A method for constructing a fuzzy model for characteristics of infectious disease patients

Abstract

The application discloses a kind of fuzzy model construction methods for the characteristics of infectious disease patient, comprising: 1 by collecting the medical record of patient, examination report, medical history, lifestyle information, construct a comprehensive patient feature set;2, the representative, uniqueness and anti-noise score of each patient are calculated after selecting the patient representative with representation ability;3, around high total score representation, a fuzzy interval two type granularity model containing fuzzy feature description of multiple patients is constructed.The application can quickly construct a patient group model with better anti-noise ability, stronger interpretability and more information content under the condition of only a small amount of disease-related data, so as to quickly model susceptible population, describe the fuzzy characteristics affecting disease infection, and assist doctors in understanding the characteristics of infectious diseases and susceptible population characteristics.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of data mining, and particularly relates to an infectious population feature modeling algorithm for patient data. BACKGROUND

[0002] With the rapid development of big data technology, using data mining technology to assist doctors in understanding disease characteristics and designing personalized diagnosis and treatment plans has become a popular innovative field. Medical patient data refers to data related to disease prevention and emergency response collected in the process of disease prevention and health management. However, big data is usually large in data volume and contains unavoidable errors in the statistical process, and it is difficult for the human brain to extract valuable information from it alone.

[0003] Clustering technology is one of the most important tasks in unsupervised learning, and the purpose is to divide data samples into different clusters, so that the data points in the same group are more similar than the data points in different groups. The clustering algorithm usually generates multiple clusters and the data prototype of the cluster. This feature makes it usually used for pre-grouping of data, and the data prototype represents the information of the cluster. However, the clusters generated by the traditional clustering technology have the following problems, which cannot provide more valuable information to doctors.

[0004] 1) Contains less information. The clusters formed by the existing clustering algorithm contain too little information, and the description of the cluster only has one patient feature information representing the cluster, and there is no feature range, feature membership degree, etc.

[0005] 2) Poor interpretability. The clusters formed by the existing clustering algorithm are either completely irregular shapes or ordinary circles, and these shapes cannot effectively represent the valuable feature range, making it difficult for users to understand the results.

[0006] 3) Poor noise resistance. The algorithm does not consider excluding noise in the data set, so that the final formed cluster model is disturbed by noise data when the data volume is large. SUMMARY

[0007] In order to solve the above-mentioned deficiencies of the current technology, the present application proposes a feature fuzzy model construction method for infectious disease patients, so as to quickly construct a disease patient group model with better noise resistance, stronger interpretability and more information content under the condition of only a small amount of disease-related data, thereby quickly modeling the susceptible population, describing the fuzzy features affecting disease infection, and assisting doctors in understanding the characteristics of infectious diseases and the features of susceptible population.

[0008] In order to achieve the above-mentioned application purposes, the present application adopts the following technical solutions:

[0009] The present invention provides a method for constructing a feature fuzzy model for patients with infectious diseases, characterized by the following steps:

[0010] Step 1: Collect various characteristics of patients with infectious diseases and construct a dataset of patients with infectious diseases. , Indicates the first A patient with an infectious disease The j-th feature in The total number of patients with infectious diseases. The total number of characteristics for each infectious disease patient;

[0011] Step 2: Select representatives of patients with infectious diseases;

[0012] Step 2.1: Use the DBSCAN algorithm to... Divided into A cluster of infectious disease patients And count the number of infectious disease patients in each cluster of infectious disease patients. and the number of patients with noise-induced infectious diseases ,in, Indicates the first A cluster of patients with infectious diseases, express The number of patients with infectious diseases in China ,lie in Any q-th infectious disease patient in the list is denoted as , ;

[0013] Step 2.2: Calculate the number of representatives of infectious disease patients. ,in, The selected ratio is used to calculate the first ratio according to equation (3). A cluster of infectious disease patients The number of infectious disease patients represented ;

[0014] (3)

[0015] Step 2.3, Calculation After calculating the Euclidean distance between any two patients with infectious diseases and sorting them in descending order, select the top... The mean of the Euclidean distances is used as density radius Thus obtain density radius ;

[0016] Step 2.4: Calculate using equation (5) The qth infectious disease patient in China Representative score ;

[0017] (5)

[0018] In formula (5), represents a positive activation function function; represents the Euclidean distance between the qth infectious disease patient in and the hth infectious disease patient in ;

[0019] Step 2.5, calculate the uniqueness score of the qth infectious disease patient in according to formula (6) ; After normalization, the normalized patient uniqueness score is obtained ;

[0020] (6)

[0021] In formula (6), represents the representative score of the hth infectious disease patient in ;

