Automatic early warning system for abnormal sales volume of medicines

Abstract

The invention provides an automatic early warning system for abnormal sales volume of medicines, and the system comprises a data management module which collects and processes patient business data at regular time, and obtains a bottom data pool; the manual management and control module selects a management and control object, configures a medicine sales volume early warning rule, judges whether the management and control object meets an early warning condition or not, and automatically sends early warning information if the judgment result is yes; the sales volume abnormity judgment module is used for extracting abnormal data in the bottom layer data pool, constructing a sales volume abnormity prediction model and carrying out real-time sales volume monitoring on medicines to obtain a sales volume abnormity trend; and the sales volume prediction module extracts sales volume influence factors in the underlying data pool, constructs a sales volume prediction model, and carries out real-time sales volume prediction on drugs to obtain a predicted purchase volume and a predicted purchase opportunity. According to the invention, the transaction trend of the drug sales volume can be effectively identified, the drug sales trend can be predicted, personalized drug abnormal sales volume automatic early warning can be realized, the drug use and consumption conditions can be monitored, and drugs with abnormal sales volume can be timely found and dealt with.

Description

Technical Field

[0001] The present invention relates to the technical field of drug management, and in particular to an automatic early warning system for abnormal drug sales. Background Art

[0002] Typically, a hospital's drug inventory system is equipped with a basic early warning function that reminds pharmacy managers to restock when drugs are about to run out, or to increase stock during the peak flu season. However, in the face of sudden disease conditions in clinical practice, such as the sudden outbreak of new influenza such as COVID-19, SARS, and H1N1, as well as sudden events such as environmental pollution, natural disasters, and major accidents, the demand for drugs may rise sharply. In these cases, traditional inventory early warning systems cannot predict and warn of the sharp increase in drug consumption in a timely manner, nor can they autonomously learn and adapt to complex drug consumption patterns, resulting in drug shortages. This not only wastes the precious time of doctors and patients, seriously affects patients' medical experience, but may also have an adverse impact on the hospital's operations and management. Summary of the invention

[0003] In order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide an automatic early warning system for abnormal drug sales, which can effectively identify abnormal trends in drug sales, predict drug sales trends, and realize multi-algorithm, personalized automatic early warning of abnormal hospital drug sales. The system also allows manual setting of rules to accurately monitor drug use and consumption in a variety of ways, and promptly discover and respond to drugs with abnormal sales.

[0004] To achieve the above object, the present invention provides the following solution: an automatic early warning system for abnormal drug sales, comprising:

[0005] The data management module is used to collect patient business data from multiple hospitals on a regular basis every day, and perform quality inspection, data regularization and data standardization operations on the patient business data to obtain the underlying data pool;

[0006] The manual control module is used to select a control object, configure a drug sales warning rule according to the control object, and then determine whether the control object meets the warning condition according to the drug sales warning rule. If the judgment result is yes, the warning information is automatically sent;

[0007] The sales anomaly judgment module is used to extract the abnormal data in the underlying data pool, mark and separate the abnormal data to obtain abnormal judgment data, use the abnormal judgment data to perform machine learning modeling to obtain a sales anomaly prediction model, and then use the sales anomaly prediction model to monitor the real-time sales of drugs to obtain the sales anomaly trend;

[0008] The sales forecasting module is used to extract the sales influencing factors from the underlying data pool, perform data cleaning, feature selection and optimization on the sales influencing factors, and then use the sales influencing factors to perform machine learning modeling to obtain a sales forecasting model, and then use the sales forecasting model to perform real-time sales forecasting of drugs to obtain predicted purchase volume and predicted purchase timing;

[0009] Among them, the data management module, the manual control module, the sales abnormality judgment module and the sales forecasting module are interconnected.

[0010] Optionally, the data management module includes:

[0011] The data collection unit is used to collect patient business data in multiple hospital inpatient management systems on a daily basis using view technology, stored procedure technology or Webservice technology, record the daily collection results, perform incremental updates and data synchronization based on the collection results, and determine whether there is erroneous data in the historical collection data. If the determination result is yes, re-synchronize the data and remind of data synchronization anomalies;

[0012] A data processing unit, used to automatically detect the patient business data using a data quality detection tool, perform data regularization and preprocessing on the patient business data, and standardize the drug-related data in the patient business data to obtain an underlying data pool; wherein the drug-related data includes a generic name of a drug and a pharmacological class of a drug;

[0013] The data storage unit is used to select a data warehouse and store the underlying data pool in the data warehouse by using a column storage method.

