A drug classification management system based on deep forest multi-label classification learning model

Abstract

The present invention discloses a drug classification management system based on a deep forest multi-label classification learning model. The data acquisition module is used to collect drug feature data and perform normalization preprocessing on the drug feature data to obtain normalized drug feature data; the data analysis module is used to analyze the pre-processed drug feature data; the label classification module is used to provide drug label classification services, and classify drugs with different data features based on the data features of the drugs; the environmental monitoring module is used to monitor the warehouse where the drugs are stored and provide real-time environmental data to facilitate the storage and management of the drugs; the data visualization module is used to provide a visual interface to display drug label classification data to help drug administrators obtain drug information in real time. The present invention performs multi-label classification on drugs to achieve hierarchical management of drugs and efficient drug information retrieval.

Description

Technical field:

[0001] The present invention belongs to the technical field of drug classification management, and in particular relates to a drug classification management system based on a deep forest multi-label classification learning model. Background technology:

[0002] In recent years, with the continuous development of medical science and technology, the number of medical drug categories has increased year by year, and the classification and management of medical drugs has become a key concern for major medical institutions. Traditional manual classification methods require the training of specialized personnel and consume a large amount of manpower and material resources.

[0003] Patent ZL201810407511.3 discloses a "drug classification management data analysis system." First, a partitioning module is used to divide a designated area into several zones, and the number of pharmacies within each rectangular zone is obtained. Then, a drug sales statistics module is used to count the drug sales of each pharmacy within each rectangular zone, to determine the number of drugs corresponding to each type of drug sold. A temperature prediction and comparison module is then used to predict the future outdoor environment of each rectangular zone. Finally, a cloud server receives data from the drug statistics module and the temperature prediction and comparison module, and displays the data for each rectangular zone on the display. However, this system does not consider detailed multi-label classification of drugs.

[0004] Patent KR1020210189242 discloses a "drug classification system using big data," which classifies drugs according to preset criteria based on input information from a server; recommends a drug as a hospital device or associated pharmacy device based on the drug information input to the classification server; determines the classification criteria for type and dosage information based on the drug's efficacy; and completes the classification of the drug according to the classification server. However, this system does not consider the impact of the drug's storage environment on the drug.

[0005] Therefore, it is very necessary to design a drug classification management system based on a deep forest multi-label classification learning model (Multi Label Classification with Deep Forest, referred to as MLCDForest) to classify drugs according to their characteristics and storage environment, to help hospitals and pharmacies manage drugs rationally, reduce management costs, and improve management efficiency.

[0006] The information disclosed in this background technology section is only intended to enhance understanding of the overall background of the invention and should not be regarded as an admission or any form of suggestion that the information constitutes the prior art already known to a person skilled in the art. Summary of the invention:

[0007] The purpose of the present invention is to provide a drug classification management system based on a deep forest multi-label classification learning model, which manages drugs according to their characteristics, facilitates hospitals or pharmacies to rationally purchase drugs, improves the rationality of drug supply, avoids the occurrence of unsold or out-of-stock drugs, and thus overcomes the defects in the above-mentioned prior art.

[0008] To achieve the above-mentioned objectives, the present invention provides a drug classification management system based on a deep forest multi-label classification learning model, comprising a data acquisition module, a data analysis module, a label division module, an environmental monitoring module, and a data visualization module that are communicatively connected in sequence; the data acquisition module is used to collect the characteristic data of the drug and perform normalization preprocessing on the drug characteristic data to obtain normalized drug characteristic data; the data analysis module is used to analyze the characteristic data after preprocessing of the drug; the label division module is used to provide drug label classification services, and classify drugs with different data characteristics based on the data characteristics of the drug; the environmental monitoring module is used to monitor the warehouse where the drugs are stored and provide real-time environmental data to facilitate the storage and management of the drugs; the data visualization module is used to provide a visual interface to display the drug label classification data, helping drug administrators to obtain drug information in real time.

[0009] Preferably, in the technical solution, the characteristic data of the drug include the drug's production date, storage environment, drug form, applicable population, and main therapeutic function.

[0010] A method for using a drug classification management system based on a deep forest multi-label classification learning model, comprising the following steps:

[0011] S01. Collect characteristic data of drugs, including drug function, drug production season, drug storage environment, drug form, and drug usage rate, and perform normalization preprocessing on these data to obtain normalized data;

[0012] S02. Extract the characteristic attributes of the drug and convert them into a one-dimensional vector, i.e., an embedding vector. Then, based on the different characteristic attributes of the drug, use the drug characteristic attributes as data labels and convert them into a one-hot array consisting of 0 and 1. The embedding vector can be marked as "1" under each drug feature and "0" under other irrelevant features. Each vector can be used as a label for a certain drug. A one-hot vector is a one-dimensional vector, where the drug vector based on one-hot can be expressed as:

[0013]

[0014] Among them ω1, ω2, ω3...ω nIt is a one-dimensional vector representing the one-hot label corresponding to each drug. One-hot labels only have two numbers, 0 and 1, which can simplify the information of drugs.

