Children disease risk prediction method and system based on data mining

Abstract

The invention relates to the field of children disease prediction, in particular to a children disease risk prediction method and system based on data mining. The method comprises the following steps: firstly, acquiring disease information and time of different children in each physical examination and index data of different detection indexes, carrying out data division on the detection indexes to obtain the level of each data range of the detection indexes, and according to the level of a target child in the data range where the index data of the detection indexes of two adjacent physical examinations are located, determining the level of the target child in the data range; the method comprises the following steps of: constructing a dynamic triple according to disease information of each child and time difference between two adjacent physical examinations, analyzing change of index data of detection indexes of each child to obtain an optimal degree of the dynamic triple, performing association rule mining on the dynamic triple of each detection index of all children and the disease information, and determining the optimal degree of the dynamic triple. And predicting the child disease risk by using the calculated confidence coefficient of each association rule. According to the invention, the accuracy of children disease risk prediction by the mined association rule can be improved.

Description

Technical Field

[0001] The present invention relates to the field of child disease prediction, and in particular to a method and system for predicting child disease risks based on data mining. Background Art

[0002] Childhood is a critical period of individual development and also a stage with a higher risk of disease. With changes in lifestyle and eating habits, children's health issues are becoming increasingly complex, and the incidence of children's diseases is increasing year by year. Since many children's diseases have no obvious symptoms in the early stages, only through regular examinations and risk predictions can disease problems be discovered and intervened in time. Therefore, it is particularly important to mine and analyze children's information data to identify and manage children's disease risks.

[0003] In related technologies, association rule mining is usually performed on the data information of physical examination indicators and disease information of children's groups to identify the association between children's physical indicator characteristics and disease symptoms, thereby predicting children's disease risks. However, since children are in a stage of rapid growth and development, their physical indicator data are highly dynamic, resulting in the inability of association rules mined by existing methods to accurately predict children's disease risks. Summary of the invention

[0004] In order to solve the technical problem that the association rules mined by the existing methods cannot accurately predict the risk of childhood diseases, the purpose of the present invention is to provide a method and system for predicting the risk of childhood diseases based on data mining. The technical solutions adopted are as follows: The present invention proposes a method for predicting the risk of childhood diseases based on data mining, the method comprising: Obtain disease information, time, and indicator data of different test indicators of different children in each physical examination within a preset time period; Taking any detection indicator as the target detection indicator, dividing the target detection indicator data according to the distribution of the indicator data of the target detection indicator of all children in all physical examinations, and obtaining multiple data ranges of the target detection indicator and the level of each data range; taking any child as the target child, constructing multiple dynamic triples of the target detection indicator of the target child according to the level of the data range of the indicator data of the target detection indicator of the target child in two adjacent physical examinations and the difference in time between the two adjacent physical examinations; According to the changes in the indicator data of the target detection indicators of each child at different physical examinations, the stability coefficient of the target detection indicators is obtained; according to the changes in the indicator data of the target detection indicators of the target child at different physical examinations and the number of physical examinations of the target child, the trust level of the target detection indicators of the target child is obtained; any dynamic triplet of the target detection indicators of the target child is used as the target dynamic triplet, and based on the stability coefficient of the target detection indicators and the trust level of the target detection indicators of the target child, the preferred degree of the target dynamic triplet is obtained; Performing association rule mining on the dynamic triples of each detection index and disease information of all children to obtain multiple association rules, wherein the association rule includes a preceding term and a succeeding term, and the preceding term includes multiple dynamic triples; obtaining the confidence of each association rule according to the preferred degree of each dynamic triple in the preceding term of the same association rule and the preferred degree of each dynamic triple in the preceding term of the association rule with the same preceding term; The disease risk of the child to be predicted is predicted based on the indicator data of each detection indicator of the child to be predicted and the confidence of each association rule.

[0005] Furthermore, the obtaining of multiple data ranges of target detection indicators and the level of each data range includes: Get the standard deviation of the target test indicators of all children in all physical examinations and average ; For target detection indicators, it will be less than As a data range, the range will be greater than or equal to As a data range, the range will be greater than or equal to and less than The range is evenly divided into a preset first number of data ranges, and the range greater than or equal to and less than The range is evenly divided into a preset first number of data ranges, and the range greater than or equal to and less than The range is evenly divided into a preset second number of data ranges, wherein the preset first number is equal to a value of 2, and the preset second number is equal to a value of 6; All the divided data ranges are numbered using non-zero natural numbers in order from left to right to obtain a sequence number for each data range, and the sequence number for each data range is used as the level of each data range.

[0006] Furthermore, the multiple dynamic triplets for constructing the target detection index of the target child include: Any two consecutive physical examinations of the target child were considered as one physical examination group; The absolute value of the difference between the two physical examinations of each physical examination group is taken as the time interval of each physical examination group; The triplet consisting of the data range level of the indicator data of the target detection indicator of the previous physical examination of the target child in each physical examination group, the data range level of the indicator data of the target detection indicator of the next physical examination, and the time interval of each physical examination group is used as the dynamic triplet of the target detection indicator of the target child.

