基于蒙特卡洛k-means聚类算法的舰船器材分类研究

Research on Warship Spare Parts Cluster method Based onMonte Carlo K-means Cluster Algorithm

摘要:

器材合理分类是建立预测模型的基础,某型舰船仪表器材数据较少、分类指标因素不足,使用传统方法易产生过拟合的问题。提出蒙特卡洛K-means算法,利用样本方差进行器材消耗聚类分析。该方法首先利用MC法计算初始聚类中心,参考SBC分类法制定聚类种类数k,通过方差聚类建模来优化仪表器材的分类,最终得到仪表器材的聚类结果。实例计算表明,该方法能够有效改进K-means方法的分类结果,无需考虑其他备件指标因素影响,适用于数据量过小和存在白噪声的模型。

Abstract:

The reasonable classification of spare-parts is the basis of establishing the prediction model, the data of a certain warship spare-parts is less and the classification index factor is insufficient, so it is easy to use the traditional method to produce the problem of overfitting. A Monte Carlo K-means algorithm is proposed, and the sample variance is used for the spare-parts consumption volatility cluster analysis. Firstly, using Monte Carlo to calculate the initial clustering center, and refers to the SBC method to formulate the number of clustering categories k. The classification of instrument spare-parts is optimized by variance clustering modeling, and finally the cluster results of instrument spare-parts are obtained. The example shows that the method can effectively improve the classification results of K-means method without considering other index factors.It is suitable for the model with too small amount of data and white noise.