Credit medical limit evaluation method based on decision boosting tree

Abstract

The invention relates to the technical field of medical treatment, and discloses a credit medical limit evaluation method based on a decision boosting tree, and the method can achieve the adjustment of an initial limit according to the disease type of a patient, can achieve the more precise matching of the medical demands of the patient, and avoids a one-time limit setting mode. For patients with relatively high rehabilitation cost, the credit medical limit evaluation system can provide certain economic support and reduce the economic burden of the patients; according to the method, the overdue condition of the patient in hospital and the overdue condition of the patient in non-hospital are scored differently, so that the credit condition of the patient can be reflected more intuitively, and the influence of the overdue condition and the like caused by the patient in hospital on the credit line is small; according to the mode, a forward excitation signal can be transmitted, but overdue behaviors under special conditions are also understood and contained, and the inclusiveness is beneficial to enhancing the credit awareness of the patient and promoting the patient to maintain own credit records more positively in the future.

Description

Technical Field

[0001] The present invention relates to the technical field of medical credit, and in particular to a method for evaluating medical credit quota based on a decision boosting tree. Background Art

[0002] Credit medical care is a medical model in which patients are assessed for their credit and granted a certain credit limit, allowing patients to receive treatment first and pay later. Within this credit limit, the medical expenses incurred by seeing a doctor can be deducted in advance using the credit limit, and the patient completes the payment in one lump sum within a certain period of time after the end of the treatment.

[0003] The use of credit medical care can mainly avoid patients having to go through the payment process before proceeding with the subsequent diagnosis and treatment process during a single visit, reduce payment links and waiting times, and alleviate patients' short-term financial payment pressure.

[0004] The existing credit medical model usually uses a single reference indicator to assess the patient's credit, and only evaluates the credit limit based on the patient's social resume and economic situation. This evaluation method cannot grant patients credit limits based on their actual needs when seeking medical treatment, which will lead to uneven supply and demand. Especially for some patients who need long-term rehabilitation cases, when the granted credit limit is used up, they still need to queue up to pay by themselves, which will result in patients being unable to enjoy the convenience brought by credit medical care. Therefore, a credit medical credit limit evaluation method based on a decision boosting tree is proposed to solve the above problem. Summary of the invention

[0005] Technical issues solved

[0006] In view of the shortcomings of the existing technology, the present invention provides a credit medical credit limit evaluation method based on a decision boosting tree, which has the advantages of reducing the payment process and waiting time of patients during the medical treatment process, alleviating the short-term financial payment pressure of patients, and solving the problem that patients must go through the payment process before proceeding with the subsequent medical treatment process during a medical treatment.

[0007] Technical Solution

[0008] To achieve the above object, the present invention provides the following technical solution: a method for evaluating medical credit limit based on a decision boosting tree, comprising the following steps:

[0009] S1, establish a database DB for storing patient information, the patient registers for an outpatient clinic and submits personal information to the database DB, and matches the existing patient data in the database DB with the submitted patient personal information to determine whether the patient has a credit limit, if so, proceed to S3, otherwise proceed to S2;

[0010] S2, construct an initial credit evaluation system based on the patient's social resume evaluation factor set A1 to grant the patient an initial credit limit E1 and enter S3;

[0011] S3, the patient visits the doctor, the doctor diagnoses the patient's condition, and divides the patient into short-term recovery case B1 and long-term recovery case B2 according to the recovery time. If the patient is a short-term recovery case B1, the patient proceeds to S5, and if the patient is a long-term recovery case B2, the patient proceeds to S4;

[0012] S4, constructing a credit limit adjustment evaluation system based on the patient's long-term rehabilitation case B2, and adjusting the patient's credit limit E1 to a credit limit E4 through the credit limit adjustment evaluation system;

[0013] S5, the patient receives treatment according to the doctor's diagnosis, and the treatment expenses are paid by the credit line E4, generating the used credit line D1, and obtaining the patient's treatment history;

[0014] S6, settlement is carried out after the treatment is completed. During the settlement process, the patient's medical insurance reimbursement amount Y b To achieve quota settlement and obtain the settlement amount D2;

[0015] S7. Establish a machine learning model based on the patient's medical records and the actual settlement amount D2 to update the one-time amount adjustment system;

[0016] S8. The patient repays the settlement amount D2. Based on the patient's subsequent repayment record, A n A secondary credit limit adjustment evaluation system is established to obtain the credit limit E4, the patient's credit limit information is stored in the database DB, and the original patient information is updated.