[0022] Step 2.6, calculate the non-noise score of the qth infectious disease patient in according to formula (8) :

[0023] (8)

[0024] In formula (8), represents the hth feature of the qth infectious disease patient in ; represents the minimum value of the hth feature in ; represents the maximum value of the hth feature in ;

[0025] Step 2.7, calculate the final candidate score of the qth infectious disease patient in according to formula (9) :

[0026] (9)

[0027] Step 2.8, for ​​​​​​​​​Ranking the final candidate scores of all infectious disease patients in descending order, and taking the top infectious disease patients as representatives of infectious disease patients , thereby obtaining clusters of infectious disease patients representatives of infectious disease patients , wherein represents the s-th representative of infectious disease patients .

[0028] Step 3, based on the infectious disease patient representatives selected in step 2, a feature fuzzy model of the infectious disease patient population is constructed;

[0029] Step 3.1, the left end point fitness value of the feature fuzzy model is constructed, and the genetic algorithm is used to process , to obtain the optimal left end point of the feature fuzzy model of the infectious disease patient population , wherein represents the optimal left end point of the s-th feature fuzzy model constructed around on the j-th feature

[0030] Step 3.2, the right end point fitness value of the feature fuzzy model is constructed, and the genetic algorithm is used to process , to obtain the optimal right end point of the feature fuzzy model of the infectious disease patient population , wherein represents the optimal right end point of the s-th feature fuzzy model constructed around on the j-th feature

[0031] Step 3.3, the re-fuzzy fitness value of the feature fuzzy model is constructed, and the genetic algorithm is used to process and , to obtain the feature fuzzy model of the final H infectious disease patient population , and , represents the left end point of the feature fuzzy model with aggressive estimation, represents the left end point of the feature fuzzy model with conservative estimation, represents the right end point of the feature fuzzy model with conservative estimation, represents the right end point of the feature fuzzy model with aggressive estimation.

[0032] The feature fuzzy model construction method for infectious disease patients according to the present application is also characterized in that in step 3.1, the left end point fitness value is constructed by formula (10):

[0033] (10)

[0034] In formula (10), the left end point of the s-th feature fuzzy model constructed around on the j-th feature, and as an individual in the genetic population; denotes a function for calculating the number of infectious disease patients, denotes the j-th feature of the n-th infectious disease patient in ; denotes the number of features satisfying the inequality; denotes a function for calculating the accuracy of the feature fuzzy model;

[0035] Further, the step 3.2 is to construct the right end point fitness value by using formula (11):

[0036] (11)

[0037] In formula (11), the right end point of the s-th feature fuzzy model constructed around on the j-th feature, and as an individual in the genetic population.

[0038] Further, the step 3.3 is to construct the re-fuzzy fitness value by using formula (12):

[0039] (12)

[0040] In formula (12), the parameter set of the s-th feature fuzzy model constructed around on the j-th feature, and as an individual in the genetic population, each individual corresponding parameter set contains: the left end point of the s-th feature fuzzy model constructed around on the j-th feature with a radical estimate ; the left end point of the s-th feature fuzzy model constructed around on the j-th feature with a conservative estimate ; the right end point of the s-th feature fuzzy model constructed around on the j-th feature with a conservative estimate ; the right end point of the s-th feature fuzzy model constructed around on the j-th feature with a radical estimate ; ​​​represents the optimal individual, and s.t. represents the constraint relationship.

[0041] The electronic device comprises a memory and a processor, and the memory is used for storing a program supporting the processor to execute the feature fuzzy model construction method, and the processor is configured to execute the program stored in the memory.

[0042] The computer readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the feature fuzzy model construction method are executed.

[0043] Compared with the prior art, the beneficial effects of the present application are embodied in:

[0044] 1. The present application uses a two-type fuzzy set as a model framework for a patient population, which has stronger interpretability than the clusters generated by a clustering algorithm, and can convey more patient information than a one-type fuzzy granularity data, and the final fuzzy description of the characteristics of the patient population makes the infection population model built using the method more understandable.

[0045] 2. The present application introduces the data mining idea of density estimation, and by comprehensively considering the representativeness, uniqueness and noise evaluation of patients, the selected patient representatives have stronger representativeness, greater difference, and are less affected by noise than the data prototypes generated by a clustering algorithm.