[0014] Optionally, the manual control module includes:

[0015] A rule-making unit is used to select a control object, configure a control dimension, a control cycle, a warning threshold, and a warning sending object according to the control object, and obtain a drug sales warning rule; wherein each control dimension and each control cycle are respectively configured with different judgment and warning rules;

[0016] A data comparison unit, configured to determine whether the controlled object meets the warning condition according to the drug sales warning rule, and if the determination result is yes, send the relevant data and the warning result in the controlled object to the warning sending object;

[0017] The warning sending unit is used to view the warning results using a hyperlink, and the warning results are compared and displayed in a combination of tables and charts.

[0018] Optionally, the control objects include all drugs, specific drugs, generic names of drugs and pharmacological classes; the control dimensions include the entire hospital, departments and major departments; the control cycles include daily statistics, natural month statistics, non-natural month statistics and weekly statistics; the warning thresholds include the maximum amount of the previous N cycles, the average amount of the previous N cycles and a fixed value.

[0019] Optionally, the sales abnormality judgment module includes:

[0020] System rule unit, used to build automatically generated system rules;

[0021] An abnormal data unit is used to extract abnormal data from the underlying data pool according to the system rules and the drug sales warning rules, establish statistical reports based on the abnormal data, perform preliminary marking of abnormal conditions of the control objects, obtain initial abnormal results, and then manually screen the initial abnormal results to obtain abnormal judgment data;

[0022] a first model building unit, configured to select a first machine learning model, divide the abnormality judgment data into a first experimental group and a first reference group, train the machine learning model using a machine learning algorithm and the first experimental group, and perform performance evaluation, parameter adjustment, and model optimization using the first reference group during the training process to obtain a sales abnormality prediction model;

[0023] Abnormal sending unit: used to use the sales abnormality prediction model to monitor the real-time sales of drugs, obtain the abnormal sales trend, generate abnormal sales information based on the abnormal sales trend and send it to the corresponding staff.

[0024] Optionally, the process of constructing the statistical report includes: according to the control cycle, counting the historical sales amount, amount growth rate, sales volume, average sales amount, multi-period sales amount of the controlled object, and the number of patients, number of prescriptions and number of related diagnoses in the same period of the controlled object to complete the construction of the statistical report.

[0025] Optionally, the sales volume forecasting module includes:

[0026] An influencing factor extraction unit, used to establish a modeling database including basic information, consumption data, replenishment cycle and other factors according to the underlying data pool, obtain sales influencing factors, and perform regression to fill in missing data in the sales influencing factors;

[0027] A feature selection unit is used to estimate and match the propensity score of the sales influencing factors using the pandas library in Python, statsmodels, and imblearn, and then perform a balance check and causal effect analysis on the matched indexes and data to obtain the correlation coefficient between the sales influencing factors and the results;

[0028] a second model building unit, configured to select a second machine learning model, divide the sales influencing factors into a second experimental group and a second reference group, train the machine learning model using a machine learning algorithm and the second experimental group, and perform performance evaluation, parameter adjustment, and model optimization using the second reference group during the training process to obtain a sales prediction model;

[0029] The early warning sending unit is used to use the sales forecasting model to perform real-time sales forecasting of drugs, obtain the predicted purchase volume and predicted purchase timing, and send the predicted purchase volume and predicted purchase timing to corresponding staff.

[0030] Optionally, the basic information includes name, dosage form, specification, manufacturer, code, drug type, generic name of drug and pharmacological category, the consumption data includes the charging time, sales volume and sales trend of the controlled object, the replenishment cycle includes the warehousing time, outbound time, inventory and replenishment cycle interval of the controlled object, and the other factors include seasonal marks, holiday marks, number of patients and high-frequency diagnoses.