[0017] Furthermore, the social resume evaluation factor set A1 in S2 includes x evaluation factors, each evaluation factor x has h rating levels, each evaluation factor x is assigned a weight w, and each evaluation level h is assigned a score j by the formula:

[0018]

[0019] To obtain the patient's social resume score F, the social resume score F value interval F∈[F min , F max ], divided into four sub-intervals from small to large:

[0020]

[0021] Among them, F1, F2, F3, F4 correspond to σ1, σ2, σ3, σ4 respectively, among which σ1, σ2, σ3, σ4 correspond to different amounts, and σ1<σ2<σ3<σ4, according to the formula:

[0022]

[0023] According to E1=N(F), the credit limit E1 is obtained.

[0024] Furthermore, the long-term rehabilitation cases B2 in S4 include different diseases m that require long-term rehabilitation, each disease m that requires long-term rehabilitation is divided into different severity levels y, different severity levels y correspond to different rehabilitation times k, each different disease m corresponds to a different initial quota E2, each different disease m corresponds to a different initial weight W1, and the mapping function F(m, y, k) is defined according to the above adjustment factors:

[0025]

[0026] In order to avoid the value of F(m, y, k) being too large or too small, the output value of the mapping function F(m, y, k) is truncated so that its output value is between the interval [η1, η2], through the formula:

[0027]

[0028] Where F(m, y, k) m in is the minimum output value of the mapping function, F(m, y, k) max is the maximum output value of the mapping function. According to the mapping function F new (m, y, k) to adjust the weight W1 and get the adjusted weight W2:

[0029] W2=W1×F(m,y,k) new

[0030] The adjusted credit limit E3 is obtained based on the adjusted weight W2 and the initial credit limit E2 of each m:

[0031]

[0032] To obtain the adjustment amount E3 and adjust the primary credit amount E1, to obtain the secondary credit amount E4:

[0033] E4=E1+E3

[0034] where α m , β m , γ m is the coefficient, k min，m is the shortest recovery time for patients with m diseases, k max，m The longest recovery time for patients suffering from disease m.

[0035] Furthermore, the settlement amount D2 in S6 is calculated according to the formula:

[0036] D2=D1-Y b

[0037] D2 is the settlement amount, Y b The amount of medical insurance reimbursement. When the patient does not use medical insurance, Y b =0.

[0038] Furthermore, D2 is the actual settlement amount required when the patient's disease type is m and its severity level is y. The gradient boosting tree model is established under the above conditions to estimate the patient's treatment costs and obtain the estimated cost D'2. By comparing the difference Δ between the estimated cost D'2 and the credit limit E4, the value range of [η1, η2] is adjusted to construct the initialization model:

[0039]

[0040] Where N is the number of samples, D2i is the actual treatment cost of the i-th sample, is the initial estimate; calculate the residual: for each sample i, calculate the residual between the current model prediction and the actual value:

[0041]

[0042] Among them, D2i is the actual treatment cost of sample i, is the estimated cost of sample i, where X i =(mi,yi,D 2i ) is the feature vector of the i-th sample;

[0043] Use decision trees to fit the residual r ti , get the predicted value h of the tth tree t (x i ), the number of decision trees is 1-t;

[0044]

[0045] Where L is the loss function, θ is the tree ht learning rate, and t is the number of decision trees.

[0046] Add the new decision tree to the existing prediction model to update the overall prediction to get the final prediction model:

[0047]

[0048] Used to predict the treatment cost of new samples according to: To estimate the cost of treatment The deviation value Δ from the credit limit E4 is used to update the value range of the mapping function F(m, y, k) according to the size of Δ, where when Δ>0, η2 decreases, and when Δ<0, η2 increases, according to:

[0049] η′2=η2+sign(Δ)×ε2

[0050] To adjust η2, where ε2 is the adjustment step determined according to the distribution and size of Δ, and sign(Δ) is the adjustment function.