[0015] S03. In the basic information of the multi-label classification dataset (X, Y) of drugs, X is the drug attribute, Y is the label, the number of samples is n, and the total number of labels is m; the given label space is Y = {Y1, Y2, ..., Y m ), a sample xi with k drug features is assigned a value from a subset y in the label space Y; the correlation between two different labels is evaluated by the Cramer correlation coefficient, and the degree of imbalance between different labels can be measured by the imbalance ratio; the overall imbalance of the labels is measured by MeanIR, and the significance of MeanIR is measured by the standard coefficient of variation CV; the Cramer correlation coefficient formula is:

[0016]

[0017] Where x is the drug attribute, It is represented as the mean of the drug attributes, y represents the label, Expressed as the mean of the label attributes;

[0018] The probability mean of the drug label MeanIR is:

[0019]

[0020] Where IRLbl(y) represents the probability that label y is the category of the drug;

[0021] The expression of the standard coefficient of variation CV is as follows:

[0022]

[0023] IRLbl(y) is calculated as follows:

[0024]

[0025] where labelset i is the label set, h(y, labelset i ) is the one-hot code of the drug label;

[0026] h(y, labelset i ) is calculated as follows:

[0027]

[0028] S04. Perform a sliding window-based multi-granularity scan on the drug data to generate low-dimensional drug feature vectors, which are then combined into different labels. Cascade random forests are then used to learn the generated drug feature vectors. The cascade random forest is used to make probability predictions for each label and concatenate them onto the original features. The class is selected based on the maximum aggregate value to obtain the drug label classification result.

[0029] S05. The classified drugs are placed in corresponding warehouses according to their storage characteristics. An integrated chip that measures temperature, humidity, and light intensity with high precision is used to monitor the drug environment in real time, and this data is then incorporated into the drug labels.

[0030] S06. The drug data generated in the above steps are visualized through the data visualization module to realize the visualization of drug classification management. The visualization interface displays the feature label category corresponding to each type of drug, the real-time environmental data of the drug, and the quantity of the drug.

[0031] Preferably, in the technical solution, the overall imbalance of the labels in step S03 is measured using SCUMBLE, and the formula for the sparsity of drug labels SCUMBLE is as follows:

[0032]

[0033] IRLbl i represents the probability of the i-th drug label, and h is the element in the one-hot array of drug labels.

[0034] Preferably, in the technical solution, in steps S05 and S06, the environmental data include ambient temperature, humidity, and light intensity.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] Based on the Deep Forest multi-label classification learning model, it can more accurately predict drug classification, improving the accuracy of drug classification while efficiently utilizing limited warehouse space. This improves the rationality of drug supply and avoids drug unsold or out-of-stock situations. Description of the drawings:

[0037] Figure 1 This is a schematic diagram of the structure of the drug classification management system based on the deep forest multi-label classification learning model of the present invention;

[0038] Figure 2 This is a flowchart of the implementation of the drug classification management system based on the deep forest multi-label classification learning model of the present invention;

[0039] Figure 3This is a schematic diagram of the distribution of environmental monitoring warehouses in the present invention;

[0040] Figure 4 It is a flow chart of drug classification of the present invention. Specific implementation method:

[0041] The specific embodiments of the present invention are described in detail below, but it should be understood that the protection scope of the present invention is not limited by the specific embodiments.

[0042] Unless expressly stated otherwise, throughout the specification and claims, the term "comprise" or variations such as "include" or "comprising", etc., will be understood to include the stated elements or components but not to exclude other elements or other components.

[0043] like Figure 1 As shown, the present invention provides a drug classification management system based on a deep forest multi-label classification learning model, comprising a data acquisition module, a data analysis module, a label division module, an environmental monitoring module, and a data visualization module that are communicatively connected in sequence; the data acquisition module is used to collect characteristic data of drugs including the production date, storage environment, drug form, applicable population, and main function of the drugs, and perform normalization preprocessing on the drug characteristic data to obtain normalized drug characteristic data; the data analysis module is used to analyze the characteristic data after preprocessing of the drugs; the label division module is used to provide drug label classification services, and classify drugs with different data characteristics based on the data characteristics of the drugs; the environmental monitoring module is used to monitor the warehouse where the drugs are stored and provide real-time environmental data to facilitate the storage and management of the drugs; the data visualization module is used to provide a visual interface to display the drug label classification data to help drug administrators obtain drug information in real time.