[0031] The present invention discloses the following technical effects by providing an automatic early warning system for abnormal drug sales:

[0032] 1. Early warning and advance filing: Traditional inventory management relies heavily on inventory volume, and managers will use this to evaluate the storage situation of drugs. Based on inventory warning, the present invention adds an abnormal sales warning function. This function uses consumption as a benchmark indicator, and can keenly capture and analyze changes in the rate of drug consumption, and identify drug consumption trends early. When the machine learning model detects a trend of abnormal sales growth, it will immediately warn pharmacy staff, allowing them to take action in advance, strengthen filing and preparation work, and improve the efficiency and response speed of medical services in the actual operation of the hospital. Drugs can be used quickly in emergencies to avoid delays in treatment due to waiting for the pharmacy to dispense medicines.

[0033] 2. Self-updating early warning standards to optimize the working mode of pharmacy staff: It combines the results of manual control and the data of the system's built-in rules, and uses machine learning algorithms to self-iterate and optimize. This means that the system can learn and identify patterns from historical data, and automatically adjust rules to adapt to changing environments and needs without manual intervention by pharmacy staff. The system has changed from passive warnings based on rule parameters to active self-updating of early warning standards. It not only has strong generalization capabilities, but also changes the working mode of pharmacy staff and reduces their workload.

[0034] 3. Comprehensiveness of drug consumption and inventory management: 1) The present invention not only allows users to manage inventory by manually setting rules, but also adopts two machine learning models, respectively for the judgment of abnormal trends in drug sales and sales forecasts. This multi-dimensional early warning mechanism enables the system to cover both early mutations and later replenishment needs, thereby improving the comprehensiveness of drug consumption and inventory management. 2) The present invention also takes into account that in actual hospital procurement, the same variety may purchase drugs of different specifications produced by multiple different manufacturers. At the same time, in clinical applications, a drug may also have multiple similar alternative drugs. Therefore, the system can not only control a single drug, but also control drugs in groups, such as grouping them according to the generic name or pharmacological class of the drug. This drug group-based control method makes sales abnormality judgment and inventory management more flexible.

[0035] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0037] Figure 1 A schematic diagram of a system architecture provided by an embodiment of the present invention;

[0038] Figure 2 A schematic diagram of a system judgment flow provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0039] 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.

[0040] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0041] like Figure 1-2 As shown, the present invention provides an automatic warning system for abnormal drug sales, including a data management module, a manual control module, a sales abnormality judgment module and a sales forecast module that are interconnected. Among them:

[0042] 1. Data management module

[0043] It is used to collect patient business data from multiple hospitals on a regular basis every day, and perform quality inspection, data regularization and data standardization operations on the patient business data to obtain the underlying data pool. The data management module includes:

[0044] 1.1 Data Acquisition Unit

[0045] It is used to collect patient business data on a daily basis in multiple hospital inpatient management systems using view technology, stored procedure technology or Webservice technology, record daily collection results, and perform real-time incremental updates based on the collection results.

[0046] The service model is adopted and a fault-tolerant mechanism is added to ensure the accuracy of the data. The data synchronization between this system and HIS (hospital inpatient management system) is performed on a daily basis, and the daily collection results are recorded to achieve data synchronization. It also determines whether there is erroneous data in the historical collection data. If the judgment result is yes, the data synchronization is performed again and the abnormal data synchronization reminder is given.

[0047] 1.2 Data Processing Unit

[0048] Used to automatically detect the accuracy and standardization of the patient business data using data quality detection tools, thereby improving data quality. Perform data regularization (such as normalizing date formats, processing illegal data, etc.) and preprocessing on the patient business data, and standardize the drug-related data in the patient business data (such as standardizing drug generic names, pharmacological class labels, etc.) to obtain the underlying data pool; wherein the drug-related data includes drug generic names and drug pharmacological classes;

[0049] 1.3 Data Storage Unit

[0050] It is used to select a data warehouse (such as the data warehouse ClickHouse) and store the underlying data pool in the data warehouse using column storage to improve the efficiency of large-scale data aggregation queries.

[0051] 2. Manual control module

[0052] It is used to select a control object, configure a drug sales warning rule according to the control object, and then determine whether the control object meets the warning condition according to the drug sales warning rule. If the judgment result is yes, the warning information is automatically sent. The manual control module includes:

[0053] 2.1 Rule-making unit

[0054] It is used to select a control object, configure the control dimension, control cycle, warning threshold and warning sending object according to the control object, and obtain the drug sales warning rule; wherein each control dimension and each control cycle are respectively configured with different judgment and warning rules.

