World Journal of Traditional Chinese Medicine

: 2019  |  Volume : 5  |  Issue : 1  |  Page : 9--17

Research on characteristic manifestations of Xiaochaihu decoction based on association rules mining

Li-Ying Zhu1, Chun-Hua Jia1, Eunhee Park2,  
1 Department of Golden Chamber, Preclinical School, Beijing University of Chinese Medicine, Beijing, China
2 Department of Physical Therapy, TaeGu Science University, Daegu, South Korea

Correspondence Address:
Chun-Hua Jia
No. 11 of North 3rd Ring East Road, Chaoyang, Beijing, 100029


Objective: The aim of this study is to explore the characteristic manifestations and contraindications of Xiaochaihu decoction (XD [It also stands for all the seven herbs in XD]) syndrome and to clarify the principle of using XD. Methods: By searching on the Encyclopedia of Traditional Chinese Medicine, 254 cases whose symptoms were relieved by XD and 48 cases who misused XD in ancient time were selected. Moreover, they are divided into two groups. Microsoft Excel 2016 was used as a storage tool. These two groups of cases were all treated with XD because of their similar manifestations, but therapeutic efficacies are reverse. The manifestations mislead doctors should be selected out to clarify the indications of using XD. Thus, the professional data mining software RapidMiner was used as a platform to do the frequent statistic and to create association rules between manifestations and herbs of both groups, respectively. After comparing the results of two groups, we can make a conclusion. Results: The characteristics of XD syndrome are alternative chills and fever, hypochondriac pain, and inflammatory swelling and pain (ear or breast). When any of these manifestations occur, XD can be used. Wiry pulse paired with alternative chills and fever, headache, thirst, or hypochondriac pain is also regarded as the indications of using XD. XD should not treat for fever paired with other manifestations, such as headache, delirium, or aversion to cold. In addition, the manifestations with high concurrence frequency do not mean the high correlation. Moreover, only according to the frequency, statistics cannot make correct conclusions. Conclusions: (1) XD is more fit for the excess syndrome such as excess fire in the liver or gallbladder meridians. (2) The medical records in which the XD is misused are also worth analyzing. (3) Association rules cannot be substituted with frequency statistics.

How to cite this article:
Zhu LY, Jia CH, Park E. Research on characteristic manifestations of Xiaochaihu decoction based on association rules mining.World J Tradit Chin Med 2019;5:9-17

How to cite this URL:
Zhu LY, Jia CH, Park E. Research on characteristic manifestations of Xiaochaihu decoction based on association rules mining. World J Tradit Chin Med [serial online] 2019 [cited 2020 Apr 5 ];5:9-17
Available from:

Full Text


Xiaochaihu decoction (XD) is a famous formula recorded in Shanghanlun (Treatise on Cold-Induced Diseases), a classic medical book written by Zhang Zhongjing in the East Han Dynasty. At present, this prescription is still extensively used to cure a wide range of diseases by traditional Chinese medical doctors.[1] The text where the XD appears for the first time in the book elaborates its ingredients and the manifestations it aims at.

Febrile disease caused by cold or wind

After 5–6 days, when the patient has alternative chills and fevers, he/she feels a distention and a sensation of oppression in the chest and costal region, reluctant to speak and eat, restless and nauseous, and prescribes XD.

Xiaochaihu decoction

Chaihu (Radix Bupleuri) 8 liang

Huangqin (Radix Scutellariae) 3 liang

Renshen (Radix Ginseng) 3 liang

Banxia (Rhizoma Pinelliae) 0.5 sheng

Gancao (Radix Glycyrrhizae Praeparata) 3 liang

Shengjiang (Rhizoma Zingiberis Recens) 3 liang

Dazao (Fructus Ziziphi Jujubae) 12 pcs.

Stew the above seven drugs in one 12 sheng of water till 6 sheng are left. Filter the decoction and stew it till 3 sheng are left. Take 3 doses a day with 1 sheng as a dose.

