Blood inventory early warning method and system based on deep learning

Abstract

The present invention proposes a blood inventory early warning method and system based on deep learning, the method includes converting the blood usage information of the target hospital into a vector and storing it in a first database; and converting the blood usage information of hospitals of the same category as the target hospital into a vector and storing it in a second database; simplifying the attributes of the first database and the second database based on the rough set theory; dividing the data in the simplified information database into a training set, a validation set and a test set according to a preset ratio; inputting the blood usage information of hospitals of the same category as the target hospital in a second preset time period into the model, training the model, and verifying and selecting the model with the smallest total loss on the validation set, and predicting the blood inventory through the test set test. Based on this method, an early warning system is also proposed. The present invention uses deep learning technology to analyze the information of hospitals of the same category and the blood usage of patients in the hospital in the past, and infers a reasonable blood inventory, which solves the problem that hospitals cannot accurately and reasonably set inventory early warning quantities.

Description

Technical Field

[0001] The present invention belongs to the technical field of blood inventory early warning, and in particular, to a blood inventory early warning method and system based on deep learning. Background Art

[0002] Due to the high demand for blood and low blood collection in most parts of my country, the current blood shortage problem is still a difficult problem facing my country's public health and blood supply. According to the blood usage of various departments in the hospital, setting appropriate inventory warnings to ensure the hospital's blood usage is an important task of the blood transfusion department. Due to the large number of patients in the hospital and the complex blood usage situation, the blood transfusion department staff cannot accurately set the inventory warning quantity that meets the hospital's blood usage situation. Too low inventory warning quantity leads to insufficient blood reserves in the hospital, affecting the treatment of patients. Too high inventory warning quantity is prone to excessive blood reserves, resulting in blood failure and waste, and will also lead to uneven blood distribution in hospitals in the region. Setting a scientific and reasonable blood inventory can reduce the blood expiration rate and alleviate the problem of insufficient blood supply while ensuring the daily blood use of the hospital, ensuring that the blood inventory remains in a stable state and improving the blood supply response capability.

[0003] With the application of information technology in the medical field, using information technology to assist blood transfusion staff in setting appropriate inventory warning quantities can reduce the situation of insufficient and excessive blood reserves, which is of great significance to alleviating the relationship between regional blood supply and demand. The current blood inventory warning system can only simply assist blood transfusion staff in setting inventory warnings, and issue warnings when the blood inventory reaches this value, but it cannot scientifically analyze whether the set warning value is reasonable. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes a blood inventory warning method and system based on deep learning. It solves the problem that hospitals cannot accurately and reasonably set inventory warning quantities, avoids the occurrence of insufficient blood reserves and excessive blood reserves leading to high blood scrapping rates, and is of great significance to alleviating the regional blood supply and demand relationship.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] The blood inventory early warning method based on deep learning includes the following steps:

[0007] The target hospital's blood usage information is converted into a vector and stored in a first database; and the target hospital's blood usage information is converted into a vector and stored in a second database;

[0008] Based on the rough set theory, attributes of the first database and the second database are simplified to obtain a simplified information database; data in the simplified information database are divided into a training set, a validation set and a test set according to a preset ratio;

[0009] The blood usage information of the target hospital and the same type of hospitals in the second preset time period is input into the LSTM model, the LSTM model is trained, and the model with the smallest total loss is selected and verified on the validation set, and the blood inventory is predicted through the test set test.

[0010] Furthermore, the selection criteria for hospitals of the same category as the target hospital are:

[0011] Obtaining several items of basic hospital information; the basic hospital information includes but is not limited to the region, hospital level, hospital category, and surgical volume;

[0012] The similarity value is set according to the similarity between each basic information and the corresponding basic information of the target hospital; the weight of each hospital's basic information is set to obtain the relevance to the target hospital.

[0013] Furthermore, the step of converting the blood usage information of the target hospital into a vector and storing it in the first database specifically includes: converting the blood usage information of the target hospital f in the first preset time period t into a vector F t And stored in the first database A.

[0014] Furthermore, the step of converting the blood use information of hospitals of the same category as the target hospital into a vector and storing it in the second database specifically includes: for each hospital of the same category as the target hospital p i (0≤i≤n), convert the blood usage information of the same type of hospital in the second preset time period d into a vector And stored in information database B; where n is the number of hospitals of the same category as the target hospital.