[0051] Furthermore, the learning rate θ of the tree ht is adjusted according to the number of diseases m suffered by the patient. The specific formula is:

[0052]

[0053] Where k is the total number of different diseases suffered by the patient, i is the i-th disease suffered by the patient, i = (1, 2, ..., k), yi is the corresponding level of the i-th disease of the patient, α and β are hyperparameters used to control the influence of disease level and number of diseases on the learning rate, and γ is an offset used to adjust the base value of the learning rate.

[0054] Furthermore, the patient's subsequent repayment record in S8 n Including the number of overdue times q1 times, overdue time q2 days, and repayment punctuality rate q3, assign points to each evaluation factor, and calculate the total score, and then according to:

[0055]

[0056] Adjust the credit limit E4 to credit limit E5, where n1 is the score for one overdue payment, n2 is the score for one overdue payment, and M k is the initial score. 10n3 is the repayment punctuality rate, and every 10% is n3 points. j = {j1, j2}, where j is the weight, j1 is the weight of the patient during hospitalization, and j2 is the weight of the patient during non-hospitalization.

[0057] Furthermore, the repayment punctuality rate q3 is calculated according to the formula:

[0058]

[0059] Among them, δ1 is the loan that has been repaid and due, and δ2 is the loan that has expired.

[0060] Beneficial Effects

[0061] Compared with the prior art, the present invention provides a credit medical quota evaluation method based on a decision boosting tree, which has the following beneficial effects:

[0062] 1. This credit medical quota evaluation method based on the decision-making boosting tree can more accurately match the patient's medical needs by adjusting the quota according to the patient's case and avoid the "one-size-fits-all" quota setting method. For patients with high rehabilitation costs, the credit medical quota evaluation system can provide certain financial support to reduce their financial burden.

[0063] 2. This method of evaluating medical credit limit based on a decision-boosting tree can more intuitively reflect the patient's credit situation by scoring the overdue situation during the patient's hospitalization and the overdue situation during the non-hospitalization period differently, so that the overdue situation caused by the patient during the hospitalization period has less impact on his credit limit. This method can convey a positive incentive signal, that is, encouraging patients to be honest and trustworthy, but also understanding and tolerating overdue behavior under special circumstances. This tolerance helps to enhance patients' credit awareness and promote them to more actively maintain their credit records in the future. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 A flow chart of a medical credit limit evaluation method based on a decision boosting tree proposed by the present invention;

[0065] Figure 2 A schematic diagram of social resume evaluation factors of a credit medical quota evaluation method based on a decision boosting tree proposed by the present invention;

[0066] Figure 3 This is a schematic diagram of a long-term rehabilitation case of a credit medical limit evaluation method based on a decision boosting tree proposed in the present invention. DETAILED DESCRIPTION

[0067] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0068] See also Figure 1 , in order to avoid patients having to go through the payment process before proceeding with the subsequent diagnosis and treatment process during a single visit, reduce payment links and waiting times, and alleviate patients' short-term financial payment pressure.

[0069] This embodiment provides a method for evaluating medical credit limit based on a decision boosting tree, comprising the following steps:

[0070] S1, establish a database DB for storing patient information, the patient registers for an outpatient clinic and submits personal information to the database DB, and matches the existing patient data in the database DB with the submitted patient personal information to determine whether the patient has a credit limit, if so, proceed to S3, otherwise proceed to S2;

[0071] S2, construct an initial credit evaluation system based on the patient's social resume evaluation factor set A1 to grant the patient an initial credit limit E1 and enter S3;

[0072] S3, the patient visits the doctor, the doctor diagnoses the patient's condition, and divides the patient into short-term recovery case B1 and long-term recovery case B2 according to the recovery time. If the patient is a short-term recovery case B1, the patient proceeds to S5, and if the patient is a long-term recovery case B2, the patient proceeds to S4;

