Hospital medical insurance non-toll income sub-account method based on LSTM (Long Short Term Memory)

Abstract

The invention relates to an LSTM (Long Short Term Memory)-based hospital out-of-medical-insurance self-fee income account-dividing method, which comprises the following steps of: constructing an LSTM model, inputting a predicted feature vector # imgabs3 # into the LSTM model for training based on normalized feature information # imgabs0 # in an out-of-medical-insurance self-fee income amount database and an actual proportion # imgabs2 # stored in a # imgabs1 # bank in a normalized history, the prediction proportion # imgabs5 # stored in the # imgabs4 # bank in the future is output, whether the prediction proportion # imgabs7 # stored in the # imgabs6 # bank in the future conforms to the distribution condition of the day or not is judged, the distribution proportion of hospital medical insurance exclusive self-fee income among all bank accounts is predicted and automatically adjusted by building the LSTM model, the efficiency and scientificity of fund division are remarkably improved, and the cost is reduced. According to the invention, tedious offline examination and approval and inter-bank transfer processes are avoided, balanced storage of funds is realized, the ability of a hospital management layer to make a fund distribution decision based on historical data and a current trend is enhanced, and the account distribution efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of big data resources, and in particular to an LSTM-based hospital non-medical insurance self-paid income account splitting method. Background Art

[0002] In recent years, in order to facilitate people's medical treatment and improve their medical experience, major hospitals have opened up mobile phone mobile settlement payment methods. In order to realize online settlement, hospitals usually apply for the hospital's official WeChat Alipay merchant and configure a bank account in each merchant number to deposit the medical expenses paid online by the public.

[0003] As the level of smart services in hospitals is getting higher and higher, the proportion of online settlement such as registration and payment by the public through mobile phones is also increasing. However, since WeChat and Alipay merchants providing online smart medical services can only configure one bank account, the hospital deposits too much money in a certain bank and too little money in other banks. In order to balance the hospital's deposits in various banks, the hospital usually has to conduct offline process approval, large-amount inter-bank transfers and other operations. The process is very inconvenient and affects the efficiency of account splitting. Therefore, a hospital non-medical insurance self-paid income account splitting method based on LSTM is proposed to solve the above problem. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention provides a method for splitting hospital self-paid income outside of medical insurance based on LSTM, which has the advantages of efficient splitting and solves the problem of low splitting efficiency.