[0015] Furthermore, the process of reducing the attributes of the first database and the second database based on the rough set theory is as follows:

[0016] Setting conditional attributes and decision attributes; the decision attributes are the blood usage of each blood type; the attributes other than the decision attributes are conditional attributes;

[0017] The importance of decision attributes is calculated by using rough set dependency, and the attribute relevance is calculated by using mutual information, and the first core attribute set a of the first database A is obtained as follows: a = {a1, ..., a n} and the second core attribute set b of the second database B = {b1, ..., b n};

[0018] Calculate the first core attribute set a = {a1, ..., a n} and the second core attribute set b={b1,....,b n} to obtain the third core attribute set c = {c1, ..., c l};

[0019] From the first database A, a third core attribute set c={c1, ..., c l} data constitute the reduced information base A'.

[0020] Furthermore, the process of dividing the data in the simplified information database into a training set, a validation set and a test set according to a preset ratio specifically includes:

[0021] The data in the information base A' are used as a training set, a validation set, and a test set according to 70%, 15%, and 15% of the information vectors, respectively.

[0022] Furthermore, the blood usage information of the target hospital and the hospitals of the same category in the second preset time period is input into the LSTM model, the LSTM model is trained, and the process of selecting the model with the smallest total loss is verified on the validation set:

[0023] The blood usage information of the target hospital and the same type of hospitals in the second preset time period d is Input into the LSTM model, take the minimum loss function as the optimization goal, use the training set to train the LSTM model with a preset number of times as a cycle, and obtain the blood use information vector within the d+1 date As predictive blood use information;

[0024] The actual blood usage of each blood type on date d+1 is analyzed with the predicted blood usage of each blood type, and the prediction error rate vector of each blood type is calculated. Total error rate As the loss function of the LSTM model; the LSTM model is verified for a preset number of times using the verification set, and the model with the smallest total loss function is tested in the test set.

[0025] Furthermore, the process of predicting blood inventory through test set testing includes:

[0026] The test set is used to test and obtain the sum of loss functions, and whether the sum of loss functions meets the preset requirements is determined, and the model that meets the preset requirements is set as the final prediction model.

[0027] Furthermore, the method also includes obtaining the blood usage of each blood category in the future preset days according to the final prediction model, and setting a blood classification warning according to the blood inventory.

[0028] The present invention also proposes a blood inventory early warning system based on deep learning, the system comprising a conversion module, a simplification module and a learning module;

[0029] The conversion module is used to convert the blood usage information of the target hospital into a vector and store it in the first database; and convert the blood usage information of hospitals of the same category as the target hospital into a vector and store it in the second database;

[0030] The reduction module is used to reduce the attributes of the first database and the second database based on the rough set theory to obtain a reduced information database; the data in the reduced information database is divided into a training set, a validation set and a test set according to a preset ratio;

[0031] The learning module is used to input the blood usage information of the target hospital and the same type of hospitals in the second preset time period into the LSTM model, train the LSTM model, and verify and select the model with the smallest total loss on the validation set, predict the blood inventory through the test set test, and set a blood classification warning based on the blood inventory.

[0032] The effects provided in the content of the invention are only the effects of the embodiments, not all the effects of the invention. One of the above technical solutions has the following advantages or beneficial effects:

[0033] The present invention proposes a blood inventory early warning method and system based on deep learning, the method includes converting the blood usage information of the target hospital into a vector and storing it in a first database; and converting the blood usage information of the target hospital's same category hospitals into a vector and storing it in a second database; based on the rough set theory, the first database and the second database are attribute simplified to obtain a simplified information database; the data in the simplified information database is divided into a training set, a validation set and a test set according to a preset ratio; the blood usage information of the target hospital's same category hospitals in the second preset time period is input into the LSTM model, the LSTM model is trained, and the model with the smallest total loss is selected on the validation set, and the blood inventory is predicted by the test set test. A blood inventory early warning method based on deep learning is also proposed. A blood inventory early warning system based on deep learning is also proposed. The present invention uses deep learning technology to analyze the information of hospitals of the same category and the blood usage of patients in the hospital in the past, and infers a reasonable blood inventory, which solves the problem that hospitals cannot accurately and reasonably set the inventory early warning amount, avoids the occurrence of the problem of insufficient blood preparation and excessive blood preparation leading to a high blood scrap rate, and is of great significance to alleviating the regional blood supply and demand relationship. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] like Figure 1 This is a general flow chart of a blood inventory early warning method based on deep learning according to Embodiment 1 of the present invention;

[0035] like Figure 2 This is a flow chart of constructing a target hospital database and a database of hospitals of the same category as the target hospital in the blood inventory early warning method based on deep learning in Example 1 of the present invention;

