| With the rapid development of information,the medical data from wearable equipment,electronic medical records,portable monitors and others has rapid growth,and storage structure diversification.The traditional storage structure and computing model can not solve the problem of storage and calculation of these data well.At the same time,the growth of large medical data can solve the problem of insufficient data samples in traditional machine learning methods,but the computing power of stand-alone data processing can not meet the requirements,and professional personnel are required to carry out artificial extraction of relevant data characteristics.The extraction process of the feature is cumbersome and influenced by the subjective factors of the experts,affecting the results of the analysis.The innovations involved are below:(1)In order to solve the above problems,this paper builds a cluster analysis platform to solve these problems.The platform uses HDFS in Hadoop to store large-scale unstructured data and uses the MapReduce computing framework to analyse the large data sets.Kafka and Storm are adopted to improve the overall performance of the platform.(2)We realizes the parallelization of BP neural network based on MapReduce.The experimental results show that the accuracy of the analysis can be improved effectively and the training time can be reduced.(3)In view of the shortcomings of traditional extraction methods in traditional machine learning methods,we try to use the deep learning methods and apply it to physiological signal processing.At present,there are few studies on physiological signal research based on deep learning in China.In this paper,we use DBN,CNN and SAE networks in depth to obtain abstract features to avoid manual interference and combine the traditional classifier SVM for automatic classification.The experimental results show that the proposed methods have a good effect. |
