Research On Machine Learning Technology Based On Hyperledger Fabric Blockchain

With the advent of the big data era,it is urgent to design efficient machine learning algorithms to analyze large-scale data.In practice,data is usually generated by multiple participants.Some data may involve security issues such as privacy.There are islands and monopolies in the data,which is extremely unfavorable to the model construction task that requires a large amount of data to complete training.The storage,security and transmission of data have stimulated the research of blockchain.Blockchain technology has also received financial support from many government departments in recent years,and its development momentum is very fast,and the rapid development of blockchain technology has also been widely valued by all sectors of society.Combined with blockchain and machine learning,this paper designs an architecture for machine learning on hyperledger fabric blockchain to help solve the problems related to data security in machine learning.In the way of empirical research,this paper discusses the effectiveness of machine learning on hyperledger fabric chain.Two classification models with the same parameter combination,logical regression and decision tree,are constructed on the telecom operator customer churn data set,and machine learning is carried out on and off the chain respectively.Through visual analysis,this paper finds out the characteristics of high churn customers from three perspectives: user type,service attribute and contract information,and outputs the portrait of high churn customers.Under the chain,a series of preprocessing work is done for the data set,such as data conversion,data cleaning and so on.In order to reduce data redundancy,feature selection is carried out by combining data analysis and Pearson correlation coefficient method,and three features with weak influence on the results are eliminated.In the process of model training,in order to further improve the accuracy of analysis,the best selection of modeling data is determined through grid search,cross test and setting different thresholds.The classification accuracy of the logistic regression model on the chain is 81.59%,which is consistent with the results under the chain.Similarly,the classification accuracy of the decision tree model on the chain is slightly higher than that of the logistic regression model with 82.50%,and its accuracy is completely consistent with that under the chain.Experiments show that under the same data set and the same parameter combination,the classification accuracy on the chain is completely consistent with that under the chain.Machine learning on the chain is feasible and effective.While solving the problems related to data security in machine learning,through the deployment of intelligent contracts on the chain,the construction of machine learning model can be completed only by calling contract commands,which greatly improves the work efficiency.