Subtractions and additions of the decoction

To be restless but not nauseous: Take away Banxia and Renshen, add one piece of Gualoushi (Fructus Trichosanthis).With thirst for water: Take away Banxia, increase Renshen to 4.5 liang, and add 4 liang of Gualoushi.With abdominal pain: Take away Huangqin, and add 3 liang of Shaoyao (Radix Paeoniae).With lump below the costal margin: Take away Dazao, and add 4 liang of Muli (Concha Ostreae).With palpitation and dysuria: Take away Huangqin, and add 4 liang of Fuling (Poria).No thirst of water but with a slight fever: Take away Renshen, and add 3 liang of Guizhi (Ramulus Cinnamomi). Cover the patient with a quilt to obtain a light perspirationAccompanied by coughing: Take away Renshen, Dazao and Shengjiang, and add 0.5 sheng of Wuweizi (Fructus Schisandrae) and 2 liang of Ganjiang (Rhizoma Zingiberis).

The text shows that this formula is a cure for a series of symptoms and signs. Meanwhile, there is also a guide to applying XD in the book.

Febrile disease caused by cold or wind with Xiaochaihu decoction

When one of the symptoms of the syndrome is observed, a diagnosis of XD syndrome can be established.

The text is significant but succinct; therefore, there are a lot of explanations of it since then. The different opinions center on what exactly the indication of applying this formula is.

Before introducing the viewpoints of different scholars, some concepts should be clarified first. In Shanghan Lun, diseases are categorized into six groups according to their seriousness. The six groups are named as Taiyang (Initial Yang) Syndrome, Shaoyang (Lesser Yang) Syndrome, Yangming (Greater Yang) Syndrome, Taiyin (Initial Yin) Syndrome, Shaoyin (Lesser Yin) Syndrome, and Jueyin (Greater Yin) Syndrome. XD is the typical formula of Shaoyang Syndrome. The symptoms and signs of the Shaoyang Syndrome are invariably bitterness in the mouth, a parched throat, and vertigo, which are called the outline of the Shaoyang Syndrome. While the main manifestations of XD are alternative chills and fevers, feeling a distention, and a sensation of oppression in the chest and costal region, reluctant to speak and eat, and restless. In addition, there are also the probable manifestations which are nausea, thirst, abdominal pain, lump below the costal margin, palpitation, dysuria, slight fever, and cough.

Some doctors do not agree with the words “one of the symptoms of the syndrome” of the guideline. They deny that diagnosis can be made with only one symptom. They prefer that the “one” should refer to a series of symptoms. For example, You Zaijing thought the characteristic symptoms are alternative chills and fever and a sensation of distention and oppression in the chest and costal region as the records in his book Shanghan Guanzhuji (A String of Beads from Treatise on Cold-Induced Diseases).[2] In the book Shanghan Bianzheng Guangzhu (Various Dialectical Comments on the Treatise on Cold-Induced Diseases),[3] Wang Hu expounded his view that outline of the Shaoyang Syndrome accompanied with one of main manifestations of XD is the indication to use XD.

If someone can make a diagnosis according to only one symptom, this symptom is the particular one or anyone of the syndrome? Different scholars have different opinions. Cheng Wuji's opinion can be summarized from his book Zhujie Shanghanlun (Treatise on Cold-Induced Diseases with Notes).[4] that when patients just have one of the probable manifestations, XD should be used. Ke Qin expressed his opinions in Shanghan Laisuji (Renewal of Treatise on Cold-Induced Diseases)[5] that the characteristic symptom is one of the outline of the Shaoyang Syndrome.

Most scholars came up with their ideas just through analyzing the traditional Chinese medical theory. Everyone thought that his view was most correct. Under these circumstances, more clinical and statistical evidence are needed. Some researchers have tried to explain from these aspects. Wu [6] selected some cases treated with XD from Encyclopedia of Traditional Chinese Medicine (TCM). Alternative chills and fever had the highest rate of occurrence. Zhang et al.[7] collected 544 medical cases cured by XD from TCM Journals between 1960 and 2008. With frequency statistics, they found that the basic indications of applying XD are the loss of appetite, bitter taste in the mouth, fullness, and pain in the chest and hypochondria, dizziness, irritability and restlessness, nausea, white tongue fur, and wiry pulse. Tao [8] analyzed 763 cases treated with XD recorded in 407 papers between 1994 and 2008 with the method of frequency statistic. It shows that the high-frequency symptoms are like that recorded in Shanghanlun. With association rules, Wang [9] analyzed 380 papers describing the clinical applications of XD from 1970 to 2009. The results show that the confidence percentage with manifestations of wiry pulse, dizziness, nausea, and vomiting as the premise and XD as the conclusion can be 83%.

