论文部分内容阅读
针对原有千车故障数统计方法上的不足,本文从改进统计方法着手,提出一种新的统计方法即重新定义千车故障数,然后利用数据挖掘中的聚类分析方法将具有相同特征的批次综合起来考虑,建立通用的运筹模型.针对缺失数据、近期预测这两个问题,本文对通用模型进行调整,“学习”出同类数据间的不同权值,然后利用加权数据,并通过拟合曲线来求出预测值.由于远期预测中数据的严重缺乏,则是从纯粹统计学的角度出发,计算得到预测值.预测模型通用性强,适用面较广.本文应用了SAS和MATLAB两种软件来求解上述模型,预测结果准确率较高,并且符合实际情况.
Aiming at the deficiencies of the original method of counting the number of failures of a thousand cars, this paper starts with the improvement of statistical methods, proposes a new statistical method that redefines the number of failures per thousand cars, and then uses the cluster analysis method in data mining In this paper, the general model is adjusted to “learn” the different weights between the same type of data, then the weighted data is used, and through the proposed And predicts the value of the forecasting curve.Because of the serious lack of data in the long-term forecasting, the forecast value is calculated from the purely statistical point of view.The forecasting model is versatile and suitable for a wide range of applications.In this paper, SAS and MATLAB Two kinds of software to solve the above model, the prediction accuracy is high, and in line with the actual situation.