[0007] Furthermore, the stability coefficient of the target detection index is obtained including: According to the order of the physical examination time, the sequence of indicator data of the target detection indicators of the target child in all physical examinations is used as the data sequence of the target detection indicators of the target child; The absolute value of the difference between each indicator data and the next adjacent indicator data in the target child's target detection indicator data sequence is used as the change amount of each indicator data in the target child's target detection indicator data sequence; the average value of the change amounts of all indicator data in the target child's target detection indicator data sequence is used as the overall change amount of the target child's target detection indicator data sequence; Select extreme value indicator data from all indicator data in the data sequence of the target detection indicator of the target child, wherein the extreme value indicator data includes maximum value indicator data and minimum value indicator data, wherein the maximum value indicator data is greater than two adjacent indicator data, and the minimum value indicator data is less than two adjacent indicator data; The overall change amount of the target detection index data sequence of the target child and the amount of the extreme value index data are integrated and used as a natural constant The negative exponential function with the base is used to map the data stability of the target detection index of the target child; The average of the data stability of the target detection index of all children is taken as the overall stability of the target detection index; Take any two children as a child group, input the data sequence of the target detection index of the two children in each child group into the dynamic time warping algorithm, and use the output result of the dynamic time warping algorithm as the data change similarity of the target detection index of each child group, and take the average value of the data change similarity of the target detection index of all child groups as the weight parameter of the target detection index; The product value of the weight parameter and the overall stability is used as the stability coefficient of the target detection index.

[0008] Furthermore, the obtaining of the confidence level of the target detection index of the target child includes: The child corresponding to the maximum value of the data stability of the target detection index is used as a reference child; Inputting the data sequence of the target detection index of the target child and the reference child into a dynamic time warping algorithm, and using the output result as a first trust parameter of the target detection index of the target child; The number of physical examinations of the target child is used as the numerator, the maximum number of physical examinations of all children is used as the denominator, and the ratio is used as the second trust parameter of the target child; The first trust parameter and the second trust parameter are integrated to obtain a trust level of the target detection indicator of the target child.

[0009] Furthermore, the obtaining of the preferred degree of the target dynamic triplet includes: The stability coefficient of the target detection index and the trust level of the target detection index of the target child are integrated and normalized to obtain the preferred degree of the target dynamic triplet.

[0010] Further, obtaining a plurality of association rules includes: The Apriori algorithm is used to mine association rules for the dynamic triples of each detection index and disease information of all children, and multiple association rules are obtained.

[0011] Further, obtaining the confidence of each association rule includes: The average of the preference levels of all dynamic triples in the antecedent of each association rule is taken as the overall preference level of the antecedent of each association rule; Taking any association rule as a target association rule, selecting a plurality of first reference association rules and a plurality of second reference association rules of the target association rule from all other association rules except the target association rule, wherein the first reference association rule and the target association rule have the same antecedent and the same consequent, and the second reference association rule and the target association rule have the same antecedent; The cumulative value of the overall preference of all the predecessors of the first reference association rules is used as the numerator, the cumulative value of the overall preference of all the predecessors of the second reference association rules is used as the denominator, and the ratio is used as the confidence of the target association rule.

[0012] Furthermore, predicting the disease risk of the child to be predicted includes: Using the indicator data of each detection indicator of the child to be predicted, all association rules are matched with the previous items to obtain the matching association rules of the child to be predicted; If the confidence of the matching association rule is greater than a preset threshold, the child to be predicted suffers from the disease information contained in the latter item of the matching association rule.

[0013] The present invention also proposes a child disease risk prediction system based on data mining, the system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor implementing any one of the steps of a child disease risk prediction method based on data mining when executing the computer program.

