利用遗传算法优化抽样调查的数据质量

Improving the Quality of Survey Data By Using GA

对居民家庭进行抽样调查时,常常出于隐私或缺乏专业知识背景的原因,使得居民对被调查的经济变量不愿或无法给出准确值,从而影响调查数据的质量.在此情况下,国外知名的调查(如,美国的消费金融调查)会将调查变量的取值范围划分成一系列连续的、长度不等的区间,请受访者选择,以此来消除受访者的戒备和疑虑心理,收集家庭的经济变量的信息,但这种区间的划分是一种经验划分.从降低数据分析误差的角度提出了一种优化区间长度的设计思路,利用遗传算法给出了问题求解的方法,最后通过一个实际例子加以说明.方法对提高调查变量的数据质量颇有帮助.

In a large sample survey about household, in order to keep ones privacy, the respondent usually unwanted to report the true value of their economic variables, for example, income or asset. A new method of partitioning range of economic variable into small intervals and asking the respondent chose the one his variable satisfied can avoid this case. In this paper, we proposed an idea of designing the intervals, and give a practical case with an answer solved by using the CA method.