[0073] S4, constructing a credit limit adjustment evaluation system based on the patient's long-term rehabilitation case B2, and adjusting the patient's credit limit E1 to a credit limit E4 through the credit limit adjustment evaluation system;

[0074] S5, the patient receives treatment according to the doctor's diagnosis, and the treatment expenses are paid by the credit line E4, resulting in the used credit line D1;

[0075] S6, settlement is carried out after the treatment is completed. During the settlement process, the patient's medical insurance reimbursement amount Y b To achieve quota settlement and obtain the settlement amount D2;

[0076] S7. Establish a machine learning model based on the patient's medical records and the actual settlement amount D2 to update the one-time amount adjustment system;

[0077] S8. The patient repays the settlement amount D2. Based on the patient's subsequent repayment record, A n A secondary credit limit adjustment evaluation system is established to obtain the credit limit E4, the patient's credit limit information is stored in the database DB, and the original data is updated.

[0078] During the actual medical consultation process, the patient can provide personal information to the database DB at the same time to match it with the user information in the database. The user information can be the user's ID number, so as to ensure the accuracy of data matching; when there is no corresponding credit limit data for the patient in the database DB, the user can choose whether to participate in credit medical care. When the patient chooses to participate in the use of credit medical care, he can submit his personal social resume to obtain the credit limit; after the credit limit is obtained, the patient will be interviewed in a follow-up manner. The doctor will diagnose the patient's condition. When the patient's case is a short-term recoverable case, such as a cold, fever, diarrhea and other diseases, the patient can go directly to the doctor, and the treatment expenses incurred during the consultation process will be paid in advance through the credit limit; if the patient's disease is a case B2 that requires long-term rehabilitation, the patient's credit limit will be adjusted according to the disease type m, the severity level y of the disease, and the recovery time j, so as to meet the needs of patients who need long-term treatment and avoid insufficient credit limit and the inability to enjoy the convenience of credit medical care during the consultation.

[0079] The above method can, on the one hand, avoid the need for patients to go through the payment process before proceeding with the subsequent diagnosis and treatment process during one visit, reduce payment links and waiting times, and alleviate patients' short-term financial payment pressure.

[0080] At the same time, the effect of three adjustments based on the patient's credit limit can ensure that patients with chronic diseases are given the corresponding credit limit, avoiding the inability to enjoy the convenience brought by credit medical care due to insufficient credit limit.

[0081] according to Figure 2 , the social resume evaluation factor set A1 in S1 includes x evaluation factors, each evaluation factor x has h rating levels, each evaluation factor x is assigned a weight w, and each evaluation level h is assigned a score j through the formula:

[0082]

[0083] To obtain the patient's social resume score F, the social resume score F value interval F∈[F min , F max ], and divide it into four sub-intervals from small to large:

[0084]

[0085] Among them, F1, F2, F3, F4 correspond to σ1, σ2, σ3, σ4 respectively, among which σ1, σ2, σ3, σ4 correspond to different amounts, and σ1<σ2<σ3<σ4, according to the formula:

[0086]

[0087] According to E1=F, the credit limit E1 is obtained.

[0088] The evaluation factors in this embodiment include the patient's age, income, occupation, and education level. By considering the patient's personal characteristics, such as age, income, and occupation, their economic ability and needs can be more accurately assessed, thereby providing more personalized services or cost support plans. This helps ensure that patients can obtain rehabilitation services that match their economic conditions and reduce their financial burden.

[0089] As a preferred technical solution in this embodiment, in actual use, σ1, σ2, σ3, and σ4 can be set to 3,000 yuan, 5,000 yuan, 8,000 yuan, and 10,000 yuan respectively. Setting the initial amount to the above amount can satisfy patients for some ordinary medical visits. The above amount only represents a more reasonable situation in actual implementation, and does not represent all initial amounts. The initial amount can be set according to actual usage according to different hospital charges and differences between regions.