[0046] 3. The present application fully considers the amount of information contained in the infection population characteristic model and the accuracy of the description, and introduces a genetic algorithm to optimize the two indicators simultaneously. Focusing on the patient itself, the final model has high coverage and strong specificity. A good balance is achieved between the interpretability and accuracy of the patient population. And by problem splitting, the construction process of the feature fuzzy model is divided into three sub-steps, which effectively reduces the construction time of the infection population characteristic model. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 A flowchart of the present application for constructing a medical patient population model;

[0048] Figure 2 A schematic diagram of the fuzzy two-type granularity data modeling process used in the present application;

[0049] Figure 3 A schematic diagram of the fuzzy two-type granularity data on the jth feature. DETAILED DESCRIPTION

[0050] In this embodiment, an infection population modeling method based on a type-two fuzzy granularity model is used to model medical patient information. First, it collects patients' medical-related information to construct a comprehensive set of patient features. Next, it calculates the representativeness, uniqueness, and noise resistance scores for each patient and selects patients with high total scores as representatives. Finally, it constructs a fuzzy interval type-two granularity model around these patient representatives, incorporating fuzzy feature descriptions of various patients. See also... Figure 1 The modeling method is performed in the following steps:

[0051] Step 1: Collect characteristic information such as medical records, examination reports, medical history, and lifestyle habits of patients with infectious diseases, and construct a dataset of patients with infectious diseases. , Indicates the first A patient with an infectious disease The j-th feature in The total number of patients with infectious diseases. The total number of characteristics for each infectious disease patient;

[0052] Step 2: Select representatives of patients with infectious diseases; such as Figure 2 As shown in the first step, representative patients are selected from the patient-feature dataset.

[0053] Step 2.1: Use the DBSCAN algorithm to... Divided into A cluster of infectious disease patients And count the number of infectious disease patients in each cluster of infectious disease patients. and the number of patients with noise-induced infectious diseases ,in, Indicates the first A cluster of patients with infectious diseases, express The number of patients with infectious diseases in China ,lie in Any q-th infectious disease patient in the list is denoted as , ;

[0054] Step 2.2: Calculate the number of representatives of infectious disease patients. ,in, The selected ratio is p, which is 0.25 in this embodiment. The value of the selected ratio can be adjusted according to specific needs; thus, the first step is calculated according to formula (3). A cluster of infectious disease patients The number of infectious disease patients represented ;like Figure 2 In this context, the number of patient representatives H is 3, which also indicates that three patient models will ultimately be formed.

[0055] (3)

[0056] Step 2.3, Calculation After calculating the Euclidean distance between any two patients with infectious diseases and sorting them in descending order, select the top... The mean of the Euclidean distances is used as density radius Thus obtain density radius ;

[0057] Step 2.4: Calculate using equation (5) The qth infectious disease patient in China Representative score Representativeness is represented by the number of the patient's nearest neighbor data; the larger the value, the larger the number of nearest neighbors and the better the representativeness.

[0058] (5)

[0059] In equation (5), This represents the activation function for positive numbers; express The qth infectious disease patient With the hth infectious disease patient The Euclidean distance between them;

[0060] Step 2.5: Calculate according to formula (6) The qth infectious disease patient in China Uniqueness score After normalization, the normalized patient uniqueness score is obtained. Uniqueness is measured by its distance from other highly representative data; the greater the distance, the higher the uniqueness.

[0061] (6)

[0062] In equation (6), express The hth infectious disease patient in China Representative score;

[0063] Step 2.6: Calculate according to formula (8) The qth infectious disease patient in China Non-noise score :

[0064] (8)

[0065] In equation (8), express The Middle A patient with an infectious disease The One characteristic, express The Middle The minimum value of each feature. express The Middle The maximum value of each feature;

[0066] Step 2.7: Calculate according to equation (9) The Middle A patient with an infectious disease Final candidate scores :

[0067] (9)

[0068] Step 3: Based on the infectious disease patient representatives selected in Step 2, construct a fuzzy model of the infectious disease patient population; such as... Figure 2 The second step involves constructing a fuzzy model of the patient representative.

[0069] Step 3.1: Construct the fitness values ​​of the left endpoint of the feature fuzzy model, and use the left endpoint genetic algorithm to... The optimal left endpoint of the fuzzy model of the characteristics of patients with infectious diseases is obtained through processing. ,in, express China The optimal left endpoint of the constructed fuzzy model for the s-th feature is on the j-th feature; the optimization objective is composed of two indicators: the model covers a large number of patients and the model accurately represents the range of features.

[0070] (10)

[0071] In equation (10), express China The left endpoint of the constructed fuzzy model for the s-th feature on the j-th feature is taken as an individual in the population; A function representing the number of patients with infectious diseases. Indicates that it is located at the th The j-th characteristic of the nth infectious disease patient in a cluster of infectious disease patients This indicates the number of features that satisfy the inequality; A function representing the calculation of the accuracy of a feature-based fuzzy model; such as... Figure 3 As shown, the genetic algorithm will on the j-th feature Expand to the left .