[0055] 1) Control objects, including all drugs, specific drugs, generic names of drugs, and pharmacological classes;

[0056] 2) Configuration rules control dimensions, including the entire hospital, department, and large department;

[0057] 3) Configure the control cycle of the rules, including daily statistics, natural month / non-natural month statistics, and weekly statistics;

[0058] 4) Configure the warning threshold of the rule, including the maximum amount of the previous N periods, the average amount of the previous N periods, and the fixed value;

[0059] 5) Each control dimension and period can be configured with warning rules. Taking daily statistics as an example, the rules are as follows:

[0060] The increase of daily sales amount compared with the warning threshold is greater than YY%, and the daily sales amount is greater than XX yuan;

[0061] The average sales amount in the past M days has increased by more than YY% compared with the warning threshold, and the average daily sales amount is more than XX yuan;

[0062] The sales amount for M consecutive days increased by more than YY% compared with the warning threshold, and the average daily sales amount was more than XX yuan;

[0063] 6) Configure the warning sending object. When the warning conditions are triggered, the system automatically sends a warning.

[0064] 2.2 Data comparison unit

[0065] It is used to judge whether the controlled object meets the warning conditions according to the drug sales warning rules. If the judgment result is yes, the relevant data and warning results in the controlled object are sent to the warning sending object.

[0066] 2.3 Warning sending unit

[0067] When the warning conditions are met, it is sent to the warning object. After the warning information is sent, users can view the warning results through hyperlinks such as the system platform, SMS platform, and corporate WeChat platform. The warning results can be displayed in a combination of tables and charts. Click on the table data, and the chart will automatically generate a short-term result bar chart / line chart, which can clearly show the comparison between the warning items of the drug in the short term and the dynamic threshold.

[0068] 3. Sales abnormality judgment module

[0069] It is used to extract abnormal data from the underlying data pool, mark and separate the abnormal data, obtain abnormal judgment data, use the abnormal judgment data for machine learning modeling, obtain a sales abnormality prediction model, and then use the sales abnormality prediction model to monitor the real-time sales of drugs to obtain abnormal sales trends. The sales abnormality judgment module includes:

[0070] 3.1 System rule unit

[0071] Used to build automatically generated system rules. The system will be configured with some rules for judging abnormal data when it leaves the factory, and rules will also be configured for manual control. The abnormal data found by these two rules are combined and summarized into the data set required for modeling.

[0072] 3.2 Abnormal Data Unit

[0073] It is used to extract abnormal data in the underlying data pool according to the system rules and the drug sales warning rules, establish statistical reports based on the abnormal data, perform preliminary marking of abnormal situations of the controlled objects, obtain initial abnormal results, and then manually screen the initial abnormal results, further select valuable changes and separate data to obtain abnormal judgment data.

[0074] The process of constructing the statistical report includes: according to the control cycle, statistics on the historical sales amount, amount growth rate, sales volume, average sales amount, multi-period sales amount of the controlled object, and the number of patients, prescriptions and related diagnoses of the controlled object in the same period to complete the construction of the statistical report.

[0075] 3.3 The first model building unit

[0076] Used to select a first machine learning model, divide the abnormal judgment data into a first experimental group and a first reference group, use a machine learning algorithm (such as XGBoost / LightGBM / TabNet / SVM algorithm) and the first experimental group to train the machine learning model, and use the first reference group to perform performance evaluation, parameter adjustment and model optimization during the training process to obtain a sales anomaly prediction model.

[0077] Among them, the performance of the model can be judged by the mean square error, root mean square error, mean absolute error, F1 / F2 score, and ROC curve, the best performing model can be selected, and the model parameters can be adjusted according to the test results to optimize the prediction accuracy.

[0078] 3.4 Abnormal sending unit

[0079] It is used to use the sales abnormality prediction model to monitor the real-time sales of drugs, obtain the abnormal sales trend, generate abnormal sales information based on the abnormal sales trend and send it to the corresponding staff.

[0080] 4. Sales forecast module

[0081] It is used to extract the sales influencing factors in the underlying data pool, perform data cleaning, feature selection and optimization on the sales influencing factors, and then use the sales influencing factors to perform machine learning modeling to obtain a sales forecasting model, and then use the sales forecasting model to perform real-time sales forecasting of drugs to obtain predicted purchase volume and predicted purchase timing. The sales forecasting module includes:

[0082] 4.1 Influencing Factor Extraction Unit

[0083] It is used to establish a modeling database including basic information, consumption data, replenishment cycle and other factors according to the underlying data pool, obtain sales influencing factors, and perform regression to fill in the missing data in the sales influencing factors.