[0090] See also Figure 3 In this embodiment, the long-term rehabilitation cases B2 in S4 include different diseases m that require long-term rehabilitation. Each disease m that requires long-term rehabilitation is divided into different severity levels y. Different severity levels y correspond to different rehabilitation times k. Each different disease m has a different initial quota E2. Each different disease m corresponds to a different initial weight W1. The mapping function F(m, y, k) is defined according to the above adjustment factors:

[0091]

[0092] In order to avoid the value of F(m, y, k) being too large or too small, the output value of the mapping function F(m, y, k) is truncated so that its output value is between the interval [η1, η2], through the formula:

[0093]

[0094] Where F(m, y, k) min is the minimum output value of the mapping function, F(m, y, k) max is the maximum output value of the mapping function. According to the mapping function F new (m, y, k) to adjust the weight W1 and get the adjusted weight W2:

[0095] W2=W1×F(m,y,k) new

[0096] The adjusted credit limit E3 is obtained based on the adjusted weight W2 and the initial credit limit E2 of each m:

[0097]

[0098] To obtain the adjustment amount E3 and adjust the primary credit amount E1, to obtain the secondary credit amount E4:

[0099] E4=E1+E3

[0100] where α m , β m , γ m is the coefficient, k min，m is the shortest recovery time for patients with m diseases, k max，m It is the longest recovery time of the patient suffering from disease m. When the patient is a short-term recovery case, E3=0.

[0101] Furthermore, the settlement amount D2 in S6 is calculated according to the formula:

[0102] D2=D1-Y b

[0103] D2 is the settlement amount, Y b The amount of medical insurance reimbursement. When the patient does not use medical insurance, Y b =0.

[0104] In the actual evaluation process, the preferred m may include cases that require long-term rehabilitation, such as nervous system injuries, bone and joint injuries, cardiovascular diseases, and chronic respiratory diseases. Of course, the above cases are only part of the long-term rehabilitation cases, and the number of cases or severity levels can be increased according to actual usage needs. The severity level y can be divided into level 1 (mild), level 2 (moderate), level 3 (severe), and level 4 (extremely severe). Similarly, the severity level can be further refined in the actual evaluation process to ensure the accuracy of the evaluation. And according to its severity level y, it can be divided into different rehabilitation times k. α m , β m , γ m The coefficients are determined by expert evaluation based on the patient's own physical condition, the hospital's medical level, and regional influence. Of course, the above factors only represent some of the factors that may affect the disease type m. In actual use, more factors that may affect the disease type m can be used to determine α. m , β m , γ m Equal coefficients.

[0105] At the same time, when a patient suffers from two chronic diseases, the patient should be evaluated twice through the credit adjustment system during the credit adjustment process to obtain the final adjusted credit amount of E4=E1+E3+E′3, where E′3 is the adjusted credit amount obtained after the second adjustment.

[0106] D2 is the actual settlement amount required when the patient's disease type is m and its severity level is y. The gradient boosting tree model is established under the above conditions to estimate the patient's treatment costs and obtain the estimated cost D'2. By comparing the difference Δ between the estimated cost D'2 and the credit limit E4, the value range of [η1, η2] is adjusted to construct the initialization model:

[0107]

[0108] Where N is the number of samples, D 2i is the actual treatment cost of the ith sample, is the initial estimate; calculate the residual: for each sample i, calculate the residual between the current model prediction and the actual value:

[0109]

[0110] Among them, D 2i is the actual treatment cost of sample i, is the estimated cost of sample i, where X i =(mi,yi,D 2i ) is the feature vector of the i-th sample;

[0111] Use decision trees to fit the residual r ti , get the predicted value h of the tth tree t (x i ), the number of decision trees is 1-t;

[0112]

[0113] Where L is the loss function, θ is the tree ht learning rate, and t is the number of decision trees.