[0044] like Figure 2 As shown, a method for using a drug classification management system based on a deep forest multi-label classification learning model includes the following steps:

[0045] S01. Collect characteristic data of drugs, including drug function, drug production season, drug storage environment, drug form, and drug usage rate, and perform normalization preprocessing on these data to obtain normalized data;

[0046] S02. Extract the characteristic attributes of the drug and convert them into a one-dimensional vector, i.e., an embedding vector. Then, based on the different characteristic attributes of the drug, use the drug characteristic attributes as data labels and convert them into a one-hot array consisting of 0 and 1. The embedding vector can be marked as "1" under each drug feature and "0" under other irrelevant features. Each vector can be used as a label for a certain drug. A one-hot vector is a one-dimensional vector, where the drug vector based on one-hot can be expressed as:

[0047]

[0048] Among them ω1, ω2, ω3...ω n It is a one-dimensional vector representing the one-hot label corresponding to each drug. One-hot labels only have two numbers, 0 and 1, which can simplify the information of drugs.

[0049] S03. In the basic information of the multi-label classification dataset (X, Y) of drugs, X is the drug attribute, Y is the label, the number of samples is n, and the total number of labels is m; the given label space is Y = {Y1, Y2, ..., Y m}, a sample xi with k drug features is assigned a value from a subset y in the label space Y; the correlation between two different labels is evaluated by the Cramer correlation coefficient, and the degree of imbalance between different labels can be measured by the imbalance ratio; the overall imbalance of the labels is measured by MeanIR or SCUMBLE, and the significance of MeanIR is measured by the standard coefficient of variation CV; the Cramer correlation coefficient formula is:

[0050]

[0051] Where x is the drug attribute, It is represented as the mean of the drug attributes, y represents the label, Expressed as the mean of the label attributes;

[0052] The probability mean of the drug label MeanIR is:

[0053]

[0054] Where IRLbl(y) represents the probability that label y is the category of the drug;

[0055] The expression of the standard coefficient of variation CV is as follows:

[0056]

[0057] IRLbl(y) is calculated as follows:

[0058]

[0059] where labelset i is the label set, h(y, labelset i ) is the one-hot code of the drug label;

[0060] h(y, labelset i ) is calculated as follows:

[0061]

[0062] The formula for the sparsity SCUMBLE of drug labels is as follows:

[0063]

[0064] IRLbl i represents the probability of the i-th drug label, and h is the element in the one-hot array of drug labels;

[0065] S04. Perform a sliding window-based multi-granularity scan on the drug data to generate low-dimensional drug feature vectors, which are then combined into different labels. Cascade random forests are then used to learn the generated drug feature vectors. The cascade random forest is used to make probability predictions for each label and concatenate them onto the original features. The class is selected based on the maximum aggregate value to obtain the drug label classification result.

[0066] S05. Store the classified medicines in corresponding warehouses according to their storage characteristics, such as Figure 3 As shown, the drug warehouse is divided into a light-shielding room, a freezing room, and a normal temperature room. An integrated chip that measures temperature, humidity, and light intensity with high precision is used to detect the environmental data of the drugs in real time. The environmental data includes ambient temperature, humidity, and light intensity, and these data are spliced ​​on the drug label.

[0067] S06. The drug data generated in the above steps are visualized through the data visualization module to realize the visualization of drug classification management. The visualization interface displays the feature label category corresponding to each type of drug, the real-time environmental data of the drug, and the quantity of the drug.

[0068] Figure 4As shown, drugs are classified according to their properties and the environment in which they are located; for example, erythromycin ointment can be used for small-area burns, acne vulgaris, impetigo and other purulent skin diseases. The standard storage environment is sealed in a cool and dry place (i.e., not exceeding 20°C), the actual storage temperature is 18°C, the production date is June 1, 2022, and the remaining inventory is 2,000 bottles. Then the characteristic labels of the drug are prescription drugs, surgical drugs, burn drugs, skin drugs, summer of 2022, ointment, general usage rate, and cool. Dry, 18℃, 2000 bottles in stock; dextran, used for anti-shock, prevention and treatment of cerebral thrombosis and thromboangiitis obliterans, used by intravenous drip or injection, the standard storage environment is sealed below 25℃, the actual storage temperature is 20℃, the production date is March 5, 2022, and there are 300 bottles left in stock. The characteristic labels of this drug are over-the-counter drugs, internal medicine drugs, neurological drugs, spring 2022, liquid, high usage rate, sealed below 25℃, 20℃, and 300 bottles in stock.

[0069] By labeling and classifying the characteristic data of drugs, including drug functions, drug production season, drug storage environment, drug form, drug usage rate and other data, and monitoring the temperature and light of the drug storage environment in real time, the reliability of drug management is improved and the rational management of drug supply is achieved.

