Abstract:

The firing quality statistics of a basic brick plant were comprehensively researched using R language+WPS,and the analysis tools such as data perspective table,correlation analysis,association analysis,cluster analysis,principal component analysis,linear models,decision trees,random forests,and artificial neural networks.The association between cause and intermediate variables was studied and the main influencing factors were found.The winner from the above tool family was selected to analyze the relations among the cause,intermediate and output variables.According to the found causes of the defects and the harm degree,the targeted governance was adopted,significantly improving the product qualified rate and enhancing the economic benefits.

Key words: traditional materials, data analysis, model evaluation, quality management, open source software

摘要：

使用R语言+WPS,综合运用数据透视表、相关分析、关联分析、聚类分析、主成分分析等数据分析工具,以及线性、决策树、随机森林、人工神经网络等模型及其评估方法研究了某碱性砖企业的烧成质量统计数据,利用上述工具研究了原因变量和中间变量的联系,找出了主要影响因素,从上述模型族中选出优胜者,研究了原因变量、中间变量和输出变量之间的关系。根据所发现的缺陷规律及其成因,进行了有的放矢的治理,取得了显著提高质量的成效。

关键词: 传统材料, 数据分析, 模型评估, 质量管理, 开源软件