[0005] To achieve the above purpose, the present invention provides the following technical solution: a method for dividing hospital self-paid income outside medical insurance based on LSTM, comprising the following specific steps: S1: Build an LSTM model based on the normalized feature information in the database of out-of-pocket income outside of medical insurance And the normalized history is stored in The actual proportion of banks , construct the prediction feature vector , the predicted feature vector Input into the LSTM model for training, and output the future stored in the The predicted proportion of banks ; S1.1: Obtain characteristic information from the database of out-of-pocket income amount outside of medical insurance and history stored in The actual proportion of banks , for feature information and history stored in The actual proportion of banks After normalization, the calculation expression is: In the formula, The maximum feature information , The minimum feature information , is the normalized feature information , For the largest history, store it in The actual proportion of banks , Store the minimum history in The actual proportion of banks , After normalization, the history is stored in The actual proportion of banks ; S1.2: Based on normalized feature information And the normalized history is stored in The actual proportion of banks , construct the prediction feature vector ; S1.3: Construct an LSTM model. The LSTM model includes a forget gate, an input gate, and an output gate. The calculation expression is: In the formula, is the input of the forget gate, is the sigmoid activation function, is the weight matrix of the forget gate, is the bias term of the forget gate, is the input of the input gate, is the candidate unit state, is the weight matrix of the input gate, is the weight matrix of the candidate unit state, is the bias term of the input gate, is the bias term of the candidate unit state, is the hyperbolic tangent activation function, is the cell state at the current time step, is the input of the output gate, is the hidden state at the current time step, is the weight matrix of the output gate, is the bias term of the output gate; S1.4: The predicted feature vector Divided into training set, validation set and test set, by minimizing the prediction ratio The actual ratio The loss function between is used to train the LSTM model, and the gradient descent method is used to update the parameters of the LSTM model. The calculation expression is: In the formula, is the loss function, is the number of training samples, is the number of banks, is a positive integer, is the learning rate, is the loss function About parameters The gradient of S1.5: Use cross entropy loss as the evaluation indicator to evaluate the performance of the LSTM model. The calculation expression is: In the formula, is the loss function, For the The samples for The true label of each category, 0 means it does not belong to this category, and 1 means it belongs to this category. For prediction The samples belong to The probability of each category satisfies ; S2: Determine the future and store it in The predicted proportion of banks Whether it meets the allocation of the day, if yes, then go to S3, if no, then go to S4; S2.1: Get the LSTM model and store it in the next step The predicted proportion of banks , construct the predicted proportion vector ,satisfy ; S2.2: Get the current day's deposits to The actual proportion of banks , construct the actual scale vector ,satisfy ; S2.3: Calculate the predicted proportion vector one by one and the actual scale vector The absolute value of the difference ,Will The absolute value of the difference Add together to get the total difference , determine the total difference Is it greater than the first threshold? If yes, then go to S3, if no, then go to S2.4. The calculation expression is: In the formula, is the total k differences The average value of is the compensation coefficient; S2.4: The absolute value of the maximum difference and the minimum absolute value of the difference Subtract to get the maximum difference , determine the maximum difference Is it greater than the second threshold? If yes, then go to S3, if no, then go to S2.5. The calculation expression is: In the formula, is the n largest difference The average value of is the compensation coefficient; S2.5: The total difference Divide by the total difference Number of , and get the average difference , determine the average difference Is it greater than the third threshold? If yes, then go to S3, if no, then go to S4. The calculation expression is: In the formula, for Total Difference The average value of is the compensation coefficient; S3: Store in the first The predicted proportion of banks Deposit the day's out-of-pocket income outside of medical insurance into the Bank, deposit the future into The predicted proportion of banks Deposit it into the database of out-of-pocket income amount outside medical insurance; S4: Build a polynomial regression model and store the history in The predicted proportion of banks Input to the polynomial regression model for training, and the output is adjusted and stored in the next The predicted proportion of banks , according to the adjusted future deposit The predicted proportion of banks Deposit the day's out-of-pocket income outside of medical insurance into the Bank, deposit the adjusted future amount into The predicted proportion of banks Deposit it into the database of out-of-pocket income amount outside medical insurance; S4.1: Get the predicted scale vector and the actual scale vector , divided into training set and test set; S4.2: For A polynomial regression model is constructed for each bank, and the calculation expression is: In the formula, Deposit to the first The forecast ratio of each bank is For the The size of the hospital, For the Hospital No. Quarterly impact factor, is the error term; S4.2.1.1: Get the last year’s Hospital outpatient volume , last year Number of surgeries per hospital and last year Number of hospital beds ; S4.2.1.2: Construct Hospital size Model, the calculation expression is: In the formula, is the intercept term, For the Hospital outpatient volume The regression coefficient of For the Number of surgeries per hospital The regression coefficient of No. Number of hospital beds The regression coefficient of is the error term; S4.2.1.3: Construct the design matrix , the calculation expression is: In the formula, is the design matrix, is the size of the hospital; S4.2.1.4: Calculate regression coefficients , the calculation expression is: In the formula, for The transposed matrix of for The inverse matrix of S4.2.2.1: Get the last year’s Hospital No. Quarterly out-of-pocket income outside of medical insurance and last year The amount of out-of-pocket income of a hospital in a year outside of medical insurance ; S4.2.2.2: Construct the Hospital No. Quarterly Impact Factor Model, the calculation expression is: In the formula, For the Hospital No. Quarterly actual ratio, For the Hospital No. Quarterly forecast ratio, ∈ ; S4.3: Using mean square error The polynomial regression model is evaluated using the expression: S4.4: The adjusted proportional vector after output Divide it into E subsets, each subset is used as the test set in turn, and the remaining E-1 subsets are used as training sets. Repeat this process E times, each time selecting a different subset as the test set, and calculate the average performance of E evaluations as the performance indicator of the model.

[0006] Compared with the prior art, the technical solution of this application has the following beneficial effects: The hospital's automatic account splitting method for self-paid income builds an LSTM model to predict and automatically adjust the distribution ratio of the hospital's self-paid income outside of medical insurance among various bank accounts, significantly improving the efficiency and scientificity of fund splitting. It not only avoids cumbersome offline approval and inter-bank transfer processes, and achieves balanced deposit of funds, but also enhances the hospital management's ability to make fund allocation decisions based on historical data and current trends, thereby improving account splitting efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 It is a flow chart of the present invention; Figure 2 This is a flow chart of step 1 of the present invention; Figure 3 This is a flow chart of step 2 of the present invention; Figure 4 This is a flow chart of step 4 of the present invention; Figure 5 The scale of the hospital is Flow chart of the calculation method; Figure 6 The impact factor of this invention Flowchart of the calculation method. DETAILED DESCRIPTION

[0008] The following will be combined with the drawings in 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.