Their jobs are significant but are still needed to improve. First, all the cases they selected have positive results. Records with negative results are also worth analyzing. Second, one word of symptom in Chinese sometimes contains two or more symptoms. When organizing the data, every symptom should be separated from that kind of words. Third, the tools used to do statistics and data analysis should be introduced. Forth, high frequency does not equal to the high association. Therefore, the data should not be analyzed only by frequency statistics. Finally, when using association rules, the two parameters of support and confidence should be clearly defined. In addition, proper criteria should be used to evaluate the rules.

In this paper, the cases treated with XD with both positive and negative results recorded in medical books before 1949 were collected. Moreover, we tried to mine the characteristic manifestations of XD by operating association rules with RapidMiner.


Sources of dataset

Encyclopedia of Traditional Chinese Medicine is a large electronic series, which contains 1156 ancient medical books before 1949. It is the great achievement of the National Key Project of electronic publishing. Since the purpose is to find the clinical evidence of applying XD, the cases that treated with XD are needed. Therefore, as the keyword, “Xiaochaihu” was used to obtain the relevant records from the Medical Records Part of the encyclopedia, which mainly contains the books from the Ming Dynasty to the Chinese Republic. According to the following inclusion and exclusion criteria, a total of 254 cases treated with XD or its modified formula were included. The medical records were from 70 medical books, and >30% of records were from Xu Mingyi Lei'an (A supplement to the Classified Medical Records of Distinguished Physicians), which was pressed in 1770 and included medical records before Qing Dynasty [Table 1]. The terminologies of manifestations used in this article refer to the WHO International Standard Terminologies on Traditional Medicine in the Western Pacific Region, which has documented the common technical terms used in traditional medicine. Moreover, based on Pharmacopoeia of the People's Republic of China, the ingredients of the selected medicinal formulae were listed. All the 109 symptoms or signs and 122 herbs in the collected cases were organized as the attributes to set up a database.{Table 1}

Inclusion criteria

(1) The cases are cured by XD or its modified formulas; (2) the is detailed describing of manifestations in the case; (3) the modified XD must include Chaihu and at least another two herbs of the original XD; (4) the XD are taken orally in these cases.

Exclusion criteria

Cases with the following criteria were excluded: (1) <50% herbs that belong to XD in the formula; (2) XD being used to consolidate the treatment effect after taking other formulae; (3) being invalid or exacerbation of the disease after taking XD.

When searching the Encyclopedia of Traditional Chinese Medicine, we found that there are some failure cases. The patients' conditions were getting worse after taking XD. These cases are significant and valuable. Because when writing a prescription, doctors believed that some of the manifestations of the patient were suitable for XD. However, XD was no effect or even harmful to these patients. Therefore, we cannot use XD just according to these manifestations. By finding the basis, they misused XD, and comparing with the result of the 254 cases cured by XD, we can get more reliable conclusions. Accordingly, 48 cases that misused XD were collected, and 70 symptoms and signs were organized as the attributes to establish another database.

Inclusion criteria

(1) Being invalid or exacerbation of the disease after taking XD or its modified formulas; (2) There is detailed describing of manifestations in the case; (3) The modified XD must include Chaihu and at least another two herbs of the original XD; (4) The XD are taken orally in these cases.

Exclusion criteria

Cases with the following criteria were excluded: (1) Less than 50% herbs that belong to XD in the formula; (2) XD being used with any other Chinese patent drugs.

Statistical analysis

In this article, Microsoft Excel 2016 was used as a storage tool, and then, the RapidMiner was used as a platform to do the primary statistics and analyze the frequency and the association rules of the manifestations and XD.