[0114] Add the new decision tree to the existing prediction model to update the overall prediction to get the final prediction model:

[0115]

[0116] Used to predict the treatment cost of new samples according to: To estimate the cost of treatment The deviation value Δ from the credit limit E4 is used to update the value range of the mapping function F(m, y, k) according to the size of Δ, where when Δ>0, η2 decreases, and when Δ<0, η2 increases, according to:

[0117] η′2=η2+sign(Δ)×ε2

[0118] To adjust η2, where ε2 is the adjustment step determined according to the distribution and size of Δ, and sign(Δ) is the adjustment function. When Δ>0, the output value of sign(Δ) is 1, when Δ=0, the output value of sign(Δ) is 0, and when Δ<0, the output value of sign(Δ) is -1.

[0119] in is the loss function used to measure the model prediction and the true value y i The difference between. is the sum of the predictions of the t-1 trees for the i-th sample. i is the true label of the i-th sample, where argmin θt Represents finding the optimal parameter θ when minimizing the loss function L t process.

[0120] In the actual evaluation process, the preferred m may include cases that require long-term rehabilitation, such as nervous system injuries, bone and joint injuries, cardiovascular diseases, and chronic respiratory diseases. Of course, the above cases are only part of the long-term rehabilitation cases, and the number of cases or severity levels can be increased according to actual usage needs. The severity level y can be divided into level 1 (mild), level 2 (moderate), level 3 (severe), and level 4 (extremely severe). Similarly, the severity level can be further refined in the actual evaluation process to ensure the accuracy of the evaluation. And according to its severity level y, it can be divided into different rehabilitation times k. α m , β m , γ m The coefficients are determined by expert evaluation based on the patient's own physical condition, the hospital's medical level, and regional influences. Experts have deep professional knowledge and rich practical experience in their respective fields, thus obtaining more comprehensive and accurate evaluation results. It is worth noting that Figure 3 The diseases shown only represent a part of the cases in actual situations. During the actual evaluation process, they can be improved according to the patient's actual condition.

[0121] Preferably, in this embodiment, the learning rate θ of the tree ht can be adjusted according to the number of diseases m suffered by the patient, and the specific formula is:

[0122]

[0123] Where k is the total number of different diseases suffered by the patient, i is the i-th disease suffered by the patient, i = (1, 2, ..., k), yi is the corresponding level of the patient's i-th disease, α and β are hyperparameters used to control the degree of influence of the disease level and the number of diseases on the learning rate, γ is an offset used to adjust the basic value of the learning rate, where log(1+k) is used as part of the learning rate q formula to reflect the impact of the number of diseases k on the credit limit prediction. As the number of diseases increases, the value of log(1+k) will also increase, but the rate of increase will gradually decrease, which helps to avoid too high a learning rate q when the number of diseases is large, thereby maintaining the rationality and stability of the predicted amount.

[0124] In the above embodiment, by dynamically adjusting the learning rate q, the scheme can flexibly respond to the number and level of diseases of different patients. This flexibility enables the model to more accurately reflect the patient's credit status, because the number and level of diseases are often important factors in determining the credit limit. The scheme takes into account the impact of the number and level of diseases on the credit limit, making the allocation of credit limits more fair. Patients with more diseases or higher levels may face higher medical expenditure risks, so it is reasonable to give them higher credit limits.

[0125] In order to assign a quota that matches the treatment cost to the patient according to the type of illness they suffer, this embodiment can more accurately match the patient's medical needs through the above evaluation system, avoiding a "one-size-fits-all" quota setting method. For patients with high rehabilitation costs, the credit medical quota evaluation system can provide certain financial support to reduce their financial burden.

[0126] In order to make the use of the quota compatible with medical insurance, the settlement quota D2 in S5 in this embodiment is calculated according to the formula:

[0127] D2=D1-Y b

[0128] D2 is the settlement amount, Y b B is the reimbursement amount for medical insurance. When the patient does not use medical insurance, Y b B=0.

[0129] The patient's subsequent repayment record in S6A n Including the number of overdue times q1 times, overdue time q2 days, and repayment punctuality rate q3, assign points to each evaluation factor, and calculate the total score, and then according to:

[0130]

[0131] Adjust the credit limit E4 to credit limit E5, where n1 is the score for one overdue payment, n2 is the score for one overdue payment, and M kis the initial score, 10n3 is the repayment punctuality rate, and every 10% is n3 points, j = {j1, j2}, where j is the weight, the weight of the repayment during the patient's hospitalization is j1, and the weight of the repayment during the patient's non-hospitalization period is j2. The repayment punctuality rate q3 is based on the formula:

[0132]

[0133] Among them, δ1 is the loan that has been repaid and due, and δ2 is the loan that has expired.