[0084] Basic information: name, dosage form, specification, manufacturer, code, drug type, generic name, pharmacological category, etc.;

[0085] Consumption data: charging time, sales volume, sales trend, etc. of drugs / generic drug names / pharmacological categories;

[0086] Replenishment cycle: the entry time, exit time, inventory quantity, replenishment cycle interval, etc. of the drug / generic name / pharmacological class;

[0087] Other factors: seasonal markers, holiday markers, number of patients, high-frequency diagnoses, etc.

[0088] Data cleaning: For missing data, we use the pandas library in Python combined with statsmodels and scikit-learn to perform regression filling, and then delete the data that still has missing data after filling.

[0089] 4.2 Feature Selection Unit

[0090] It is used to estimate and match the propensity scores of the sales influencing factors using the pandas library, statsmodels and imblearn in Python, and then perform balance check and causal effect analysis on the matched indexes and data to obtain the correlation coefficient between the sales influencing factors and the results; the correlation coefficient is used to adjust the weights of each factor in the machine learning training model.

[0091] 4.3 Second model building unit

[0092] Used to select a second machine learning model, divide the sales influencing factors into a second experimental group and a second reference group, use a machine learning algorithm (such as XGBoost / LightGBM / TabNet / SVM algorithm) and the second experimental group to train the machine learning model, and use the second reference group to perform performance evaluation, parameter adjustment and model optimization during the training process to obtain a sales prediction model.

[0093] Among them, the performance of the model can be judged by the mean square error, root mean square error and mean absolute error, the best performing model can be selected, and the model parameters can be adjusted according to the test results to optimize the prediction accuracy.

[0094] 4.4 Warning sending unit

[0095] It is used to use the sales forecasting model to perform real-time sales forecasting of drugs, obtain the predicted purchase volume and predicted purchase timing, and send the predicted purchase volume and predicted purchase timing to corresponding staff.

[0096] 5. Permission management

[0097] Permissions can be assigned to different accounts, and each role can have different permissions corresponding to different functions. This includes: creating and managing user accounts, assigning roles and permissions; permissions to enter each functional module of the program; permissions to set drug control rules; permissions to view warning information, etc.

[0098] Therefore, the present invention provides an automatic early warning system for abnormal drug sales, which can effectively identify abnormal trends in drug sales and predict drug sales trends, thereby realizing multi-algorithm, personalized automatic early warning of abnormal hospital drug sales. The system also allows manual setting of rules to accurately monitor drug usage and consumption in a variety of ways, and promptly discover and respond to drugs with abnormal sales.

[0099] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0100] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. An automatic early warning system for abnormal drug sales, characterized in that: include: The data management module is used to collect patient business data from multiple hospitals on a regular basis every day, and perform quality inspection, data regularization and data standardization operations on the patient business data to obtain the underlying data pool; The manual control module is used to select a control object, configure a drug sales warning rule according to the control object, and then determine whether the control object meets the warning condition according to the drug sales warning rule. If the judgment result is yes, the warning information is automatically sent; The sales anomaly judgment module is used to extract the abnormal data in the underlying data pool, mark and separate the abnormal data to obtain abnormal judgment data, use the abnormal judgment data to perform machine learning modeling to obtain a sales anomaly prediction model, and then use the sales anomaly prediction model to monitor the real-time sales of drugs to obtain the sales anomaly trend; The sales forecasting module is used to extract the sales influencing factors from the underlying data pool, perform data cleaning, feature selection and optimization on the sales influencing factors, and then use the sales influencing factors to perform machine learning modeling to obtain a sales forecasting model, and then use the sales forecasting model to perform real-time sales forecasting of drugs to obtain predicted purchase volume and predicted purchase timing; Among them, the data management module, the manual control module, the sales abnormality judgment module and the sales forecasting module are interconnected.