[0014] The present invention has the following beneficial effects: The present invention takes into account that the association rules mined by the existing methods cannot accurately predict the risk of children's diseases. Therefore, the disease information, time and indicator data of different detection indicators of different children in each physical examination are first obtained. Subsequently, various data and information can be integrated to mine the association rules between the detection indicators and the disease, and the risk of children's diseases can be accurately predicted. Considering that the rapid growth of children and changes in environmental factors will make the indicator data of the detection indicators more dynamic, it is easy to lose the information about the disease represented by the detection indicators in the subsequent process of mining association rules. Therefore, the target detection indicators are first divided into data, the level of each data range of the target detection indicators is obtained, and the dynamic triples of the target detection indicators are constructed. Subsequently, the dynamic triples can be used to accurately mine the association rules, so that the mined information has a stronger credibility and improves the accuracy of disease risk prediction. Considering that for children in the growth and development stage, there may be some detection errors. The dynamics of the indicator detection data are too strong, resulting in unstable or difficult to capture data changes, resulting in poor mining effects on association rules. Therefore, the stability coefficient is first used to reflect the degree of stability of the data changes of the target detection indicator, providing a data basis for the subsequent evaluation of the confidence of the association rules, thereby improving the accuracy of disease risk prediction. At the same time, considering that there are individual differences between different children, the change characteristics of the indicator data of the target detection indicators of different children are different. Therefore, the value of the target detection indicator of the target child in the association rule mining can be reflected by the degree of trust, and then the reliability of the target dynamic triplet in the subsequent mined association rules can be reflected by the degree of preference. Since some detection indicator data of children in the developmental stage are more dynamic, the mined association rules are less authentic in disease prediction. Therefore, the present invention further reflects the authenticity of each association rule in disease prediction through the confidence, thereby achieving accurate prediction of the disease risk of the child to be predicted. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0016] Figure 1 A flow chart of a method for predicting children's disease risks based on data mining provided by an embodiment of the present invention; Figure 2 A schematic diagram of multiple data ranges and levels of each data range of target detection indicators provided by an embodiment of the present invention.

[0017] Reference numerals: Level 1 is the level of the first data range, Level 2 is the level of the second data range, and Level 12 is the level of the twelfth data range. DETAILED DESCRIPTION

[0018] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation method, structure, features and effects of a method and system for predicting the risk of childhood diseases based on data mining proposed by the present invention, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0019] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0020] The following is a detailed description of a method and system for predicting children's disease risks based on data mining provided by the present invention in conjunction with the accompanying drawings.

[0021] See also Figure 1 , which shows a flow chart of a method for predicting children's disease risks based on data mining provided by an embodiment of the present invention, the method comprising: Step S1: Obtain disease information, time, and index data of different detection indicators of different children in each physical examination within a preset time period.

[0022] In order to determine whether children are at risk of disease, they will often go to the hospital for physical examinations. During each physical examination of the child, the hospital will record the indicator data of various test indicators such as the child's height, weight or blood pressure and save them in the background database. At the same time, the hospital will also record the disease information such as diabetes, pneumonia or asthma suffered by the child and save it in the background database. At the same time, the time of each physical examination of the child will be recorded and saved in the background database along with the indicator data and disease information.

[0023] Therefore, the embodiment of the present invention first obtains the disease information, time and indicator data of different detection indicators of different children in each physical examination within a preset time period from the hospital's background database, and then integrates various data and information to explore the association rules between the detection indicators and the disease, and accurately predict the disease risk of children.

[0024] Among them, the preset time period is set to 1 year, and the specific value of the preset time period can also be set by the implementer according to the specific implementation scenario, and is not limited here.

[0025] Step S2: Taking any detection indicator as the target detection indicator, dividing the target detection indicator data according to the distribution of the indicator data of the target detection indicator of all children in all physical examinations, and obtaining multiple data ranges of the target detection indicator and the level of each data range; taking any child as the target child, constructing multiple dynamic triplets of the target detection indicator of the target child according to the level of the data range of the indicator data of the target detection indicator of two adjacent physical examinations of the target child and the difference in time between the two adjacent physical examinations.

[0026] In the process of mining association rules based on the indicator data of children's detection indicators, the indicator data will be converted into discrete attributes, for example, the weight will be converted into multiple different levels such as underweight, normal, overweight, etc., and then these discrete attributes will be used as conditions and the child's disease information will be used as the result to generate corresponding association rules. In practice, the rapid growth of children and changes in environmental factors will make the indicator data of the detection indicators more dynamic, and it is easy to lose the information about the disease represented by the detection indicators in the subsequent association rule mining process. Therefore, the embodiment of the present invention first analyzes any detection indicator, takes any detection indicator as the target detection indicator, and divides the target detection indicator according to the distribution of the indicator data of the target detection indicators of all children in all physical examinations, and obtains multiple data ranges of the target detection indicator and the level of each data range. Subsequently, dynamic triples can be constructed based on the level of the indicator data of the target detection indicators of the children in each physical examination in the corresponding data range, so as to facilitate the subsequent accurate mining of association rules.

[0027] Preferably, in one embodiment of the present invention, the method for acquiring multiple data ranges of target detection indicators and the level of each data range specifically includes: Since the number of child samples analyzed in the embodiment of the present invention is large, the indicator data of the target detection indicators for a large number of children conform to the normal distribution. According to the 3-Sigma principle, about 68% of the data will fall between the mean plus or minus one standard deviation, and about 95% of the data will fall between the mean plus or minus two standard deviations. Therefore, the standard deviation of the indicator data of the target detection indicators of all children in all physical examinations is first obtained. and average .

