本文選題：Dirichlet過程混合模型　切入點：分布式聚類　出處：《系統(tǒng)仿真學報》2017年12期

【摘要】：為了處理大量分布式存儲的農(nóng)場環(huán)境數(shù)據(jù),為作物增產(chǎn)提供異常環(huán)境參考并制定預防策略,本文結(jié)合農(nóng)場環(huán)境數(shù)據(jù)的特點,在Hadoop平臺中實現(xiàn)了對農(nóng)場環(huán)境數(shù)據(jù)的Dirichlet過程混合模型聚類,并提出了基于聚類分析的農(nóng)場環(huán)境異常檢測方法。在Map Reduce框架下,Map階段完成樣本點到模型的分配;Reduce階段對模型與類簇個數(shù)進行更新。通過實驗驗證了分布式Dirichlet聚類的性能,分析結(jié)果表明該方法可以應用于大量農(nóng)場環(huán)境數(shù)據(jù)的異常檢測。
[Abstract]:In order to deal with a large number of distributed farm environmental data, provide abnormal environmental reference for crop yield increase and formulate prevention strategies, this paper combines the characteristics of farm environmental data. The Dirichlet process hybrid model clustering of farm environment data is realized in Hadoop platform. A method of farm environment anomaly detection based on cluster analysis is proposed. The model and cluster number are updated in the phase of distribution from sample points to models in the Map Reduce framework. The performance of distributed Dirichlet clustering is verified by experiments. The analysis results show that this method can be applied to the anomaly detection of a large number of farm environmental data.
【作者單位】： 上海大學機電工程與自動化學院;上海市電站自動化技術(shù)重點實驗室;
【基金】：上海市科委重點項目(14DZ1206302)
【分類號】：S126;TP311.13
，

本文編號：1683723