[0009] See also Figure 1-6 In this embodiment, a method for dividing hospital self-paid income outside of medical insurance based on LSTM includes the following specific steps: S1: Build an LSTM model based on the normalized feature information in the database of out-of-pocket income outside of medical insurance And the normalized history is stored in The actual proportion of banks , construct the prediction feature vector , the predicted feature vector Input into the LSTM model for training, and output the future stored in the The predicted proportion of banks , using the LSTM model to capture time series features, accurately predict the proportion of future bank deposits, and provide a scientific basis for fund allocation; S1.1: Obtain characteristic information from the database of out-of-pocket income amount outside of medical insurance and history stored in The actual proportion of banks , for feature information and history stored in The actual proportion of banks After normalization, the calculation expression is: In the formula, The maximum feature information , The minimum feature information , is the normalized feature information , For the largest history, store it in The actual proportion of banks , Store the minimum history in The actual proportion of banks , After normalization, the history is stored in The actual proportion of banks ; S1.2: Based on normalized feature information And the normalized history is stored in The actual proportion of banks , construct the prediction feature vector ; S1.3: Construct an LSTM model. The LSTM model includes a forget gate, an input gate, and an output gate. The calculation expression is: In the formula, is the input of the forget gate, is the sigmoid activation function, is the weight matrix of the forget gate, is the bias term of the forget gate, is the input of the input gate, is the candidate unit state, is the weight matrix of the input gate, is the weight matrix of the candidate unit state, is the bias term of the input gate, is the bias term of the candidate unit state, is the hyperbolic tangent activation function, is the cell state at the current time step, is the input of the output gate, is the hidden state at the current time step, is the weight matrix of the output gate, is the bias term of the output gate; S1.4: The predicted feature vector Divided into training set, validation set and test set, by minimizing the prediction ratio The actual ratio The loss function between is used to train the LSTM model, and the gradient descent method is used to update the parameters of the LSTM model. The calculation expression is: In the formula, is the loss function, is the number of training samples, is the number of banks, is a positive integer, is the learning rate, is the loss function About parameters The gradient of S1.5: Use cross entropy loss as the evaluation indicator to evaluate the performance of the LSTM model. The calculation expression is: In the formula, is the loss function, For the The samples for The true label of each category, 0 means it does not belong to this category, and 1 means it belongs to this category. For prediction The samples belong to The probability of each category satisfies ; S2: Determine the future and store it in The predicted proportion of banks Whether it is consistent with the allocation of the day, if yes, then go to S3, if not, then go to S4, and verify the accuracy of the model prediction by comparing the prediction with the actual allocation, so as to ensure the rationality and reliability of the fund allocation decision; S2.1: Get the LSTM model and store it in the next step The predicted proportion of banks , construct the predicted proportion vector ,satisfy ; S2.2: Get the current day's deposits to The actual proportion of banks , construct the actual scale vector ,satisfy ; S2.3: Calculate the predicted proportion vector one by one and the actual scale vector The absolute value of the difference ,Will The absolute value of the difference Add together to get the total difference , determine the total difference Is it greater than the first threshold? If yes, then go to S3, if no, then go to S2.4. The calculation expression is: In the formula, is the total k differences The average value of is the compensation coefficient; S2.4: The absolute value of the maximum difference and the minimum absolute value of the difference Subtract to get the maximum difference , determine the maximum difference Is it greater than the second threshold? If yes, then go to S3, if no, then go to S2.5. The calculation expression is: In the formula, is the n largest difference The average value of is the compensation coefficient; S2.5: The total difference Divide by the total difference Number of , and get the average difference , determine the average difference Is it greater than the third threshold? If yes, then go to S3, if no, then go to S4. The calculation expression is: In the formula, for Total Difference The average value of is the compensation coefficient; S3: Store in the first The predicted proportion of banks Deposit the day's out-of-pocket income outside of medical insurance into the Bank, deposit the future into The predicted proportion of banks The amount is stored in the database of out-of-pocket income outside of medical insurance, and funds are allocated based on the accurate forecast ratio to optimize resource allocation. At the same time, the database is updated to support subsequent forecasts and decisions; S4: Build a polynomial regression model and store the history in The predicted proportion of banks Input to the polynomial regression model for training, and the output is adjusted and stored in the next The predicted proportion of banks , according to the adjusted future deposit The predicted proportion of banks Deposit the day's out-of-pocket income outside of medical insurance into the Bank, deposit the adjusted future amount into The predicted proportion of banks The amount is stored in the database of out-of-pocket income outside of medical insurance. When the prediction is inaccurate, the prediction result is further adjusted through the polynomial regression model to improve the prediction accuracy and ensure more accurate and reasonable allocation of funds. S4.1: Get the predicted scale vector and the actual scale vector , divided into training set and test set; S4.2: For A polynomial regression model is constructed for each bank, and the calculation expression is: In the formula, Deposit to the first The forecast ratio of each bank is For the The size of the hospital, For the Hospital No. Quarterly impact factor, is the error term; S4.2.1.1: Get the last year’s Hospital outpatient volume , last year Number of surgeries per hospital and last year Number of hospital beds ; S4.2.1.2: Construct Hospital size Model, the calculation expression is: In the formula, is the intercept term, For the Hospital outpatient volume The regression coefficient of For the Number of surgeries per hospital The regression coefficient of No. Number of hospital beds The regression coefficient of is the error term; S4.2.1.3: Construct the design matrix , the calculation expression is: In the formula, is the design matrix, is the size of the hospital; S4.2.1.4: Calculate regression coefficients , the calculation expression is: In the formula, for The transposed matrix of for The inverse matrix of S4.2.2.1: Get the last year’s Hospital No. Quarterly out-of-pocket income outside of medical insurance and last year The amount of out-of-pocket income of a hospital in a year outside of medical insurance ; S4.2.2.2: Construct the Hospital No. Quarterly Impact Factor Model, the calculation expression is: In the formula, For the Hospital No. Quarterly actual ratio, For the Hospital No. Quarterly forecast ratio, ∈ , the first quarter is from January 1 to March 31, the second quarter is from April 1 to June 30, the third quarter is from July 1 to September 30, and the fourth quarter is from October 1 to December 31; S4.3: Using mean square error The polynomial regression model is evaluated using the expression: S4.4: The adjusted proportional vector after output Divide it into E subsets, each subset is used as the test set in turn, and the remaining E-1 subsets are used as training sets. Repeat this process E times, each time selecting a different subset as the test set, and calculate the average performance of E evaluations as the performance indicator of the model.

