Improved Research And Applications Of Fuzzy Min-Max Neural Network

With the rapid development of information technology,global data shows explosive growth.Both the amount of data and the type of data are constantly rising,and the research on data has become more and more complex,and there is often a wealth of value and knowledge behind the massive data.Data mining is an emerging technology dedicated to extracting valuable information,knowledge,internal laws or potential patterns from massive amounts of data.It can realize the transformation of data from numerical value to value and knowledge.However,in the face of the complex data types and huge amount of data in the era of big data,data mining technology is also facing unprecedented challenges.At present,classification and clustering algorithms are the two key research contents of data mining.Classification is to derive rules through training from data with existing labels,and apply the rules to classify other data.Clustering is to divide the data according to similarity into several clusters that are relatively similar but differ significantly from each other.The fuzzy min-max neural network is a neural network that can be used for both classification and clustering for online training,avoiding the shortcomings of traditional networks.This paper has done the following work around the improvement research and application of fuzzy min-max neural network:First of all,this paper chooses the most popular variant of fuzzy min-max neural network to strengthen the fuzzy min-max classification neural network for improvement.The performance of the algorithm is improved by replacing the membership function in the enhanced fuzzy min-max classification neural network with the inclusion measure in the fuzzy lattice theory.And it was verified on the public data sets.The experimental results show that the improved classification algorithm performs well in Precison,F₁ score,and Accuracy.Secondly,the effectiveness of the improved method of enhanced fuzzy min-max classification neural network and the improved method combined with the inclusion measure of the fuzzy lattice has been proved.Therefore,this paper tries to extend the rule improvement of the enhanced fuzzy min-max classification neural network to the original fuzzy min-max clustering neural network,and at the same time adopts the fuzzy lattice inclusion measure as the membership function.Experiments on the public data sets show that the improved clustering algorithm performs well in terms of Precison,adjusted RAND index,CH index and other indicators.Then,the improved fuzzy min-max classification neural network based on the fuzzy inclusion measure is applied to the online sales forecasting problem of agricultural products,taking corn sales data as an example.By dividing corn sales data into corresponding pre-set sales grade categories,the forecast of corn sales can be realized.And the improved fuzzy min-max clustering neural network based on the inclusion measure of fuzzy lattice is applied to the problem of agricultural circular economy region division,taking Heilongjiang Province as an example.Sixteen indicators were selected to form a region division indicator system for agricultural circular economy.Heilongjiang Province is divided into 4 agricultural circular economy development zones,and the resources of each zone are comprehensively analyzed,and the industrial characteristics give development suggestions suitable for each region.This paper proposes two improved versions of fuzzy min-max neural networks,respectively for classification and clustering.And proved the effectiveness and reliability of the two improved models in the public data sets.The improved method proposed in this paper can improve the similarity measurement ability in classification and clustering methods,reduce the original algorithm parameters,and also improve the limitations of the original algorithm operation space and the data type that can be processed,and it is more in line with today's complex data types in this era of big data with huge amount of data.At the same time,the improved method has also been successfully applied to problems in the agricultural field,successfully solving the problem of online sales forecasting of agricultural products and the problem of agricultural circular economy region division.It shows that while the method has theoretical significance,it also has rich practical significance.