Association rules are a data mining methodology that seeks to find frequent connections between attributes in a data set.[10] It is widely used in data analysis for direct marketing, catalog design, and other business decision-making processes. While the amount of clinical data increased rapidly, association rules mining was successfully applied into various kinds of studies both in modern [11],[12] and traditional Chinese medical categories.[13],[14]

There are two main factors that dictate whether or not frequency patterns get translated into association rules: confidence and support.[10] The confidence is the main parameter used when creating rules. This setting tells the software how often an association must be found in a dataset for it to be considered a rule. Association rules are laid out using premises (sometimes called antecedents) and conclusions (sometimes called consequents). The premise is the first attribute in the association, and the conclusions are all attributes associated with the given premise.[15] The support is an easier measure to calculate. This is simply the number of times that the rule did occur, divided by the number of observations in the dataset.

In RapidMiner, the minimum support can be set in the FP-Growth operator, which is the first modeling operator needed for association rules. This operator calculates the frequent itemsets found in the data.[15] The FP in FP-Growth stands for frequency pattern. Frequency pattern analysis is handy for many kinds of data mining and is a necessary component of association rule mining. Without having frequencies of attribute combinations, we cannot determine whether any of the patterns in the data occur often enough to be considered rules. The min-confidence parameter can be set in the Create Association Rule's operator. Only the combinations that satisfy the minimum thresholds on support and confidence were considered to mine meaningful rules (Cross, 2016, North, 2012).[10]

Before setting these two parameters, we will introduce the evaluating parameters first. There are six parameters to evaluate the association rules, lift, Chi-square (χ2), cosine, all-confidence, max-confidence, and Kulczynski (Kulc). Among them, only lift and χ2 are not null-invariant. That means their values will greatly be influenced by the null-transaction. Sometimes, in one dataset, the other four parameters may show totally different result. In this circumstance, imbalance ratio (IR) should be introduced to give a better explanation. Through comparison, Kulc and IR are recommended to be the evaluation index.[16] If Kulc is far more than 0.5 and IR is near to zero, the result is better.



To calculate the two parameters above and to find an effective mode, the min-support of 0.1 and the min-confidence of 0.1 were specified in this study.

In the 254 cases, the association rules were first created between the symptoms and herbs. Moreover, the rules of (symptom ⇒ all 7 herbs of XD) are most attractive. And then, we created the association rules between manifestations to do further analysis. Furthermore, 48 cases which misused XD were organized to create association rules between manifestations. This will explore some manifestations which are similar to that in the 254 cases group. Moreover, these manifestations may mislead the doctors to treat with XD. Moreover, in these 48 cases, most of them used the XD without modifying. That means the failure of treatment was due to the whole formula. Moreover, we do not care about which herbs in XD caused the failure; thus, the association rules between the symptoms and herbs is not necessary in this part. After comparing the manifestations between 254 cases group and 48 cases group, the characteristics of XD will be selected out.


Descriptive statistics of demographic attributes

In the cases that cured by XD, there are 151 males and 103 females. Only 36 cases have age record. Moreover, most of them have the age from 10 to 29. There are 25 males and 23 females in the cases that misused XD. And, only 36 cases have an age record. In other cases, that without age, we can infer that most of them are adults according to the descriptions [Table 2].{Table 2}

Association rules mining of 254 cases cured by Xiaochaihu decoction

Associations between manifestations and herbs

The association rules we also concerned are which with the manifestations as premises and the XD or some herbs of it as the conclusions. And, mining the bases for using XD is the main purpose of this analysis. The association rules “manifestations ⇒ herbs” with the largest values of conf (P⇒C) or conf (C⇒P) were selected from the results. Besides, the association rules of manifestations paired with XD (including all seven herbs) were also selected [Table 3].{Table 3}

Since Chaihu (Radix Bupleuri) was used in every case, it was not displayed in the conclusions. Most confidence values in [Table 3] are >0.8. That means if the manifestations occur, the corresponding herbs will more likely to appear.