[0036] like Figure 3This is a flow chart of the information simplification method in the blood inventory early warning method based on deep learning in Example 1 of the present invention;

[0037] like Figure 4 This is a flow chart of a deep learning method in a blood inventory early warning method based on deep learning in Example 1 of the present invention;

[0038] like Figure 5 This is a schematic diagram of a blood inventory early warning system based on deep learning according to Example 2 of the present invention. DETAILED DESCRIPTION

[0039] In order to clearly illustrate the technical features of the present solution, the present invention is described in detail below through specific implementation methods and in conjunction with the accompanying drawings. The disclosure below provides many different embodiments or examples for realizing different structures of the present invention. In order to simplify the disclosure of the present invention, the components and settings of specific examples are described below. In addition, the present invention may repeat reference numbers and / or letters in different examples. This repetition is for the purpose of simplification and clarity, and does not itself indicate the relationship between the various embodiments and / or settings discussed. It should be noted that the components illustrated in the accompanying drawings are not necessarily drawn to scale. The present invention omits the description of known components and processing techniques and processes to avoid unnecessary limitations on the present invention.

[0040] Example 1

[0041] Embodiment 1 of the present invention proposes a blood inventory warning method based on deep learning. A deep learning algorithm is provided to analyze the daily blood usage of the hospital and predict the blood usage in the next few days, so as to reasonably set the blood inventory warning amount. Figure 1 The overall flow chart of the blood inventory early warning method based on deep learning in Example 1 of the present invention is given. The flow chart includes:

[0042] The target hospital's blood usage information is converted into a vector and stored in a first database; and the target hospital's blood usage information is converted into a vector and stored in a second database;

[0043] Based on the rough set theory, attributes of the first database and the second database are simplified to obtain a simplified information database; the data in the simplified information database is divided into a training set, a validation set and a test set according to a preset ratio;

[0044] The blood usage information of the target hospital and the same type of hospitals in the second preset time period is input into the LSTM model, the LSTM model is trained, and the model with the smallest total loss is selected and verified on the validation set, and the blood inventory is predicted through the test set test.

[0045] like Figure 2 This is a flow chart of constructing a target hospital database and a database of hospitals of the same category as the target hospital in the blood inventory early warning method based on deep learning in Example 1 of the present invention;

[0046] The selection criteria for hospitals of the same category as the target hospital are as follows: obtaining several items of basic hospital information; the basic hospital information includes but is not limited to the region, hospital level, hospital category, and surgical volume; setting a similarity value based on the degree of similarity between each basic information and the corresponding basic information of the target hospital; setting the weight of each basic information of the hospital to obtain the relevance to the target hospital.

[0047] According to the basic information of the hospital and rule E, select n hospitals p (p = p1, p2, ... p n ). The specific content of Rule E is as follows:

[0048] Assume that m items of basic hospital information s are obtained i (i=1,2,...m)(region, hospital level, hospital category, surgical volume, etc.), and quantify this information. Each piece of information is set with five values ​​according to its similarity with the hospital, namely s i (i=1,2,...m), the similarity increases from 1 to 5, and the weight of each basic information is set to ω i ,satisfy Get the relevance to this hospital

[0049] For each hospital p i (0≤i≤n), convert the blood usage information of the hospital within date d (d represents a certain day) into a vector And stored in database B.

[0050] Collection Hospital p i (0≤i≤n) Blood usage information within date d. Blood usage information mainly includes two categories: one is numerical information (such as the amount of blood used for a certain blood type, blood type ID, etc.) and the other is text information (such as gender, department name, blood type name, etc.).

[0051] Date type is also an important factor affecting patient admission to hospital. Date type needs to be included as one of the factors affecting the amount of blood used in hospitals, with weekdays, weekends and holidays represented by 0, 1 and 2 respectively.

[0052] (a) No processing is done on numerical information.

[0053] (b) For text information, convert it into numerical information, as shown below:

[0054] Example 1: The patient's gender information is "male" or "female", and it is stipulated that "male" is represented by "0" and "female" is represented by "1".

[0055] Example 2: Assuming the number of departments is n, the department numbers are specified to be represented from "0" to "n".

[0056] vector There are k dimensions in total, each dimension represents an item in the hospital's blood use information. It is divided into two parts. The first T1 item represents information that is not related to the specific amount of blood used, recorded as The latter T2 item represents information related to the specific amount of blood used, recorded as

[0057] but

[0058] The information variables corresponding to the blood usage information of all dates are stored in the blood transfusion information database A, that is, the blood transfusion information database A contains the blood usage vectors of all hospitals on all dates.

