基于Grubbs规则和MATLAB语言快速剔除异常值方法的建立及其在药物苦度评价中的应用(1)
中圖分类号 R943；TP311.1 文献标志码 A 文章编号 1001-0408(2019)02-0176-07
DOI 10.6039/j.issn.1001-0408.2019.02.07
摘 要 目的：建立基于Grubbs规则和MATLAB语言的异常值剔除方法，并评价其在药物苦度评价中的的应用效果。方法：以Grubbs规则为参考，建立基于MATLAB语言的异常值自动循环剔除方法。选择20名志愿者分别进行单组口尝试验(通草)和多组口尝试验(通草、明党参、茯苓等10种药材)；选择7个传感器进行电子舌测试(川木通)。以上述试验所得的苦度评价数据(口尝试验为苦度值，电子舌测试为传感器响应值)为数据源，选择5名研究者，采用基于Grubbs规则的查表逐一剔除法(方法一)、基于Grubbs规则的Excel软件剔除法(方法二)、基于Grubbs规则和MATLAB语言的异常值自动循环剔除法(方法三)进行异常值的判定及剔除；以异常值剔除时间和错误概率为指标，评价上述3种方法的应用效果。结果：单组口尝试验苦度评价数据中有2个异常值，3种方法的剔除时间分别为(745.400 0±25.904 4)、(288.333 3±31.253 1)、(0.000 3±0.000 0)s，错误概率分别为20.0%、0、0；多组口尝试验苦度评价数据中有6个异常值，3种方法的剔除时间分别为(3 693.107 7±75.023 3)、(1 494.761 4±53.826 9)、(0.005 2±0.000 0)s，错误概率分别为10.0%、4.0%、0；电子舌测试苦度评价数据中有3个异常值，3种方法的剔除时间分别为 (2 992.673 3±84.117 6)、(1 276.367 1±55.024 5)、(0.002 3±0.000 0)s，错误概率分别为5.7%、2.9%、0。3种方法的剔除结果一致；方法二的剔除时间显著短于方法一(P<0.01)，方法三的剔除时间显著短于方法一和方法二(P<0.01)；3种方法错误概率无显著差异(P>0.05)。结论：基于Grubbs规则和MATLAB语言的异常值自动循环剔除法可显著缩短苦度评价数据异常值的剔除时间，提高数据处理效率，可用于药物苦度评价。
关键词 Grubbs规则；MATLAB语言；异常值；剔除；药材；苦度评价
ABSTRACT OBJECTIVE： To establish the elimination method of outliers based on Grubbs rule and MATLAB language， and to evaluate the effects of it on drug bitterness evaluation. METHODS： Referring to Grubbs rule， the automatic cyclic outliers elimination method based on MATLAB language was established. Totally 20 volunteers were included in single oral taste test (Tetrapanax papyrifer) and multiple oral taste test (10 kinds of medicinal material as T. papyrifer， Changium smyrnioides， Poria cocos， etc.). Seven sensors were selected for electronic tongue test (Clematis armandii). The data of bitterness evaluation in above tests (oral taste test as bitterness value， electronic tongue test as response value of sensors) were used as the data source. Five researchers were selected and adopted table-by-table elimination method based on Grubbs rule (method one)， Excel software elimination method based on Grubbs rule (method two) and automatic cyclic outliers elimination method based on Grubbs rule and MATLAB language (method three) to judge and eliminate the outliers. The effects of above three methods were evaluated with the removal time and error rate of outliers as indexes. RESULTS： There were two outliers in the data of bitterness evaluation in single oral taste test； the elimination time of the three methods were(745.400 0±25.904 4)，(288.333 3±31.253 1)and(0.000 3±0.000 0)s， respectively； error rates were 20.0%， 0 and 0， respectively. There were six outliers in the data of bitterness evaluation in multiple oral taste test； the elimination time of three methods were (3 693.107 7±75.023 3)， (1 494.761 4±53.826 9)， (0.005 2±0.000 0)s， respectively； error rates were 10.0%， 4.0%， 0， respectively. There were three outliers in the data of bitterness evaluation in electronic tongue test； the elimination time of three methods were (2 992.673 3±84.117 6)， (1 276.367 1±55.024 5)， (0.002 3±0.000 0)s， respectively； error rates were 5.7%， 2.9%， 0， respectively. The elimination results of the three methods were consistent. The elimination time of method two was significantly shorter than that of method one (P<0.01)； the elimination time of method three was significantly shorter than those of method one and method two (P<0.01). There was no significant difference in error rate of 3 methods (P>0.05). CONCLUSIONS： The automatic cyclic elimination method of outliers based on Grubbs rule and MATLAB language can significantly shorten the elimination time of outliers in data of drug bitterness evaluation， improve the efficiency of data processing， and is suitable for drug bitterness evaluation., http://www.100md.com(刘瑞新 王艳丽 张耀 桂新景 王君明 王青晓 姚静 张璐 施钧瀚 李学林)