Seven manifestations such as inflammatory swelling and pain (ear or breast), headache, fever, hypochondriac pain, alternative chills and fever, rapid pulse, and wiry pulse are associated with XD. Among them, inflammatory swelling and pain (ear or breast) with the support of 0.126 has the highest affinity to XD (0.938). Although alternative chills and fever is the most frequent symptom (0.378), the confidence of that paired with XD is only 0.792.

Moreover, although the data are imbalance, the top nine rules are still attractive. When loss of appetite or inflammatory swelling and pain (ear or breast) occurred, Gancao, Shengjiang, and Dazao must be used. If there is hypochondriac pain, Huangqin must be used. Huangqin and Gancao must appear in the prescriptions to the patients with a headache or thirst. If patients with manifestations of wiry and rapid pulse or abdominal pain, Gancao must be used. When alternative chills and fever are simultaneous with wiry pulse, Huangqin, Shengjiang, and Dazao must be prescribed. Moreover, if vomiting, Shengjiang must be used to stop it.

Referring to Kulc and IR, the association rules of alternative chills and fever paired with XD (Rule 29) or herbs of Huangqin, Banxia, Shengjiang, and Dazao (Rule 23) are better than others.

Association of manifestations

There are only 11 frequent items among the 109 manifestations attributes [Table 4]. Moreover, there are only two association rules after operating by RapidMiner [Table 5].{Table 4}{Table 5}

According to the Kulc, rapid pulse and wiry pulse have a better correlation. Moreover, we can see that the two rules are all related to pules. Since only 104 cases have pulse manifestations records, we selected out these samples to explore the association rules. And, the rest of the cases do not have any association rules between manifestations.

The frequent items of the manifestations in the 104 cases are as follows [Table 6]. The top three items with high-support are wiry pulse (0.625), rapid pulse (0.529), and alternative chills and fever (0.423). Both hypochondriac pain (0.933) and white tongue fur (0.813) have high affinity to wiry pulse [Table 7]. However, referring to the evaluating parameters, the top three pairs with better association include fever paired with rapid pulse (0.7), alternative chills and fever paired with wiry pulse (0.682), and rapid pulse paired with wiry pulse (0.6). These results are similar to that of the 254 cases.{Table 6}{Table 7}

Association rules mining of 48 cases misused Xiaochaihu decoction

The frequent items of 70 manifestation attributes are as follows [Table 8]. Fever and alternative chills and fever are the most frequent ones (0.458). It is obvious that aversion to cold must be accompanied by fever [Table 9]. However, the data are imbalance according to the Kulc and IR. That means if patients have a fever, there is not always accompanied with aversion to cold. Under evaluating, the association rules including a headache paired with fever (Rule 3) and rapid pulse paired with wiry pulse (Rule 5) are more interesting. Moreover, Rule 6 carries no significance because of its low Kulc.{Table 8}{Table 9}

A comparison of the results of 254 cases group and 48 cases group

Some of the frequent items are the same in the two groups [Table 10]. Alternative chills and fever has the biggest support in both groups. Every item is associated with at least one other manifestations or XD except vomiting.{Table 10}

There are also different frequent items in two groups [Table 11].{Table 11}

On comparing [Table 7] and [Table 9], the same association rules in two groups are rapid pulse paired with wiry pulse and rapid pulse paired with fever. Rule 1 is better than Rule 2 in two groups [Table 12].{Table 12}

The same items and rules can be regarded as the indications of using XD by doctors, but the differences are needed more attention. They may be the determinants of using XD or not.


The imbalance of the results

In method part, we introduced the evaluating parameters, Kulc and IR. The results with higher Kulc and lower IR are better. However, most rules have low Kulc and high IR and show an extreme imbalance. Why did that happen? And whether we should deny the results only according to the two parameters or not.