[0059] Convert the blood usage information of the target hospital f in the first preset time period t into a vector F t And stored in the first database A.

[0060] Based on the rough set theory, the attributes of the first database and the second database are simplified to obtain a simplified information database; Figure 3 This is a flow chart of the information simplification method in the blood inventory early warning method based on deep learning in Example 1 of the present invention;

[0061] Input information bases A and B into the attribute reduction model.

[0062] Setting conditional attributes and decision attributes; the decision attributes are the blood usage of each blood type; the attributes other than the decision attributes are conditional attributes;

[0063] The importance of decision attributes is calculated by using rough set dependency, and the attribute relevance is calculated by using mutual information, and the first core attribute set a of the first database A is obtained as follows: a = {a1, ..., a n} and the second core attribute set b of the second database B = {b1, ..., b n};

[0064] Calculate the first core attribute set a = {a1, ..., a n} and the second core attribute set b={b1,....,b n} to obtain the third core attribute set c = {c1, ..., c l};

[0065] From the first database A, a third core attribute set c={c1, ..., c l} data constitute the reduced information base A'.

[0066] The data in the simplified information database is divided into a training set, a validation set and a test set according to a preset ratio; the blood usage information of the target hospital and the same category of hospitals in the second preset time period is input into the LSTM model, the LSTM model is trained, and the model with the smallest total loss is selected on the validation set, and the blood inventory is predicted through the test set test.

[0067] like Figure 4 This is a flow chart of a deep learning method in a blood inventory early warning method based on deep learning in Example 1 of the present invention;

[0068] The data in the information database A' is used as the training set, verification set, and test set according to 70%, 15%, and 15% of the information vectors, respectively.

[0069] The blood usage information of the target hospital and the same type of hospitals in the second preset time period d is Input into the LSTM model, take the minimum loss function as the optimization goal, use the training set to train the LSTM model with a preset number of times as a cycle, and obtain the blood use information vector within the d+1 date As predictive blood use information;

[0070] The actual blood usage of each blood type on date d+1 is analyzed with the predicted blood usage of each blood type, and the prediction error rate vector of each blood type is calculated. Total error rate As the loss function of the LSTM model; the LSTM model is verified for a preset number of times using the validation set, and the model with the smallest total loss function is tested in the test set.

[0071] The test set is used to test and obtain the sum of the loss function. It is determined whether the sum of the loss function meets the preset requirements, and the model that meets the preset requirements is set as the final prediction model.

[0072] The training process is as follows: Use the training set to save the trained model with 500 training cycles. Use the model to verify on the validation set, compare the predicted blood usage with the actual blood usage, and calculate the total loss function of the blood usage information on the validation set. After every 100 verifications, select the model with the smallest total loss function and test it on the test set. Test the test set to obtain the total loss function and set the ideal total loss function. If the set conditions are not met, retrain the model; if the set conditions are met, use it as the final prediction model.

[0073] The predicted model is used to predict the blood usage of each blood type in the next few days, and to set a reasonable inventory warning amount. For example, the third-level warning: the blood volume is lower than 5 working days for 3 consecutive days; the second-level warning: the blood volume is lower than 4 working days for 2 consecutive days; the first-level warning: the blood volume is lower than 3 working days.

[0074] The blood inventory warning method based on deep learning proposed in Example 1 of the present invention uses deep learning technology to analyze the information of similar hospitals and the blood usage of patients in the hospital in the past, and infers the reasonable blood inventory. It solves the problem that hospitals cannot accurately and reasonably set inventory warning quantities, avoids the occurrence of high blood scrapping rates due to insufficient blood reserves and excessive blood reserves, and is of great significance to alleviating the regional blood supply and demand relationship.

[0075] Example 2

[0076] Based on the deep learning method in the blood inventory early warning method based on deep learning proposed in Example 1 of the present invention, a deep learning system in the blood inventory early warning method based on deep learning is also proposed, such as Figure 5 This is a schematic diagram of a blood inventory early warning system based on deep learning according to Embodiment 2 of the present invention. The system includes a conversion module, a simplification module and a learning module;

[0077] The conversion module is used to convert the blood usage information of the target hospital into a vector and store it in the first database; and convert the blood usage information of hospitals of the same category as the target hospital into a vector and store it in the second database;

[0078] The reduction module is used to reduce the attributes of the first database and the second database based on the rough set theory to obtain a reduced information database; the data in the reduced information database is divided into a training set, a validation set and a test set according to a preset ratio;

[0079] The learning module is used to input the blood usage information of the target hospital and the same type of hospitals in the second preset time period into the LSTM model, train the LSTM model, and verify and select the model with the smallest total loss on the validation set, predict the blood inventory through the test set test, and set blood classification warning according to the blood inventory.