[0070] The foregoing descriptions of specific exemplary embodiments of the present invention are for purposes of illustration and description. These descriptions are not intended to limit the invention to the precise forms disclosed, and it is apparent that many variations and modifications are possible in light of the foregoing teachings. The exemplary embodiments have been selected and described for the purpose of explaining the specific principles of the invention and their practical application, thereby enabling those skilled in the art to realize and utilize a variety of exemplary embodiments of the invention and various options and modifications. The scope of the invention is intended to be defined by the claims and their equivalents.

Claims

1. A method for using a drug classification management system based on a deep forest multi-label classification learning model, characterized by: The drug classification management system includes a data acquisition module, a data analysis module, a label classification module, an environmental monitoring module, and a data visualization module that are sequentially connected in communication; the data acquisition module is used to collect drug feature data and perform normalization preprocessing on the drug feature data to obtain normalized drug feature data; The data analysis module is used to analyze the characteristic data of the pre-processed drugs; the label classification module is used to provide drug label classification services, and classify drugs with different data characteristics based on the data characteristics of the drugs; the environmental monitoring module is used to monitor the warehouse where the drugs are stored and provide real-time environmental data to facilitate the storage and management of drugs; the data visualization module is used to provide a visual interface to display the drug label classification data, helping drug administrators to obtain drug information in real time; the characteristic data of the drugs include the production date, storage environment, drug form, applicable population, and main function of the drugs; The steps are: S01. Collect characteristic data of drugs, including drug function, drug production season, drug storage environment, drug form, and drug usage rate, and perform normalization preprocessing on these data to obtain normalized data; S02. Extract the characteristic attributes of the drug and convert them into a one-dimensional vector, i.e., an embedding vector. Then, based on the different characteristic attributes of the drug, use the drug characteristic attributes as data labels and convert them into a one-hot array consisting of 0s and 1s. S03. In the basic information of the multi-label classification dataset (X, Y) of drugs, X is the drug attribute, Y is the label, the number of samples is n, and the total number of labels is m; the given label space is Y = {Y1, Y2, ..., Y m A sample xi with k drug features is assigned a value from a subset y in the label space Y. The correlation between two different labels is evaluated using the Cramer correlation coefficient, and the degree of imbalance between different labels can be measured by the imbalance ratio. The overall imbalance of labels is measured using MeanIR, which represents the mean probability of drug labels. The significance of MeanIR is measured using the standard coefficient of variation (CV). S04. Perform a multi-granularity scan of the drug data based on a sliding window to generate low-dimensional drug feature vectors, and then combine these feature vectors into different labels. Then, use the cascade random forest to train the cascade random forest model using the generated drug feature vector dataset. Use the trained cascade random forest model to make a probability prediction for each label and splice it onto the original features. Select the class based on the maximum aggregate value to obtain the drug label classification result. S05. The classified drugs are placed in corresponding warehouses according to their storage characteristics for storage. An integrated chip that measures temperature, humidity, and light intensity with high precision is used to monitor the drug environment in real time, and this data is then incorporated into the drug labels. S06. The drug data generated in the above steps are visualized through the data visualization module to realize the visualization of drug classification management. The visualization interface displays the feature label category corresponding to each type of drug, the real-time environmental data of the drug, and the quantity of the drug.

2. The method for using the drug classification management system based on the deep forest multi-label classification learning model according to claim 1 is characterized by: The Cramer correlation coefficient formula is: Where x is the drug attribute, It is represented as the mean of the drug attributes, y represents the label, Expressed as the mean of the label attributes.

3. The method for using the drug classification management system based on the deep forest multi-label classification learning model according to claim 1 is characterized by: The probability mean of the drug label MeanIR is: Where IRLbl(y) represents the probability that label y is the category of the drug.

4. The method for using the drug classification management system based on the deep forest multi-label classification learning model according to claim 1 is characterized by: The expression of the standard coefficient of variation CV is as follows:

5. The method for using the drug classification management system based on the deep forest multi-label classification learning model according to claim 3 or 4, characterized in that: IRLbl(y) is calculated as follows: where labelset i is the label set, h(y,labelset i ) is the one-hot code of the drug label; h(y,labelset i ) is calculated as follows:

6. The method for using the drug classification management system based on the deep forest multi-label classification learning model according to claim 1 is characterized by: The overall imbalance of labels in step S03 is measured using SCUMBLE, which represents the sparsity of drug labels. The formula for SCUMBLE is as follows: IRLbl i represents the probability of the i-th drug label, and h is the element in the one-hot array of drug labels.

7. The method for using the drug classification management system based on the deep forest multi-label classification learning model according to claim 1 is characterized by: In steps S05 and S06, the environmental data includes environmental temperature, humidity, and light intensity.

Images