[0134] As a preferred technical solution in this embodiment, M can be used in the actual evaluation process. k Set to 100 points, n1 is scored as 5, n2 is scored as 1, and n3 is scored as 2. The weight of the patient's repayment during hospitalization is set to j1, and the weight of the patient's repayment during non-hospitalization is set to j2. Then, by calculating the total score of the patient's credit record, the total score is compared with the initial score to obtain the credit limit adjustment ratio, and the credit limit E4 is adjusted to the credit limit E5 according to the credit limit adjustment ratio. This method can more intuitively reflect the patient's credit situation, so that the overdue situation caused by the patient during hospitalization has less impact on his credit limit. In the actual evaluation process, j1>j2, this evaluation method can ensure that when the patient is overdue during hospitalization, the deduction points are less, that is, the reduction of the credit limit is smaller.

[0135] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for evaluating medical credit limit based on a decision boosting tree, comprising the following steps, characterized in that: S1, establish a database DB for storing patient information, the patient registers for an outpatient clinic and submits personal information to the database DB, and matches the existing patient data in the database DB with the submitted patient personal information to determine whether the patient has a credit limit, if so, proceed to S3, otherwise proceed to S2; S2, construct an initial credit evaluation system based on the patient's social resume evaluation factor set A1 to grant the patient an initial credit limit E1 and enter S3; S3, the patient visits the doctor, the doctor diagnoses the patient's condition, and divides the patient into short-term recovery case B1 and long-term recovery case B2 according to the recovery time. If the patient is a short-term recovery case B1, the patient proceeds to S5, and if the patient is a long-term recovery case B2, the patient proceeds to S4; S4, constructing a credit limit adjustment evaluation system based on the patient's long-term rehabilitation case B2, and adjusting the patient's credit limit E1 to a credit limit E4 through the credit limit adjustment evaluation system; S5, the patient receives treatment according to the doctor's diagnosis, and the treatment expenses are paid by the credit line E4, generating the used credit line D1, and obtaining the patient's treatment history; S6, settlement is performed after the treatment is completed. During the settlement process, the amount is settled according to the patient's medical insurance reimbursement amount Yb, and the settlement amount D2 is obtained; S7. Establish a machine learning model based on the patient's medical records and the actual settlement amount D2 to update the one-time amount adjustment system; S8. The patient repays the settlement amount D2. Based on the patient's subsequent repayment record, A n A secondary credit limit adjustment evaluation system is established to obtain the credit limit E4, the patient's credit limit information is stored in the database DB, and the original patient information is updated.

2. According to claim 1, a method for evaluating medical credit limit based on decision boosting tree is characterized in that: The social resume evaluation factor set A1 in S2 includes x evaluation factors, each evaluation factor x has h rating levels, each evaluation factor x is assigned a weight w, and each evaluation level h is assigned a score j by the formula: To obtain the patient's social resume score F, the social resume score F value interval F∈[F min , F max ], divided into four sub-intervals from small to large: Among them, F1, F2, F3, F4 correspond to σ1, σ2, σ3, σ4 respectively, among which σ1, σ2, σ3, σ4 correspond to different amounts, and σ1<σ2<σ3<σ4, according to the formula: According to E1=N(F), the credit limit E1 is obtained.