Kulc is the average of conf (P⇒C) and conf (C⇒P). Therefore, the data with higher confidence will have higher Kulc. Since the confidence values are influenced by the support values, the difference between conf (P⇒C) and conf (C⇒P) is obvious sometimes due to the different support values of premises and conclusions. If the support of premise is much higher than that of the conclusion, the conf (P⇒C) will be low and the conf (C⇒P) will be high. Hence, the Kulc will also be low. While the meaning of the conf (P⇒C) is that the occurrence frequency of C when P occurs. In this article, we want to observe what the manifestation is when XD or other specific manifestations have high occurrence frequency. Therefore, sometimes, we just care about conf (P⇒C), the single oriental of the arrow, and the results with low Kulc are not meaningless.

IR is determined by the absolute D-value of sup(P) and sup(C). However, when searching the cases, for example, we just selected the cases treated with XD. Therefore, the herbs belong to XD must have an extremely high rate of occurrence. The support values of them are always >0.8. While the manifestations are diverse, the most frequent symptom is alternative chills and fever, and is only with the support of 0.37. When setting the manifestations as the premises and the herbs as the conclusions, the IR must be high. Under the circumstance, the high IR due to the cases we selected. To this kind of data, IR is just a reference rather than a determinant when evaluating a rule.

The results of association rules

In [Table 3], some confidence values are 1. That means when a symptom or sign occur, some herbs must be used. There are some rules like that, such as loss of appetite/inflammatory swelling and pain (ear or breast) ⇒ Gancao, Shengjiang, Dazao; alternative chills and fever, wiry pulse ⇒ Huangqin, Shengjiang, Dazao; headache/thirst ⇒Huangqin, Gancao; hypochondriac pain ⇒ Huangqin; (wiry and rapid pulse/abdominal pain ⇒ Gancao; vomiting ⇒ Shengjiang. Here, Chiahu has been omitted because it is used in every case.

Gancao (Licorice root) is most frequently prescribed as an important ingredient in many prescriptions of TCM. It is described in Shanghanlun that Gancao (1) harmonizes all drugs and detoxifies the adverse effects of herbs, (2) works synergistically with other drugs, (3) relieves spasm and pain, and (4) has a detoxifying action.[17] The Chinese term Gancao means the “sweet plant.” In TCM theory, sweet flavor acts on the spleen and can enhance the function of the digestive system. At present, >400 compounds have been isolated from licorice and demonstrated a variety of pharmacological activities, such as anticancer, anti-inflammatory, antiviral, hepatoprotective, expectorant, and memory enhancing activity.[18] All these above can explain the using of Gancao aiming to those manifestations.

Shengjiang (Ginger) has been used to treat many medical conditions such as dyspepsia, flatulence, nausea and abdominal pain. It can stimulate gastric emptying and antral contractions in patients with functional dyspepsia.[19] This can explain that patients with the loss of appetite or vomiting are all treated with Shengjiang. Moreover, some researches also proved that ginger volatile oil has potent anti-inflammatory and can treat pain and chronic inflammation.[20] That is why doctors used Shengjiang when patients have inflammatory swelling and pain.

Dazao (the fruits of jujube) is widely used in TCM. It is conventionally used for antifungal, antibacterial, antiulcer, anti-inflammatory, and sedatives. Moreover, it is helpful for chronic constipation and proved to be effective against neonatal jaundice.[21] As the rules showing, Dazao is always used with Shengjiang as a pair to protect the spleen and stomach qi. However, Dazao has a sweet flavor and can sometimes impede the spleen qi. Therefore, patients with nausea or vomiting won't be treated with Dazao, Gancao, or other herbs that have sweet flavor.

In TCM theory, pathogenic fire flaming is characterized by a headache, a sore throat, thirst, redden eyes, and gum bleeding. Huangqin can move heat and dampness, and can also purge fire and remove the toxin from the body. Hence, it is effective to a headache and thirst. Chaihu and Huangqin are the important couplet medicines and are also the main ingredients of XD. If patients have Shaoyang syndrome or XD patterns, XD or at least Chaihu and Huangqin will most probably be used. Alternative chills and fever are always regarded as the characteristic symptom of Shaoyang and XD syndrome in Shanghanlun. There is another article in Shanghanlun that wiry pulse indicates that the Shaoyang is attacked by pathogens. Since the Shaoyang channel goes along the hypochondria and ears, hypochondriac pain and deaf will occur when the Shaoyang channel is attacked. Therefore, aiming at these symptoms, Chaihu and Huangqin must be used.