[0028] Then, for the target detection metric, it will be less than As a data range, the range will be greater than or equal to As a data range, the range will be greater than or equal to and less than The range is evenly divided into a preset first number of data ranges, and the range greater than or equal to and less than The range is evenly divided into a preset first number of data ranges, and the range greater than or equal to and less than The range is evenly divided into a preset second number of data ranges, where the preset first number is set to 2, the preset second number is set to 6, and the specific values ​​of the preset first number and the preset second number can also be set by the implementer according to the specific implementation scenario, and are not limited here.

[0029] Finally, use non-zero natural numbers to number all the divided data ranges from left to right to obtain the sequence number of each data range, and use the sequence number of each data range as the level of each data range, see Figure 2 , which shows a schematic diagram of multiple data ranges of a target detection indicator and the level of each data range provided by an embodiment of the present invention. In an embodiment of the present invention, the target detection indicator is divided into 12 data ranges, and the levels of each data range are 1 to 12 in order from left to right.

[0030] Since there are differences in the changes in the indicator data of the target detection indicators of different children in the rapid growth stage, the embodiment of the present invention first analyzes any child, takes any child as the target child, and constructs multiple dynamic triplets of the target detection indicators of the target child according to the level of the data range of the indicator data of the target detection indicators of the target child in two adjacent physical examinations, and the difference in time between the two adjacent physical examinations. The dynamic triplets can be used to accurately mine association rules subsequently, so that the mined information has stronger credibility, thereby improving the accuracy of disease risk prediction.

[0031] Preferably, in one embodiment of the present invention, the method for acquiring multiple dynamic triplets of target detection indicators of target children specifically includes: In order to make the constructed dynamic triplet reflect the data change characteristics of the target detection indicators of the target child, firstly, any two adjacent physical examinations of the target child are taken as a physical examination group, and the absolute value of the difference between the two physical examinations of each physical examination group is taken as the time interval of each physical examination group. The time interval is used to reflect the time difference between the two adjacent physical examinations of the target child. Subsequently, the time interval can be used as an element to construct the dynamic triplet, so as to improve the accuracy of the subsequent association rule mining between the dynamically changing detection indicators and the disease.

[0032] Then, a triplet consisting of the level of the data range where the indicator data of the target detection indicator of the previous physical examination of the target child in each physical examination group is located, the level of the data range where the indicator data of the target detection indicator of the next physical examination is located, and the time interval of each physical examination group is used as the dynamic triplet of the target detection indicator of the target child, wherein each physical examination group can obtain a corresponding dynamic triplet, and if there are multiple physical examination groups for the target child, multiple dynamic triplets of the target detection indicators of the target child can be obtained. For example, for the first physical examination group of the target child, it includes the first and second physical examinations of the target child, if the level of the data range where the indicator data of the target detection indicator of the first physical examination is located is level 1, the level of the data range where the indicator data of the target detection indicator of the second physical examination is located is level 2, and the time interval between the two physical examinations is 10 days, then the triplet A dynamic triplet as an indicator of target detection for target children.

[0033] By using the same method as above, multiple dynamic triples corresponding to each detection indicator of the target child and multiple dynamic triples corresponding to each detection indicator of each child can be obtained.

[0034] Step S3: According to the changes in the indicator data of the target detection indicators of each child in different physical examinations, the stability coefficient of the target detection indicators is obtained; according to the changes in the indicator data of the target detection indicators of the target child in different physical examinations and the number of physical examinations of the target child, the trust level of the target detection indicators of the target child is obtained; any dynamic triplet of the target detection indicators of the target child is used as the target dynamic triplet, and based on the stability coefficient of the target detection indicators and the trust level of the target detection indicators of the target child, the preferred degree of the target dynamic triplet is obtained.

[0035] Taking into account that for children in the growth and development stage, the detection data of some detection indicators may be too dynamic, resulting in unstable or difficult to capture data changes of some detection indicators, resulting in poor results in subsequent mining of association rules. Therefore, the embodiment of the present invention analyzes the changes in indicator data of target detection indicators of each child in different physical examinations, and reflects the degree of stability of data changes of target detection indicators through the obtained stability coefficient, providing a data basis for subsequent evaluation of the confidence of association rules, thereby improving the accuracy of disease risk prediction.

[0036] Preferably, in one embodiment of the present invention, the method for obtaining the stability coefficient of the target detection index specifically includes: According to the order of the physical examination time, the sequence composed of the indicator data of the target detection indicators of the target child in all physical examinations is used as the data sequence of the target detection indicators of the target child, so that the data sequence of the target detection indicators of each child can be obtained by the same method.

[0037] The absolute value of the difference between each indicator data in the data sequence of the target child's target detection indicator and the next adjacent indicator data is used as the change of each indicator data in the data sequence of the target child's target detection indicator. The larger the change, the greater the change in the indicator data of the target child's target detection indicator between two adjacent physical examinations. Furthermore, the average of the change in all indicator data in the data sequence of the target child's target detection indicator can be used as the overall change of the data sequence of the target child's target detection indicator. The larger the overall change, the greater the fluctuation in the data change of the target child's target detection indicator, the stronger the dynamics, and the more unstable the data change.

