An Implementation of Rough Information Analysis Tool and Its Application to Data by Questionnaires

在本研究中，我们开发了一个数据分析工具，它可以处理基于粗糙集的概念，我们将这个工具应用于问卷调查的数据分析。对于数值表，主要采用多变量分析，特别是采用回归分析，从数值表中生成回归线。在数值表中，可以得到数据的均值和方差。然而，在分类数据表中，均值和方差可能没有定义，多变量分析可能无法应用于分析具有分类值的表。对于这样的表，我们应用了基于粗糙集的概念，并且我们已经实现了工具程序，可以处理基于粗糙集的概念。我们使用C语言和Prolog语言，实现了处理基于概念的粗糙集的程序，比如一个集的可定义性程序，一个处理表中等价关系的程序，一个检查数据依赖性的程序，一个计算标准支持度的程序，等等。准确性和覆盖率，从表中生成规则的程序等。根据这些基本程序，我们实现了几个工具程序。我们将这些工具程序应用于学生的课堂评价数据分析，该数据由60名学生对16个问题的评价组成。我们得到了几个结果，最有特点的规律是“如果每个学生都觉得老师在每节课上都很有礼貌，那么这个学生的评价是好的。”每个学生对老师的印象似乎强烈地影响着对班级的评价。

In this research, we have developed a data analysis tool, which can handle rough sets based concepts, and we applied this tool to analyzing data by questionnaires. For tables with numerical values, multivariate analysis has mainly been employed, and especially regression analysis is applied to generate regression lines from tables with numerical values. In tables with numerical values, it is possible to obtain mean and variance of data. However in tables with categorical data, mean and variance may not be defined, and multivariate analysis may not be applied to analyzing tables with categorical values. For such tables, we apply rough sets based concepts, and we have realized tool programs, which can handle rough sets based concepts.We employed C and Prolog language, and realized programs handling rough sets based concepts, like a program for the definability of a set, a program for manipulating equivalence relations in tables, a program for checking data dependency, a program for calculating criteria support,. accuracy and coverage, a program for generating rules from tables, etc. According to these basic programs, we have realized several tool programs.We have applied these tool programs to analyzing class evaluation data by students, which consists of evaluation on 16 questions by 60 students. We obtained several results, and the most characteristic rule is such that "If every student felt that the teacher was polite in every class, the evaluation of this student is good." The impression of a teacher to every student seems to strongly influence the evaluation of the class.