In [Table 3], there are some manifestations having high affinity to XD. That means when these manifestations occur, the original formula of XD will be most probably used. This is reflected by the confidence value in the association rules. Some researchers we mentioned before thought that the manifestations with high occurrence frequency must be the indications of using XD. However, the figure below reverses this opinion.

[Figure 1] shows that the symptom of alternative chills and fever with the highest support does not have the highest affinity to XD. On the contrary, the symptom of inflammatory swelling and pain with much lower support has the highest confidence. Moreover, if having any manifestations above, at least more than half of the patients will be treated with XD. However whether any of these seven manifestations can be regarded as the indication of using XD or not should still be deliberated.{Figure 1}

By comparing the frequent items and association rules between 254 cases group and 48 cases group, we can see more interesting results. For example, the same association rules in two groups are (rapid pulse ⇒ wiry pulse) and (rapid pulse ⇒ fever). That means doctors should not treat with XD only in accordance with these pairs of manifestations.

Furthermore, [Figure 2] is based on [Table 10] which shows the same frequent items of two groups. Obviously, the occurrence frequency of fever and headache in 48 cases group is much higher than another group. Meanwhile, fever and headache have a high correlation in 48 cases group. Consequently, we would better not use XD when patients with these two symptoms.{Figure 2}

In [Figure 2], alternative chills and fever has the highest support in both groups. And it has high affinity to wiry pulse as well as XD in 254 cases group. While in 48 cases group, alternative chills and fever has the same support to fever, most association rules are related to fever. Hence, the alternative chills and fever can be regarded as one of the characteristic symptoms of XD syndrome.

The wiry pulse and rapid pulse have a similar support value in 254 cases group, but most association rules are related to wiry pulse. Therefore, the wiry pulse seems to be one sign of using XD. However, in [Figure 1], wiry pulse has the lowest affinity to XD. Thus, wiry pulse must accompany other symptoms together to be the characteristics, such as alternative chills and fever, headache, thirst, or hypochondriac pain, according to [Table 7].

Vomiting has a similar and lowest frequency in two groups. Moreover, it also has no association rules with other manifestations or XD under the criterion that min-confidence are 0.1. Although there is an article in Shanghanlun that XD can treat for fever paired with vomiting, we cannot find any evidence from these medical records.

In addition, [Table 11] shows the different frequent items in two groups. Among them, hypochondriac pain and inflammatory swelling and pain (ear or breast) in 254 cases group have high affinity to XD. Thus, these two manifestations can be the characteristics of XD syndrome. In addition, in 48 cases group, delirium, and aversion to cold are associated to fever. This shows that fever is not the characteristic of XD syndrome.


The characteristics of XD syndrome are alternative chills and fever, hypochondriac pain, and inflammatory swelling and pain (ear or breast). When any of these manifestations occur, XD can be used. Wiry pulse paired with alternate chills and fever, headache, thirst, or hypochondriac pain is also can be regarded as the indications of using XD. However, these manifestations indicate that XD can be used, not must be used. XD should not treat for fever paired with some manifestations, such as headache, delirium, or aversion to cold. It shows that XD is more fit for the excess syndrome such as an excess fire in the liver or gallbladder meridiansIn this article, we select some ancient cases, in which the manifestations became worse after using XD. These doctors treated with XD just because some manifestations were similar to the XD syndrome. Analyzing these cases can help us distinguish the real characteristics of XDBy comparing some support values and confidence values, we proved that high occurrence frequency does not equal to the high association. Hence, the researchers cannot get a conclusion only according to the frequency statistic when exploring the principle of using XD or other formulas.

The defect of this article is that the dataset is not large enough. However, the medical cases in ancient time were analyzed with association rules by RapidMiner are tried for the first time. In the future, more data and multiple methods should be used to deeply research the principle of using XD and other TCM formulas.