[0038] Select extreme value indicator data from all indicator data in the data sequence of the target detection indicator of the target child, the extreme value indicator data includes maximum value indicator data and minimum value indicator data, the maximum value indicator data is greater than two adjacent indicator data, and the minimum value indicator data is less than two adjacent indicator data; The more extreme value index data in the target children's target detection index data sequence, the greater the fluctuation of the target children's target detection index data change, the stronger the dynamics, and the more unstable the data change. Therefore, the overall change amount and the number of extreme value index data in the target children's target detection index data sequence can be combined and used as a natural constant. The data stability of the target detection index of the target children is obtained by mapping the negative exponential function with as the base. The greater the data stability, the more stable and consistent the data changes of the target detection index of the target children are.

[0039] In an embodiment of the present invention, the integration of the two can be achieved by calculating the sum or product of the overall change in the data sequence of the target detection index of the target child and the number of extreme value index data, which is not limited here.

[0040] The data stability of each child's target detection index can be obtained by the same method as above, and then the average value of the data stability of all children's target detection index is taken as the overall stability of the target detection index. The greater the overall stability, the more stable and consistent the data changes of the target detection index are.

[0041] Any two children are regarded as a child group, and the data series of the target detection indicators of the two children in each child group are input into the dynamic time warping algorithm, and the output result of the dynamic time warping algorithm is used as the data change similarity of the target detection indicators of each child group. The greater the data change similarity, the more similar the data change characteristics of the target detection indicators between children, and the higher the confidence level of the data change characteristics of the target detection indicators reflected by the overall stability obtained above. Therefore, the average value of the data change similarity of the target detection indicators of all child groups can be used as the weight parameter of the target detection indicator, and the product value of the weight parameter and the overall stability can be used as the stability coefficient of the target detection indicator. The dynamic time warping algorithm is a technical means well known to those skilled in the art and will not be elaborated here.

[0042] As an example, in one embodiment of the present invention, the expression of the stability coefficient of the target detection index may be specifically, for example, as follows:

[0043] in, Represents the stability coefficient of the target detection index; Indicates Data stability of target detection indicators for each child; Indicates the number of children; Indicates the overall stability of target detection indicators; Indicates The similarity of data changes of target detection indicators for each child group; represents the number of children groups; Represents the weight parameter of the target detection index; Indicates the data stability of the target detection index for the target children; The first in the data sequence representing the target detection index of the target child The change of indicator data; The number of indicator data in the data sequence representing the target detection indicator of the target child, that is, the number of physical examinations of the target child; The overall change in the data series representing the target child's target detection index; The number of extreme value indicator data in the data series representing the target detection indicators of the target children; Indicated by natural constant An exponential function with base .

[0044] Taking into account the individual differences between different children, the change characteristics of the indicator data of the target detection indicators of different children are different, and the different number of physical examinations for different children will lead to different accuracy in capturing the data changes of the target detection indicators. Therefore, the embodiment of the present invention analyzes the changes in the indicator data of the target detection indicators of the target children in different physical examinations and the number of physical examinations of the target children, and reflects the value of the target detection indicators of the target children in association rule mining through the obtained trust level. Subsequently, the trust level of the target detection indicators of the target children and the stability coefficient of the target detection indicators can be combined to accurately calculate the preferred degree of the dynamic triple of the target detection indicators of the target children, thereby improving the accuracy of predicting the risk of children's diseases.

[0045] Preferably, in one embodiment of the present invention, the method for obtaining the trust level of the target detection index of the target child specifically includes: The child corresponding to the maximum value of the data stability of the target detection indicator is used as the reference child; The data series of the target detection indicators of the target child and the reference child are input into the dynamic time warping algorithm, and the output result is used as the first trust parameter of the target child's target detection indicator. The larger the first trust parameter is, the more similar the data change characteristics of the target child's target detection indicator are to those of the reference child, the more stable the data change of the target child's target detection indicator is, and the greater the degree of trust in the target child's target detection indicator.

[0046] The number of physical examinations of the target child is taken as the numerator, the maximum number of physical examinations of all children is taken as the denominator, and the ratio is taken as the second trust parameter of the target child. The larger the second trust parameter is, the more physical examinations the target child has had, which means that it is easier to capture the detailed changes in the indicator data of the target detection index of the target child, and the greater the degree of trust in the target detection index of the target child.

[0047] The first trust parameter and the second trust parameter are then integrated to obtain the trust level of the target detection index of the target child.

[0048] In an embodiment of the present invention, the sum or product of the first trust parameter and the second trust parameter may be used as the trust level of the target detection index of the target child to achieve a combination of the two, which is not limited here.