DOI 10.6039/j.issn.1001-0408.2019.02.07
摘 要 目的：建立基于Grubbs规则和MATLAB语言的异常值剔除方法，并评价其在药物苦度评价中的的应用效果。方法：以Grubbs规则为参考，建立基于MATLAB语言的异常值自动循环剔除方法。选择20名志愿者分别进行单组口尝试验(通草)和多组口尝试验(通草、明党参、茯苓等10种药材)；选择7个传感器进行电子舌测试(川木通)。以上述试验所得的苦度评价数据(口尝试验为苦度值，电子舌测试为传感器响应值)为数据源，选择5名研究者，采用基于Grubbs规则的查表逐一剔除法(方法一)、基于Grubbs规则的Excel软件剔除法(方法二)、基于Grubbs规则和MATLAB语言的异常值自动循环剔除法(方法三)进行异常值的判定及剔除；以异常值剔除时间和错误概率为指标，评价上述3种方法的应用效果。结果：单组口尝试验苦度评价数据中有2个异常值，3种方法的剔除时间分别为(745.400 0±25.904 4)、(288.333 3±31.253 1)、(0.000 3±0.000 0)s，错误概率分别为20.0%、0、0；多组口尝试验苦度评价数据中有6个异常值，3种方法的剔除时间分别为(3 693.107 7±75.023 3)、(1 494.761 4±53.826 9)、(0.005 2±0.000 0)s，错误概率分别为10.0%、4.0%、0；电子舌测试苦度评价数据中有3个异常值，3种方法的剔除时间分别为 (2 992.673 3±84.117 6)、(1 276.367 1±55.024 5)、(0.002 3±0.000 0)s，错误概率分别为5.7%、2.9%、0。3种方法的剔除结果一致；方法二的剔除时间显著短于方法一(P<0.01)，方法三的剔除时间显著短于方法一和方法二(P<0.01)；3种方法错误概率无显著差异(P>0.05)。结论：基于Grubbs规则和MATLAB语言的异常值自动循环剔除法可显著缩短苦度评价数据异常值的剔除时间，提高数据处理效率，可用于药物苦度评价。
关键词 Grubbs规则；MATLAB语言；异常值；剔除；药材；苦度评价
ABSTRACT OBJECTIVE： To establish the elimination method of outliers based on Grubbs rule and MATLAB language， and to evaluate the effects of it on drug bitterness evaluation. METHODS： Referring to Grubbs rule， the automatic cyclic outliers elimination method based on MATLAB language was established. Totally 20 volunteers were included in single oral taste test (Tetrapanax papyrifer) and multiple oral taste test (10 kinds of medicinal material as T. papyrifer， Changium smyrnioides， Poria cocos， etc.). Seven sensors were selected for electronic tongue test (Clematis armandii). The data of bitterness evaluation in above tests (oral taste test as bitterness value， electronic tongue test as response value of sensors) were used as the data source. Five researchers were selected and adopted table-by-table elimination method based on Grubbs rule (method one)， Excel software elimination method based on Grubbs rule (method two) and automatic cyclic outliers elimination method based on Grubbs rule and MATLAB language (method three) to judge and eliminate the outliers. The effects of above three methods were evaluated with the removal time and error rate of outliers as indexes. RESULTS： There were two outliers in the data of bitterness evaluation in single oral taste test； the elimination time of the three methods were(745.400 0±25.904 4)，(288.333 3±31.253 1)and(0.000 3±0.000 0)s， respectively； error rates were 20.0%， 0 and 0， respectively. There were six outliers in the data of bitterness evaluation in multiple oral taste test； the elimination time of three methods were (3 693.107 7±75.023 3)， (1 494.761 4±53.826 9)， (0.005 2±0.000 0)s， respectively； error rates were 10.0%， 4.0%， 0， respectively. There were three outliers in the data of bitterness evaluation in electronic tongue test； the elimination time of three methods were (2 992.673 3±84.117 6)， (1 276.367 1±55.024 5)， (0.002 3±0.000 0)s， respectively； error rates were 5.7%， 2.9%， 0， respectively. The elimination results of the three methods were consistent. The elimination time of method two was significantly shorter than that of method one (P<0.01)； the elimination time of method three was significantly shorter than those of method one and method two (P<0.01). There was no significant difference in error rate of 3 methods (P>0.05). CONCLUSIONS： The automatic cyclic elimination method of outliers based on Grubbs rule and MATLAB language can significantly shorten the elimination time of outliers in data of drug bitterness evaluation， improve the efficiency of data processing， and is suitable for drug bitterness evaluation., http://www.100md.com(刘瑞新 王艳丽 张耀 桂新景 王君明 王青晓 姚静 张璐 施钧瀚 李学林)