This article is supported by The Natural Science Foundation of China (Project No. 81373770). Moreover, we also want to express our gratitude to Professor Renquan Liu, who gave a lot of advice about statistics, to Nissi Wang, who made many suggestions about submission, and to Zheng Sun, who offered the Encyclopedia of TCM.

Financial support and sponsorship

This study was financially supported by the Natural Science Foundation of China (Project No. 81373770).

Conflicts of interest

There are no conflicts of interest.


1Li SM, Xu SJ, Li SM, Hong WX. Study on clinical experience of using xiaochaihu decoction. Chinese Journal of Integrative Medicine 2014;34:1264-6.
2You ZJ. A String of Beads from Treatise on Cold-Induced Diseases. Beijing: Academy Press; 2009.
3Wang H. Various Dialectical Comments on the Treatise on Cold-Induced Diseases. Shanghai: Shanghai Medical Publishing House; 1958.
4Cheng WJ. Treatise on Cold-Induced Diseases with Notes. Beijing: The Medicine Science and Technology Press of China; 2011.
5Ke Qin. Renewal of Treatise on Cold-Induced Diseases. Taiyuan: Shanxi Science and Technology Press; 2010.
6Wu GE. Explore of Application of Xiaochaihu Decoction in Past Dynasty. Beijing: Beijing University of Chinese Medicine; 2005. p. 20.
7Zhang XG, Jia B, Chen G, Li CG. Analysis of characteristics of modern TCM doctors applying xiaochaihu decoction. Chin J Inf TCM 2010;193:93-4.
8Tao FZ. Study on Xiaochaihu Catogrized Formulas. Nanjing: Nanjing University of Chinese Medicine; 2009. p. 25.
9Wang GB. Study on Correspondence of Xiaochaihu Decoction and its Syndrome. Nanjing: Nanjing University of Chinese Medicine; 2010. p. 2049.
10Matthew N. Data Mining for the Masses. Athens: Global Text Project; 2012.
11Cremaschi P, Carriero R, Astrologo S, Coli C, Lisa A, Parolo S, et al. An Association rule mining approach to discover lncrnas expression patterns in cancer datasets. Biomed Res Int 2015;12:15.
12Hu W, Li X, Wang T, Zheng S. Association mining of mutated cancer genes in different clinical stages across 11 cancer types. Oncotarget 2016;7:68270-7.
13Chen HY, Lin YH, Hu S, Yang SH, Chen JL, Chen YC. Identifying Chinese herbal medicine network for eczema: Implications from a nationwide prescription database. Evid Based Complement Altern Med 2015;1:19.
14Chen YC, Lin YH, Hu S, Chen HY. Characteristics of traditional Chinese medicine users and prescription analysis for pediatric atopic dermatitis: A population-based study. BMC Complement Altern Med 2016;16:173.
15Hofmann M, Klinkenberg R. RapidMiner: DATA Mining use Cases and Business Analytics Applications. Boca Raton: CRC Press; 2013.
16Han J, Kamber M, Pei J. Data Mining Concepts and Techniques. 3rd ed. San Francisco: Morgan Kaufmann Publisher; 2012.
17Kitagawa I. Licorice root. A natural sweetener and an important ingredient in Chinese medicine. Pure Appl Chemr 2002;74:1189-98.
18Tang ZH, Li T, Tong YG, Chen XJ, Chen XP, Wang YT, et al. A systematic review of the anticancer properties of compounds isolated from licorice (Gancao). Planta Med 2015;81:1670-87.
19Hu ML, Rayner CK, Wu KL, Chuah SK, Tai WC, Chou YP, et al. Effect of ginger on gastric motility and symptoms of functional dyspepsia. World J Gastroenterol 2011;17:105-10.
20Jia YL, Zhao JM, Zhang LH, Sun BS, Bao MJ, Fen-Fen LI, et al. Analgesic and anti-inflammatory effects of ginger oil. Chin Herbal Med 2011;3:150-5.
21Hasan N, Sorkhy MA, Albattah F. Ziziphus jujube of the Middle East, food and medicine. Unique J Ayurvedic Herbal Med 2014;2:7-11.