[0049] As an example, in one embodiment of the present invention, the expression of the confidence level of the target detection index of the target child may be specifically, for example, as follows:

[0050] in, represents the target child’s confidence level in the target detection indicator; The first trust parameter representing the target detection index of the target child; indicates the number of physical examinations of the target children; represents the maximum number of physical examinations for all children; Represents the second trust parameter of the target child.

[0051] Since the dynamic triples of detection indicators of different children and the dynamic triples of different detection indicators of the same child have different reliabilities in the subsequently generated association rules, the embodiment of the present invention first takes any dynamic triple of the target detection indicator of the target child as the target dynamic triplet, and obtains the preference degree of the target dynamic triplet based on the stability coefficient of the target detection indicator and the trust degree of the target detection indicator of the target child. The greater the preference degree, the stronger the reliability of the target dynamic triplet in the subsequently generated association rules, providing a data basis for the calculation of the confidence of each subsequent association rule.

[0052] Preferably, in one embodiment of the present invention, the method for obtaining the preference degree of the target dynamic triplet specifically includes: The greater the stability coefficient of the target detection index and the degree of trust in the target detection index of the target child, the more reliable the dynamic triplet of the target detection index of the target child. Therefore, the stability coefficient of the target detection index and the degree of trust in the target detection index of the target child can be synthesized and normalized, and the calculation result can be limited to range, thereby obtaining the preferred degree of the target dynamic triplet. It should be noted that since there are multiple dynamic triples of the target detection indicators of the target child, the preferred degrees of the multiple dynamic triples of the target detection indicators of the target child are the same.

[0053] In an embodiment of the present invention, the integration of the stability coefficient of the target detection index and the trust level of the target detection index of the target child can be achieved by calculating the sum or product of the two, which is not limited here.

[0054] In one embodiment of the present invention, the normalization method can be maximum and minimum value normalization, and the normalization in subsequent steps can all be performed using maximum and minimum value normalization. In other embodiments of the present invention, other normalization methods can be selected according to the specific range of values, which will not be described in detail.

[0055] As an example, in one embodiment of the present invention, the expression of the preference degree of the target dynamic triplet may be specifically, for example, as follows:

[0056] in, Indicates the preference degree of the target dynamic triple; Represents the stability coefficient of the target detection index; represents the target child’s confidence level in the target detection indicator; Represents the normalization function.

[0057] The same method as above can be used to obtain the preference level of each dynamic triplet of the target detection index of the target child, as well as the preference level of each dynamic triplet of each detection index of each child.

[0058] Step S4: Perform association rule mining on the dynamic triples of each detection indicator and disease information of all children to obtain multiple association rules, where the association rule includes a preceding term and a succeeding term, and the preceding term includes multiple dynamic triples; the confidence of each association rule is obtained according to the preference degree of each dynamic triple in the preceding term of the same association rule, and the preference degree of each dynamic triple in the preceding term of the association rule with the same preceding term.

[0059] The embodiment of the present invention predicts the disease risk of children by mining the association rules between the detection indicators of children and diseases. Therefore, the dynamic triples of the detection indicators of all children and the disease information are first mined to obtain multiple association rules, wherein the association rules include antecedents and consequents, the antecedents include multiple dynamic triples, and the consequents contain various disease information. That is to say, the structural form of the association rules is "antecedent→consequent", which is a representation of a causal relationship, the antecedent represents a condition, and the consequent represents a result associated with the condition. Subsequently, the disease risk of children can be predicted based on the generated association rules.

[0060] Preferably, in one embodiment of the present invention, the Apriori algorithm can be used to perform association rule mining on the dynamic triples of each detection indicator and disease information of all children to obtain multiple association rules, wherein the Apriori algorithm is a technical means well known to those skilled in the art and will not be elaborated here. In other embodiments of the present invention, the existing FP-growth algorithm or Eclatant algorithm may also be used to mine association rules, which is not limited here.

[0061] Since some test indicator data of children in the developmental stage are highly dynamic, the mined association rules have low authenticity in reflecting the correlation between the test indicators and the disease. The preference degree of the dynamic triples obtained above can reflect the reliability in the association rules. Therefore, it is necessary to analyze the confidence of each association rule based on the preference degree of the dynamic triples. The confidence of the association rule is the probability that the association rule containing the predecessor also contains the successor. Therefore, the confidence of each association rule can be obtained according to the preference degree of each dynamic triple in the predecessor of the same association rule and the preference degree of each dynamic triple in the predecessor of the association rule with the same predecessor. Subsequently, the prediction accuracy of children's disease risk can be evaluated based on the confidence of each association rule, thereby achieving accurate prediction of children's disease risk.