3. According to claim 2, a method for evaluating medical credit limit based on decision boosting tree is characterized in that: The long-term rehabilitation cases B2 in S4 include different diseases m that require long-term rehabilitation. Each disease m that requires long-term rehabilitation is divided into different severity levels y. Different severity levels y correspond to different rehabilitation times k. Each different disease m corresponds to a different initial quota E2. Each different disease m corresponds to a different initial weight W1. The mapping function F(m, y, k) is defined according to the above adjustment factors: In order to avoid the value of F(m, y, k) being too large or too small, the output value of the mapping function F(m, y, k) is truncated so that its output value is between the interval [η1, η2], through the formula: Where F(m, y, k) min is the minimum output value of the mapping function, F(m, y, k) max is the maximum output value of the mapping function. According to the mapping function F new (m, y, k) to adjust the weight W1 and get the adjusted weight W2: W2=W1×F(m,y,k) new The adjusted credit limit E3 is obtained based on the adjusted weight W2 and the initial credit limit E2 of each m: To obtain the adjustment amount E3 and adjust the primary credit amount E1, to obtain the secondary credit amount E4: E4=E1+E3 where α m , β m , γ m is the coefficient, k min，m is the shortest recovery time for patients with m diseases, k max，m The longest recovery time for patients suffering from disease m.

4. According to the method for evaluating medical credit limit based on decision boosting tree in claim 1, it is characterized in that: The settlement amount D2 in S6 is based on the formula: D2=D1-Y b D2 is the settlement amount, Y b The amount of medical insurance reimbursement. When the patient does not use medical insurance, Y b =0.

5. A method for evaluating medical credit limit based on a decision boosting tree according to claim 3 or 4, characterized in that: D2 is the actual settlement amount required when the patient's disease type is m and its severity level is y. The gradient boosting tree model is established under the above conditions to estimate the patient's treatment costs and obtain the estimated cost D'2. By comparing the difference Δ between the estimated cost D'2 and the credit limit E4, the value range of [η1, η2] is adjusted to construct the initialization model: Where N is the number of samples, D 2i is the actual treatment cost of the ith sample, is the initial estimate; calculate the residual: for each sample i, calculate the residual between the current model prediction and the actual value: Among them, D 2i is the actual treatment cost of sample i, is the estimated cost of sample i, where X i =(mi,yi,D 2i ) is the feature vector of the i-th sample; Use decision trees to fit the residual r ti , get the predicted value h of the tth tree t (X i ), the number of decision trees is 1-t; Where L is the loss function, θ is the tree ht learning rate, and t is the number of decision trees. Add the new decision tree to the existing prediction model to update the overall prediction to get the final prediction model: Used to predict the treatment cost of new samples according to: To estimate the cost of treatment The deviation value Δ from the credit limit E4 is used to update the value range of the mapping function F(m, y, k) according to the size of Δ, where when Δ>0, η2 decreases, and when Δ<0, η2 increases, according to: η′2=η2+sign(Δ)×ε2 To adjust η2, where ε2 is the adjustment step determined according to the distribution and size of Δ, and sign(Δ) is the adjustment function.

6. A method for evaluating medical credit limit based on decision boosting tree according to claim 5, characterized in that: The learning rate θ of the tree ht is adjusted according to the number of diseases m suffered by the patient. The specific formula is: Where k is the total number of different diseases suffered by the patient, i is the i-th disease suffered by the patient, i = (1, 2, ..., k), yi is the corresponding level of the i-th disease of the patient, α and β are hyperparameters used to control the influence of disease level and number of diseases on the learning rate, and γ is an offset used to adjust the base value of the learning rate.

7. The method for evaluating medical credit limit based on decision boosting tree according to claim 1 is characterized in that: The patient's subsequent repayment record in S8 n Including the number of overdue times q1 times, overdue time q2 days, and repayment punctuality rate q3, assign points to each evaluation factor, and calculate the total score, and then according to: Adjust the credit limit E4 to credit limit E5, where n1 is the score for one overdue payment, n2 is the score for one overdue payment, and M k is the initial score. 10n3 is the repayment punctuality rate, and every 10% is n3 points, j={j1, j2}, where j is the weight, j1 is the weight during the patient's hospitalization, and j2 is the weight during the patient's non-hospitalization period.

8. The method for evaluating medical credit limit based on decision boosting tree according to claim 7 is characterized in that: The repayment punctuality rate q3 is based on the formula: Among them, δ1 is the loan that has been repaid and due, and δ2 is the loan that has expired.

Images