[0072] Step 3.2, constructing the right end point fitness value of the feature fuzzy model, using the right end point genetic algorithm to process , to obtain the optimal right end point of the feature fuzzy model of the infectious disease patient , wherein, represents the optimal right end point of the s-th feature fuzzy model constructed around the j-th feature ;

[0073] (11)

[0074] In formula (11), represents the right end point of the s-th feature fuzzy model constructed around the j-th feature in ; as shown in formula (11), the algorithm extends Figure 3 to the right on the j-th feature , and and together constitute the initial interval data. The method of separately extending the left and right intervals fully utilizes the independence of the interval data, and effectively improves the convergence speed of the genetic algorithm.

[0075] Step 3.3, constructing the re-fuzzification fitness value of the feature fuzzy model, using the re-fuzzification genetic algorithm to process and , to obtain the feature fuzzy model of the final H infectious disease patients , and , represents the left end point of the model with aggressive estimation, represents the left end point of the model with conservative estimation, represents the right end point of the model with conservative estimation, represents the right end point of the model with aggressive estimation. The genetic algorithm individual fitness formula is (12).

[0076] (12)

[0077] In formula (12), represents the parameter set of the s-th feature fuzzy model constructed around the j-th feature in , and serves as an individual in the population. The parameter set corresponding to each individual contains: the left end point of the s-th feature fuzzy model constructed around the j-th feature in with aggressive estimation ; the left end point of the s-th feature fuzzy model constructed around the j-th feature in the left end point of the conservative estimation of the s-th constructed fuzzy feature model on the j-th feature ; around the right end point of the conservative estimation of the s-th constructed fuzzy feature model on the j-th feature ; around the right end point of the aggressive estimation of the s-th constructed fuzzy feature model on the j-th feature ; denotes the optimal individual, s.t. denotes the constraint relationship. As shown in Figure 3 , the algorithm extends into and , extends into and , re-fuzzifies the two interval endpoints, promotes the one-type fuzzy data to two-type fuzzy data, and contains more patient feature information.

[0078] As a generative model, the fuzzy model provides a fuzzy feature description of the infected population, unlike discriminative models which only have a single use. The model can provide multiple helps for doctors to diagnose and understand the characteristics of the disease.

[0079] For example, now that we have a fuzzy model of a certain infected population, where the j-th feature represents the body temperature feature of the population. If we want to judge the fever feature of the population with a confidence level of 60%, the model will output: in the case of conservative estimation, when the body temperature is in this range, the probability of the patient being ill is more than 60%. When aggressive estimation is required, the body temperature interval in which the probability of the patient being ill is greater than 60% is .

[0080] For example, the model can output a judgment of the probability of a suspected patient being ill. Now there is a new suspected patient with a body temperature of T. First, find the infected population model to which it belongs through the nearest principle . Through its body temperature feature, we can judge that when , the probability of the patient being ill is in this interval, and when , the probability of the patient being ill is in this interval. In this embodiment, an electronic device includes a memory and a processor, the memory is used to store a program supporting the processor to execute the above method, and the processor is configured to execute the program stored in the memory.

[0081] In this embodiment, an electronic device includes a memory and a processor, the memory is used to store a program supporting the processor to execute the above method, and the processor is configured to execute the program stored in the memory.

[0082] ​In the embodiment, a computer readable storage medium stores a computer program, and the computer program is run by a processor to execute the steps of the method.

[0083] In summary, the method not only classifies patients into several groups, but also provides a fuzzy feature description of the patient groups, which significantly improves the information content and interpretability of the model, benefits doctors in understanding the characteristics of infectious diseases and susceptible population characteristics, and provides auxiliary information for the diagnosis of future suspected patients. Meanwhile, the method fully considers the statistical errors and noise problems inherent in big data, can effectively eliminate the influence of these data on the accuracy of the final model, and further improves the robustness of the infectious population feature fuzzy model.