[0080] In the conversion module, the selection criteria for hospitals of the same category as the target hospital are:

[0081] Obtaining several items of basic hospital information; the basic hospital information includes but is not limited to the region, hospital level, hospital category, and surgical volume;

[0082] The similarity value is set according to the similarity between each basic information and the corresponding basic information of the target hospital; the weight of each hospital's basic information is set to obtain the relevance to the target hospital.

[0083] Convert the blood usage information of the target hospital f in the first preset time period t into a vector F t And stored in the first database A.

[0084] For each target hospital of the same category as the hospital p i (0≤i≤n), convert the blood usage information of the same type of hospital in the second preset time period d into a vector And stored in information database B; where n is the number of hospitals of the same category as the target hospital.

[0085] The process executed by the reduction module is: setting conditional attributes and decision attributes; the decision attributes are the blood usage of each blood type; the attributes other than the decision attributes are conditional attributes;

[0086] The importance of decision attributes is calculated by using rough set dependency, and the attribute relevance is calculated by using mutual information, and the first core attribute set a of the first database A is obtained as follows: a = {a1, ..., a n} and the second core attribute set b of the second database B = {b1, ..., b n};

[0087] Calculate the first core attribute set a = {a1, ..., a n} and the second core attribute set b={b1,....,b n} to obtain the third core attribute set c = {c1, ..., c l};

[0088] From the first database A, a third core attribute set c={c1, ..., c l} data constitute the reduced information base A'.

[0089] The data in the information database A' is used as the training set, verification set, and test set according to 70%, 15%, and 15% of the information vectors, respectively.

[0090] The learning module is executed as follows: the blood usage information of the target hospital and the same type of hospitals in the second preset time period d is collected. Input into the LSTM model, take the minimum loss function as the optimization goal, use the training set to train the LSTM model with a preset number of times as a cycle, and obtain the blood use information vector within the d+1 date As predictive blood use information;

[0091] The actual blood usage of each blood type on date d+1 is analyzed with the predicted blood usage of each blood type, and the prediction error rate vector of each blood type is calculated. Total error rate As the loss function of the LSTM model; the LSTM model is verified for a preset number of times using the verification set, and the model with the smallest total loss function is tested in the test set.

[0092] The test set is used to test and obtain the sum of loss functions, and whether the sum of loss functions meets the preset requirements is determined, and the model that meets the preset requirements is set as the final prediction model.

[0093] The final prediction model is used to obtain the blood usage of each blood category in the preset future days, and blood classification warnings are set based on the blood inventory.

[0094] The deep learning system in the deep learning-based blood inventory warning method proposed in Example 2 of the present invention uses deep learning technology to analyze information of similar hospitals and previous blood usage of patients in the hospital, and infers a reasonable blood inventory. This solves the problem that hospitals cannot accurately and reasonably set inventory warning quantities, avoids the occurrence of high blood scrapping rates due to insufficient or excessive blood reserves, and is of great significance to alleviating the regional blood supply and demand relationship.

[0095] 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 that there is any such actual relationship or order between these entities or operations. Moreover, the term "include", "comprise" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or equipment that includes a series of elements are inherent to the elements. In the absence of more restrictions, the elements limited by the sentence "comprise one..." do not exclude the presence of other identical elements in the process, method, article or equipment that includes the elements. In addition, the above-mentioned technical solution provided in the embodiment of the present application is consistent with the corresponding technical solution in the prior art in principle, and the part is not described in detail, so as not to repeat too much.

[0096] Although the above describes the specific implementation of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. For those skilled in the art, other different forms of modifications or deformations can be made on the basis of the above description. It is not necessary and impossible to list all the implementation methods here. On the basis of the technical solution of the present invention, various modifications or deformations that can be made by those skilled in the art without creative work are still within the scope of protection of the present invention.