[0010] The working principle of the above embodiment is: The hospital's automatic account splitting method for self-paid income builds an LSTM model to predict and automatically adjust the distribution ratio of the hospital's self-paid income outside of medical insurance among various bank accounts, significantly improving the efficiency and scientificity of fund splitting. It not only avoids cumbersome offline approval and inter-bank transfer processes, and achieves balanced deposit of funds, but also enhances the hospital management's ability to make fund allocation decisions based on historical data and current trends, thereby improving account splitting efficiency.

[0011] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0012] 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 dividing hospital's out-of-pocket income outside of medical insurance based on LSTM, characterized in that: The specific steps include: S1: Build an LSTM model based on the normalized feature information in the database of out-of-pocket income outside of medical insurance And the normalized history is stored in The actual proportion of banks , construct the prediction feature vector , the predicted feature vector Input into the LSTM model for training, and output the future stored in the The predicted proportion of banks ; S2: Determine the future storage The predicted proportion of banks Whether it meets the allocation of the day, if yes, then go to S3, if no, then go to S4; S3: Store in the first The predicted proportion of banks Deposit the day's out-of-pocket income outside of medical insurance into the Bank, deposit the future into The predicted proportion of banks Deposit it into the database of out-of-pocket income amount outside medical insurance; S4: Build a polynomial regression model and store the history in The predicted proportion of banks Input to the polynomial regression model for training, and the output is adjusted and stored in the next The predicted proportion of banks , according to the adjusted future deposit to the The predicted proportion of banks Deposit the day's out-of-pocket income outside of medical insurance into the Bank, deposit the adjusted future amount into The predicted proportion of banks Deposited into the database of out-of-pocket income outside of medical insurance.

2. A LSTM-based hospital non-medical insurance self-paid income account splitting method as claimed in claim 1, characterized in that: Step S1 constructs an LSTM model based on the normalized feature information in the database of out-of-pocket income outside of medical insurance And the normalized history is stored in The actual proportion of banks , construct the prediction feature vector , the predicted feature vector Input into the LSTM model for training, and output the future stored in the The predicted proportion of banks , including the following sub-steps: S1.1: Obtain characteristic information from the database of out-of-pocket income amount outside of medical insurance and history stored in The actual proportion of banks , for feature information and history stored in The actual proportion of banks After normalization, the calculation expression is: In the formula, The maximum feature information , The minimum feature information , is the normalized feature information , For the largest history, store it in The actual proportion of banks , Store the minimum history in The actual proportion of banks , After normalization, the history is stored in The actual proportion of banks ; S1.2: Based on normalized feature information And the normalized history is stored in The actual proportion of banks , construct the prediction feature vector ; S1.3: Build an LSTM model. The LSTM model includes a forget gate, an input gate, and an output gate. S1.4: The predicted feature vector Divided into training set, validation set and test set, by minimizing the prediction ratio The actual ratio The loss function between is used to train the LSTM model, and the gradient descent method is used to update the parameters of the LSTM model; S1.5: Use cross entropy loss as the evaluation metric to evaluate the performance of the LSTM model.