[0062] Preferably, in one embodiment of the present invention, the method for obtaining the confidence of each association rule specifically includes: The average value of the preference of all dynamic triples in the antecedent of each association rule is used as the overall preference of the antecedent of each association rule; any association rule is used as the target association rule, and multiple first reference association rules and multiple second reference association rules of the target association rule are selected from all other association rules except the target association rule, wherein the antecedents and the posterior terms of the first reference association rule and the target association rule are the same, and the antecedents of the second reference association rule and the target association rule are the same; and the cumulative value of the overall preference of the antecedents of all the first reference association rules is used as the numerator, the cumulative value of the overall preference of the antecedents of all the second reference association rules is used as the denominator, and the ratio is used as the confidence of the target association rule.

[0063] As an example, in one embodiment of the present invention, the expression of the confidence of the target association rule may be specifically, for example, as follows:

[0064] in, Indicates the confidence of the target association rule; Indicates The overall preference degree of the antecedent of the first reference association rule; The number of first reference association rules representing the target association rule; Indicates The overall preference degree of the antecedent of the second reference association rule; Indicates the number of second reference association rules of the target association rule.

[0065] The confidence of each association rule can be obtained by the same method as above.

[0066] Step S5: predicting the disease risk of the child to be predicted based on the indicator data of each detection indicator of the child to be predicted and the confidence of each association rule.

[0067] After obtaining the confidence of each association rule, the disease risk of the child to be predicted can be accurately predicted based on the indicator data of each detection indicator of the child to be predicted and the confidence of each association rule, thereby improving the accuracy of the prediction of the child's disease risk.

[0068] Preferably, in one embodiment of the present invention, the method for predicting the disease risk of a child to be predicted specifically includes: Using the indicator data of each detection indicator of the child to be predicted, all association rules are matched with the antecedents to obtain the matching association rules of the child to be predicted, wherein the dynamic triples contained in the antecedents of the matching association rules obtained after matching are the same as the dynamic triples corresponding to the detection indicators of the child to be predicted, that is, the disease information contained in the latter term of the matching association rule is more likely to be the disease caused by the child to be predicted.

[0069] At the same time, the greater the confidence of the matching association rule, the more real and reliable the association between the detection index and the disease reflected by the association rule is. Therefore, if the confidence of the matching association rule is greater than the preset threshold, it is considered that the child to be predicted suffers from the disease information contained in the latter item of the matching association rule, where the preset threshold is set to 0.7. The specific value of the preset threshold can also be set by the implementer according to the specific implementation scenario, and is not limited here.

[0070] An embodiment of the present invention provides a child disease risk prediction system based on data mining, the system comprising a memory, a processor and a computer program, wherein the memory is used to store the corresponding computer program, the processor is used to run the corresponding computer program, and when the computer program runs in the processor, it can implement the method described in steps S1 to S4.

[0071] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0072] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A method for predicting children's disease risk based on data mining, characterized in that: The method comprises: Obtain disease information, time, and indicator data of different test indicators of different children in each physical examination within a preset time period; Taking any detection indicator as the target detection indicator, dividing the target detection indicator data according to the distribution of the indicator data of the target detection indicator of all children in all physical examinations, and obtaining multiple data ranges of the target detection indicator and the level of each data range; taking any child as the target child, constructing multiple dynamic triples of the target detection indicator of the target child according to the level of the data range of the indicator data of the target detection indicator of the target child in two adjacent physical examinations and the difference in time between the two adjacent physical examinations; According to the changes in the indicator data of the target detection indicators of each child at different physical examinations, the stability coefficient of the target detection indicators is obtained; according to the changes in the indicator data of the target detection indicators of the target child at different physical examinations and the number of physical examinations of the target child, the trust level of the target detection indicators of the target child is obtained; any dynamic triplet of the target detection indicators of the target child is used as the target dynamic triplet, and based on the stability coefficient of the target detection indicators and the trust level of the target detection indicators of the target child, the preferred degree of the target dynamic triplet is obtained; Performing association rule mining on the dynamic triples of each detection index and disease information of all children to obtain multiple association rules, wherein the association rule includes a preceding term and a succeeding term, and the preceding term includes multiple dynamic triples; obtaining the confidence of each association rule according to the preferred degree of each dynamic triple in the preceding term of the same association rule and the preferred degree of each dynamic triple in the preceding term of the association rule with the same preceding term; The disease risk of the child to be predicted is predicted based on the indicator data of each detection indicator of the child to be predicted and the confidence of each association rule.

2. A method for predicting children's disease risk based on data mining according to claim 1, characterized in that: The obtaining of multiple data ranges of target detection indicators and the level of each data range includes: Get the standard deviation of the target test indicators of all children in all physical examinations and average ; For target detection indicators, it will be less than As a data range, the range will be greater than or equal to As a data range, the range will be greater than or equal to and less than The range is evenly divided into a preset first number of data ranges, and the range greater than or equal to and less than The range is evenly divided into a preset first number of data ranges, and the range greater than or equal to and less than The range is evenly divided into a preset second number of data ranges, wherein the preset first number is equal to a value of 2, and the preset second number is equal to a value of 6; All the divided data ranges are numbered using non-zero natural numbers in order from left to right to obtain a sequence number for each data range, and the sequence number for each data range is used as the level of each data range.

