techniques,and tools of thinking modes based on granular structures,problem solving methods,and information processing models.It is a new computing paradigm that has gone through the field of intelligent information processing.Granular computing can store and process data in the form of "grain",saving time and space costs,and improving the learning ability of deep neural networks.This paper begins with the traditional BP nerve as the starting point,and derives the deep learning model-depth belief network,and establishes a deep learning model for the classification of big data.In the deep learning model,in view of the deep learning model learning and convergence ability,it is proposed to use the quantitative particle swarm optimization algorithm to optimize the deep learning model,which not only accelerates the convergence ability of the deep learning model,but also improves its understanding ability.Although the quantitative particle swarm optimization algorithm optimizes the deep learning model and improves the convergence speed,it does not reduce the complexity of the deep learning model.As the depth learning model is deeper,the complexity requirements of the model are higher.Combining with the theory of granular computing,this paper proposes a deep granular computing learning model combining granular computing and deep learning models,which not only improves the convergence ability and comprehension ability,but also reduces the complexity of deep learning.Cognitive maps are another form of network that is different from common feedforward neural networks.Cognitive maps can also use nodes and side information to construct a network,which also has strong learning ability.This paper also combines granular computing with cognitive graph model,proposes a granular computing cognitive map RS-FCM,and uses real-valued coding genetic algorithm to optimize RS-FCM.Experimental verified that compared with the traditional cognitive graph model,the learning ability of granular computing cognitive model has been greatly improved.