3. The LSTM-based hospital non-medical insurance self-paid income account splitting method as claimed in claim 1, characterized in that: Step S2 determines the future storage The predicted proportion of banks Whether it meets the allocation of the day, if yes, then go to S3, if no, then go to S4, including the following sub-steps: S2.1: Get the LSTM model and store it in the next step The predicted proportion of banks , construct the predicted proportion vector ,satisfy ; S2.2: Get the current day's deposits to The actual proportion of banks , construct the actual scale vector ,satisfy ; S2.3: Calculate the predicted proportion vector one by one and the actual scale vector The absolute value of the difference ,Will The absolute value of the difference Add together to get the total difference , determine the total difference Is it greater than the first threshold? If yes, then go to S3, if no, then go to S2.

4. The calculation expression is: In the formula, is the total k differences The average value of is the compensation coefficient; S2.4: The absolute value of the maximum difference and the minimum absolute value of the difference Subtract to get the maximum difference , determine the maximum difference Is it greater than the second threshold? If yes, then go to S3, if no, then go to S2.

5. The calculation expression is: In the formula, is the n largest difference The average value of is the compensation coefficient; S2.5: The total difference Divide by the total difference Number of , and get the average difference , determine the average difference Is it greater than the third threshold? If yes, then go to S3, if no, then go to S4. The calculation expression is: In the formula, for Total Difference The average value of is the compensation coefficient.

4. The LSTM-based hospital non-medical insurance self-paid income account splitting method as claimed in claim 1, characterized in that: Step S4 constructs a polynomial regression model and stores the history in the The predicted proportion of banks Input to the polynomial regression model for training, and the output is adjusted and stored in the next The predicted proportion of banks , according to the adjusted future deposit to the The predicted proportion of banks Deposit the day's out-of-pocket income outside of medical insurance into the Bank, including the following sub-steps: S4.1: Get the predicted scale vector and the actual scale vector , divided into training set and test set; S4.2: For A polynomial regression model is constructed for each bank, and the calculation expression is: In the formula, Deposit to the first The forecast ratio of each bank is For the The size of the hospital, For the Hospital No. The quarterly impact factor, is the error term; S4.3: Using mean square error The polynomial regression model is evaluated using the expression: S4.4: The adjusted proportional vector after output Divide it into E subsets, each subset is used as the test set in turn, and the remaining E-1 subsets are used as training sets. Repeat this process E times, each time selecting a different subset as the test set, and calculate the average performance of E evaluations as the performance indicator of the model.

5. A method for dividing hospital's non-medical insurance self-paid income based on LSTM as claimed in claim 4, characterized in that: Step 4.2 Hospital size The calculation method is: S4.2.1.1: Get the last year’s Hospital outpatient volume Last year Number of surgeries per hospital and last year Number of hospital beds ; S4.2.1.2: Construct Hospital size Model, the calculation expression is: In the formula, is the intercept term, For the Hospital outpatient volume The regression coefficient of For the Number of surgeries per hospital The regression coefficient of No. Number of hospital beds The regression coefficient of is the error term; S4.2.1.3: Construct the design matrix , the calculation expression is: In the formula, is the design matrix, is the size of the hospital; S4.2.1.4: Calculate regression coefficients , the calculation expression is: In the formula, for The transposed matrix of for The inverse matrix of .

6. A LSTM-based hospital non-medical insurance self-paid income account splitting method as claimed in claim 4, characterized in that: Step 4.2 Hospital No. Quarterly Impact Factor The calculation method is: S4.2.2.1: Get the last year’s Hospital No. Quarterly out-of-pocket income outside of medical insurance and last year The amount of out-of-pocket income of a hospital in a year outside of medical insurance ; S4.2.2.2: Construct the Hospital No. Quarterly Impact Factor Model, the calculation expression is: In the formula, For the Hospital No. Quarterly actual ratio, For the Hospital No. Quarterly forecast ratio, ∈ .

Images