Claims

1. A method for constructing a feature vague model for an infectious disease patient, characterized in that, The method comprises the following steps: Step 1, collecting various features of infectious disease patients and constructing an infectious disease patient dataset , represents the jth feature in the ith infectious disease patient , is the total number of infectious disease patients, is the total number of features of each infectious disease patient;​ Step 2, selecting infectious disease patient representatives; Step 2.1: Use the DBSCAN algorithm to... Divided into A cluster of infectious disease patients And count the number of infectious disease patients in each cluster of infectious disease patients. and the number of patients with noise-induced infectious diseases ,in, Indicates the first A cluster of patients with infectious diseases, express The number of patients with infectious diseases in China ,lie in Any q-th infectious disease patient in the list is denoted as , ; Step 2.2, calculating the number of infectious disease patient representatives wherein, is the selected ratio, so that the number of infectious disease patient representatives in the infectious disease patient cluster is calculated according to equation (3) ; and Step 2.3, calculating the number of infectious disease patient representatives (3) Step 2.3, Calculation After calculating the Euclidean distance between any two patients with infectious diseases and sorting them in descending order, select the top... The mean of the Euclidean distances is used as density radius Thus obtain density radius ; Step 2.4, calculating with formula (5) qth infectious disease patient representative score ; (5) In formula (5), represents a positive activation function function; represents the qth infectious disease patient in the hth infectious disease patient between the hth infectious disease patient Step 2.

5. Calculate according to formula (6) the qth infectious disease patient a unique score After normalization, the normalized patient unique score is obtained ; (6) In formula (6), denotes the hth infectious disease patient a representative score; Step 2.

6. Calculate according to formula (8) qth infectious disease patient non-noise score : (8) In formula (8), denotes the first of the infectious disease patients the first characteristic, denotes the minimum value of the first characteristic, denotes the maximum value of the first characteristic; Step 2.

7. Calculate according to formula (9) the final candidate score for the infectious disease patient :​ (9) Step 2.8, on all infectious disease patients are ranked in descending order of final candidate score, and the top infectious disease patients are taken as representatives of infectious disease patients , thereby obtaining clusters of infectious disease patients representatives of infectious disease patients , wherein represents the s-th representative of infectious disease patients ; Step 3, constructing a feature fuzzy model of the infectious disease patient population based on the infectious disease patient representatives selected in step 2; Step 3.1, constructing fitness value of left end point of feature fuzzy model, and using genetic algorithm to process , to obtain optimal left end point of feature fuzzy model of infectious disease patient population , wherein, represents , the optimal left end point of the s-th feature fuzzy model constructed around on the j-th feature Step 3.2, constructing the fitness value of the right end point of the feature fuzzy model, and using genetic algorithm to process , to obtain the optimal right end point of the feature fuzzy model of the patient population of infectious diseases , wherein, represents the optimal right end point of the s-th feature fuzzy model constructed around the j-th feature . Step 3.3, re-fuzzing fitness value of the feature fuzzy model is constructed, and genetic algorithm is used to process and to obtain the final H feature fuzzy model of the patient population of infectious diseases , and , represents the left end point of the feature fuzzy model with aggressive estimation, represents the left end point of the feature fuzzy model with conservative estimation, represents the right end point of the feature fuzzy model with conservative estimation, represents the right end point of the feature fuzzy model with aggressive estimation.

2. The method according to claim 1, wherein the method is characterized by: The left end point fitness value in step 3.1 is constructed by using formula (10): (10) In formula (10), denotes the s-th feature fuzzy model constructed around the left end point of the j-th feature, and as an individual in the genetic population; denotes a function for calculating the number of infectious disease patients, denotes the j-th feature of the n-th infectious disease patient in ; denotes the number of features satisfying the inequality; denotes a function for calculating the accuracy of the feature fuzzy model.

3. The method according to claim 2, wherein the method is characterized by: The right end point fitness value in step 3.2 is constructed by using formula (11): (11) In formula (11), denotes around the the right end point of the s-th feature fuzzy model constructed on the j-th feature, and as an individual in the genetic population.

4. The method according to claim 2, wherein the method is characterized by: The re-fuzzy fitness value in step 3.3 is constructed by using formula (12): (12) In formula (12), denotes around the parameter set of the s-th feature fuzzy model constructed on the j-th feature, and as a individual in the genetic population, the parameter set corresponding to each individual contains: around the left end point of the s-th feature fuzzy model constructed on the j-th feature with aggressive estimation ; around the left end point of the s-th feature fuzzy model constructed on the j-th feature with conservative estimation ; around the right end point of the s-th feature fuzzy model constructed on the j-th feature with conservative estimation ; around the right end point of the s-th feature fuzzy model constructed on the j-th feature with aggressive estimation ; denotes the optimal individual, and s.t. denotes the constraint relationship.

5. An electronic device comprising a memory and a processor, characterized in that The memory is used to store a program supporting the processor to execute the feature fuzzy model construction method in any one of claims 1-4, and the processor is configured to execute the program stored in the memory.

6. A computer-readable storage medium having stored thereon a computer program, characterized in that The computer program is executed by the processor to perform the steps of the feature fuzzy model construction method in any one of claims 1-4.

Images