Research And Application Of Parallel Fuzzy Rule Classification Algorithm Based On Mapreduce

Classification problem is one of the important research content of pattern recognition and machine learning,which is extensively used in industry,business and scientific research.The classification algorithm based on fuzzy rules has the advantages of high classification accuracy,whose results are semantic,interpretability and easy to be understood by users.In the current big data environment,the emergence of the parallel computing can effectively solve the problems that the traditional single computer is time-consuming,inefficient and even impossible to deal with in the face of large-scale data sets.The MapReduce model,proposed by Google,is an easy-to-develop programming model and can process large data sets in parallel.Besides,the model can decrease the complexity of the parallel program.Furthermore,it only needs to design parallel computing tasks for the people who using the MapReduce model,which can reduce the time greatly and have high efficiency.This thesis presents a parallel fuzzy rule classification algorithm based on MapReduce model.This algorithm uses parallel computing method to extract fuzzy rules and builds fuzzy rules classifier,it has the advantage of fuzzy system in dealing with uncertain problem and the parallel computing ability of MapReduce in big data.In the experimental research part,the parallel algorithm is applied to the futures automatic trading platform named TradeBlazer(TB)and the label printing production task of printing plant.The main contents are as follows:(1)The proposed parallel algorithm is applied to the futures trading.The fuzzy rules are extracted from the futures data in the first.Secondly,the fuzzy rules are converted into TB formula to form a trading strategies which are applied to the automated trading platform TB.Thirdly,the utility of the fuzzy rules is judged by the profit and loss situation of the simulated transaction.Experimental results show that the proposed approach can effectively reduce the data processing time and has good scalability,the extracted fuzzy rules have good returns in futures trading,which also proves the availability and validity of the rule.Moreover,the rules are semantic and have certain guiding significance for the investors to make decisions.At the same time,this method can also provide a new way of futures program trading.(2)The parallel algorithm is utilized to the production process of traditional printing plant and the intelligent production of each order for label printing production is carried out.Using the proposed algorithm to extract fuzzy rules,which are based on history data that the workshop managers assigned task for every machine according to their experience.Then it can be used to construct the fuzzy rules classifier,so that the new task order is reasonably arranged to process on the corresponding machine.The result shows that the algorithm can improve the speed of data processing,and the extracted rules has good semantic representation,which can substitute the production managers to carry on the order task allocation to a certain extent and improve the efficiency of the industrial production of printing.