数据处理方法.ppt

《数据处理方法.ppt》由会员分享，可在线阅读，更多相关《数据处理方法.ppt（11页珍藏版）》请在淘文阁 - 分享文档赚钱的网站上搜索。

1、数据预处理方法Data preprocessing method一、评价指标的类型一致化Evaluation index type uniformization评价指标可以分为三类;正指标、逆指标和适度指标。有些指标取值越大越好，一般称之为正指标，如产值、利润；有些指标取值越小越好，一般称之为逆指标，如单位产品成本；还有些指标是取值越接近某一确定值越好，一般称之为适度指标，如资产负债率，投资率。Evaluation index can be divided into three categories;Positive index,reverse index and moderate index

2、.Some index value the higher the better,general called positive index,such as output,profit；Some index value the lower the better,generally called reverse index,such as unit product cost；Also some value index is the more close to a fixed value the better,generally called appropriate index,such as as

3、set liability ratio,rate of investment.逆指标转化为正指标Reverse index into positive index =,(0)i=1,2,.m,=可以是负值，i=1，2，.m (May be negtive value)=可以是负值，i=1，2，.m 以上各式中的 是对第i个参评单位、第j项逆指标的实际值进行转化后生成的正指标数值，是第i个参评单位、第j项评价指标的实际值。i=1，2，.m;j=1,2,.,n.The above all the formulas is the generated positive value transforme

4、d from the actual value of the number i eligible unit and number j item reverse index,is the actual value of number i eligible unit and number j item index.i=1，2，.m;j=1,2,.,n.适度指标转化为正指标Moderate index into positive index 设某一适度指标的理想值为,则 与 的距离越小越好，因此 与 的距离 相当于一个逆指标，因此有：Set up an ideal value fo a modera

5、te index,then the distance between and may the smaller the better.Therefore the distance is equivalent to a reverse index,so:=i=1，2，.,m 是对各评价对象的适度指标进行转化后生成的正指标值；是第i个参评单位、第j项指标的实际值；一般取c=a。Is the generated positive valua of the object moderate indexes after transformed；is the actual value of number i

6、eligible unit and the number j item indexes ；generally order c=a。二、评价指标的无量纲化Evaluation indexes dimensionless 由于各指标计量单位不同（具有不同量纲）而无法直接汇总，因此，需要进行消除量纲处理。Because of different measurement unit between indexs (with different dimensional),so they cant do direct summary,therefore,need to eliminate dimension

7、al processing.相对化处理方法 Relativization processing method标准化处理 Standardized processing 相对化处理首先必须对每个指标确定一个基准值，比如可以选择该指标在各评价对象中的平均值、最大值、历史最高水平等。Fist we must set a benchmark for each index when we give a relativization processing ,for example we can choose the the mean value,maximum value,historical highe

8、st level,etc.of the indexs in the evaluation object 正指标：=，逆指标：=(positive index)(reverse index)为第i个参评单位、第j项指标实际值，为第j个评价指标基准值，为经过相对化处理的无量纲评价值。is the actual value of number i eligible unit and number j indexes,is the basic evalua of number j index,is the dimensionless evalua after relativization proces

9、sing 前提：需进行标准化处理的变量要服从正态分布Prerequisite:the variables needed to standardized processing must be obey the normal distribution1.求出变量的数学期望 和标准差 ；Find out the variables mathematical expectation and standard deviation ;2.进行标准化处理：Standardized processing若指标中有逆指标，改变处理后指标的符号。If there are reverse indexs,change the index sign after processing.