3. The method for predicting children's disease risk based on data mining according to claim 1, characterized in that: The multiple dynamic triplets of target detection indicators for constructing the target child include: Any two consecutive physical examinations of the target child were considered as one physical examination group; The absolute value of the difference between the two physical examinations of each physical examination group is taken as the time interval of each physical examination group; The triplet consisting of the data range level of the indicator data of the target detection indicator of the previous physical examination of the target child in each physical examination group, the data range level of the indicator data of the target detection indicator of the next physical examination, and the time interval of each physical examination group is used as the dynamic triplet of the target detection indicator of the target child.

4. The method for predicting children's disease risk based on data mining according to claim 1, characterized in that: The stability coefficient of the target detection index includes: According to the order of the physical examination time, the sequence of indicator data of the target detection indicators of the target child in all physical examinations is used as the data sequence of the target detection indicators of the target child; The absolute value of the difference between each indicator data and the next adjacent indicator data in the target child's target detection indicator data sequence is used as the change amount of each indicator data in the target child's target detection indicator data sequence; the average value of the change amounts of all indicator data in the target child's target detection indicator data sequence is used as the overall change amount of the target child's target detection indicator data sequence; Select extreme value indicator data from all indicator data in the data sequence of the target detection indicator of the target child, wherein the extreme value indicator data includes maximum value indicator data and minimum value indicator data, wherein the maximum value indicator data is greater than two adjacent indicator data, and the minimum value indicator data is less than two adjacent indicator data; After integrating the overall variation of the target detection index data sequence of the target child and the quantity of the extreme value index data, the data stability of the target child's target detection index is obtained by mapping using a negative exponential function with a natural constant e as the base; The average of the data stability of the target detection index of all children is taken as the overall stability of the target detection index; Take any two children as a child group, input the data sequence of the target detection index of the two children in each child group into the dynamic time warping algorithm, and use the output result of the dynamic time warping algorithm as the data change similarity of the target detection index of each child group, and take the average value of the data change similarity of the target detection index of all child groups as the weight parameter of the target detection index; The product value of the weight parameter and the overall stability is used as the stability coefficient of the target detection index.

5. The method for predicting children's disease risk based on data mining according to claim 4, characterized in that: The confidence level of the target detection index for the target child includes: The child corresponding to the maximum value of the data stability of the target detection index is used as a reference child; Inputting the data sequence of the target detection index of the target child and the reference child into a dynamic time warping algorithm, and using the output result as a first trust parameter of the target detection index of the target child; The number of physical examinations of the target child is used as the numerator, the maximum number of physical examinations of all children is used as the denominator, and the ratio is used as the second trust parameter of the target child; The first trust parameter and the second trust parameter are integrated to obtain a trust level of the target detection indicator of the target child.

6. The method for predicting children's disease risk based on data mining according to claim 1, characterized in that: The obtaining of the preferred degree of the target dynamic triplet comprises: The stability coefficient of the target detection index and the trust level of the target detection index of the target child are integrated and normalized to obtain the preferred degree of the target dynamic triplet.

7. The method for predicting children's disease risk based on data mining according to claim 1, characterized in that: The obtaining of multiple association rules comprises: The Apriori algorithm is used to mine association rules for the dynamic triples of each detection index and disease information of all children, and multiple association rules are obtained.

8. The method for predicting children's disease risk based on data mining according to claim 1, characterized in that: The step of obtaining the confidence of each association rule comprises: The average of the preference levels of all dynamic triples in the antecedent of each association rule is taken as the overall preference level of the antecedent of each association rule; Taking any association rule as a target association rule, selecting a plurality of first reference association rules and a plurality of second reference association rules of the target association rule from all other association rules except the target association rule, wherein the first reference association rule and the target association rule have the same antecedent and the same consequent, and the second reference association rule and the target association rule have the same antecedent; The cumulative value of the overall preference of all the predecessors of the first reference association rules is used as the numerator, the cumulative value of the overall preference of all the predecessors of the second reference association rules is used as the denominator, and the ratio is used as the confidence of the target association rule.

9. The method for predicting children's disease risk based on data mining according to claim 1, characterized in that: The predicting of the disease risk of the child to be predicted includes: Using the indicator data of each detection indicator of the child to be predicted, all association rules are matched with the previous items to obtain the matching association rules of the child to be predicted; If the confidence of the matching association rule is greater than a preset threshold, the child to be predicted suffers from the disease information contained in the latter term of the matching association rule.

10. A child disease risk prediction system based on data mining, the system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 9 are implemented.

Images