Claims

1. A blood inventory early warning method based on deep learning, characterized in that: The following steps are involved: Convert the blood usage information of the target hospital into a vector and store it in the first database; and converting the blood usage information of the target hospital to a vector and storing it in the second database; the converting the blood usage information of the target hospital to a vector and storing it in the first database specifically includes: converting the blood usage information of the target hospital f in the first preset time period t into a vector F t and store it in the first database A; The step of converting the blood use information of hospitals of the same category as the target hospital into a vector and storing it in the second database specifically includes: for each hospital of the same category as the target hospital p i (0≤i≤n), convert the blood usage information of the same type of hospital in the second preset time period d into a vector and stored in the information database B; where n is the number of hospitals of the same category as the target hospital; Based on the rough set theory, the attributes of the first database and the second database are simplified to obtain a simplified information database; the data in the simplified information database is divided into a training set, a validation set and a test set according to a preset ratio; the process of simplifying the attributes of the first database and the second database based on the rough set theory is: setting conditional attributes and decision attributes; the decision attributes are the blood volume of each blood type; the attributes other than the decision attributes are conditional attributes; the importance of the decision attributes is calculated by using the rough set dependency, and the attribute correlation is calculated by using the mutual information, and the first core attribute set a of the first database A is obtained respectively. n } and the second core attribute set b of the second database B = {b1, ..., b n }; Calculate the first core attribute set a = {a1, ...., a n } and the second core attribute set b={b1,....,b n } to obtain the third core attribute set c = {c1, ..., c l }; Take out the third core attribute set c={c1, ..., c l The data of} constitutes the reduced information base A'; The blood usage information of hospitals of the same category as the target hospital in the second preset time period is input into the LSTM model, the LSTM model is trained, and the model with the smallest total loss is selected by verification on the validation set, and the blood inventory is predicted by the test set test; the blood usage information of hospitals of the same category as the target hospital in the second preset time period is input into the LSTM model, the LSTM model is trained, and the model with the smallest total loss is selected by verification on the validation set. The process is: The blood usage information of the target hospital and the same type of hospitals in the second preset time period d is Input into the LSTM model, take the minimum loss function as the optimization goal, use the training set to train the LSTM model with a preset number of times as a cycle, and obtain the blood use information vector within the d+1 date As predictive blood use information; The actual blood usage of each blood type on date d+1 is analyzed with the predicted blood usage of each blood type, and the prediction error rate vector of each blood type is calculated. Total error rate As the loss function of the LSTM model; the LSTM model is verified for a preset number of times using the verification set, and the model with the smallest total loss function is tested in the test set.

2. The blood inventory early warning method based on deep learning according to claim 1 is characterized in that: The selection criteria for hospitals of the same category as the target hospital are as follows: Obtaining several items of basic hospital information; the basic hospital information includes but is not limited to the region, hospital level, hospital category, and surgical volume; Set a similarity value based on the similarity between each basic information and the corresponding basic information of the target hospital; Set the weight of each hospital's basic information to obtain its relevance to the target hospital.

3. The blood inventory early warning method based on deep learning according to claim 1 is characterized in that: The process of dividing the data in the simplified information database into a training set, a validation set and a test set according to a preset ratio specifically includes: The data in the information base A' are used as a training set, a validation set, and a test set according to 70%, 15%, and 15% of the information vectors, respectively.

4. The blood inventory early warning method based on deep learning according to claim 1 is characterized in that: The process of predicting blood inventory by testing the test set includes: The test set is used to test and obtain the sum of loss functions, and whether the sum of loss functions meets the preset requirements is determined, and the model that meets the preset requirements is set as the final prediction model.

5. The blood inventory early warning method based on deep learning according to claim 4 is characterized in that: The method also includes obtaining the blood usage of each blood category in a preset number of days in the future according to the final prediction model, and setting a blood classification warning according to the blood inventory.

6. A blood inventory early warning system based on deep learning, used to execute the blood inventory early warning method based on deep learning according to any one of claims 1 to 5, characterized in that: The system includes a transformation module, a simplification module and a learning module; The conversion module is used to convert the blood usage information of the target hospital into a vector and store it in the first database; and convert the blood usage information of hospitals of the same category as the target hospital into a vector and store it in the second database; The reduction module is used to reduce the attributes of the first database and the second database based on the rough set theory to obtain a reduced information database; the data in the reduced information database is divided into a training set, a validation set and a test set according to a preset ratio; The learning module is used to input the blood usage information of the target hospital and the same type of hospitals in the second preset time period into the LSTM model, train the LSTM model, and verify and select the model with the smallest total loss on the validation set, predict the blood inventory through the test set test, and set a blood classification warning based on the blood inventory.

Images