2. The automatic early warning system for abnormal drug sales according to claim 1 is characterized in that: The data management module comprises: The data collection unit is used to collect patient business data in multiple hospital inpatient management systems on a daily basis using view technology, stored procedure technology or Webservice technology, record the daily collection results, perform incremental updates and data synchronization based on the collection results, and determine whether there is erroneous data in the historical collection data. If the determination result is yes, re-synchronize the data and remind of data synchronization anomalies; A data processing unit, used to automatically detect the patient business data using a data quality detection tool, perform data regularization and preprocessing on the patient business data, and standardize the drug-related data in the patient business data to obtain an underlying data pool; wherein the drug-related data includes a generic name of a drug and a pharmacological class of a drug; The data storage unit is used to select a data warehouse and store the underlying data pool in the data warehouse by using a column storage method.

3. The automatic early warning system for abnormal drug sales according to claim 2 is characterized in that: The manual control module includes: A rule-making unit is used to select a control object, configure a control dimension, a control cycle, a warning threshold, and a warning sending object according to the control object, and obtain a drug sales warning rule; wherein each control dimension and each control cycle are respectively configured with different judgment and warning rules; A data comparison unit, configured to determine whether the controlled object meets the warning condition according to the drug sales warning rule, and if the determination result is yes, send the relevant data and the warning result in the controlled object to the warning sending object; The warning sending unit is used to view the warning results using a hyperlink, and the warning results are compared and displayed in a combination of tables and charts.

4. The automatic early warning system for abnormal drug sales according to claim 3 is characterized in that: The control objects include all drugs, specific drugs, generic names of drugs and pharmacological classes; the control dimensions include the entire hospital, departments and large departments; the control cycles include daily statistics, natural month statistics, non-natural month statistics and weekly statistics; the warning thresholds include the maximum amount of the previous N periods, the average amount of the previous N periods, and a fixed value.

5. The automatic early warning system for abnormal drug sales according to claim 4 is characterized in that: The sales abnormality judgment module includes: System rule unit, used to build automatically generated system rules; An abnormal data unit is used to extract abnormal data from the underlying data pool according to the system rules and the drug sales warning rules, establish statistical reports based on the abnormal data, perform preliminary marking of abnormal conditions of the control objects, obtain initial abnormal results, and then manually screen the initial abnormal results to obtain abnormal judgment data; a first model building unit, configured to select a first machine learning model, divide the abnormality judgment data into a first experimental group and a first reference group, train the machine learning model using a machine learning algorithm and the first experimental group, and perform performance evaluation, parameter adjustment, and model optimization using the first reference group during the training process to obtain a sales abnormality prediction model; Abnormal sending unit: used to use the sales abnormality prediction model to monitor the real-time sales of drugs, obtain the abnormal sales trend, generate abnormal sales information based on the abnormal sales trend and send it to the corresponding staff.

6. The automatic early warning system for abnormal drug sales according to claim 5 is characterized in that: The process of constructing the statistical report includes: according to the control cycle, statistics on the historical sales amount, amount growth rate, sales volume, average sales amount, multi-period sales amount of the controlled object, and the number of patients, prescriptions and related diagnoses of the controlled object in the same period to complete the construction of the statistical report.

7. The automatic early warning system for abnormal drug sales according to claim 6 is characterized in that: The sales volume forecasting module includes: An influencing factor extraction unit, used to establish a modeling database including basic information, consumption data, replenishment cycle and other factors according to the underlying data pool, obtain sales influencing factors, and perform regression to fill in missing data in the sales influencing factors; A feature selection unit is used to estimate and match the propensity score of the sales influencing factors using the pandas library in Python, statsmodels, and imblearn, and then perform a balance check and causal effect analysis on the matched indexes and data to obtain the correlation coefficient between the sales influencing factors and the results; a second model building unit, configured to select a second machine learning model, divide the sales influencing factors into a second experimental group and a second reference group, train the machine learning model using a machine learning algorithm and the second experimental group, and perform performance evaluation, parameter adjustment, and model optimization using the second reference group during the training process to obtain a sales prediction model; The early warning sending unit is used to use the sales forecasting model to perform real-time sales forecasting of drugs, obtain the predicted purchase volume and predicted purchase timing, and send the predicted purchase volume and predicted purchase timing to corresponding staff.

8. The automatic early warning system for abnormal drug sales according to claim 7 is characterized in that: The basic information includes name, dosage form, specification, manufacturer, code, drug type, generic name and pharmacological category; the consumption data includes the charging time, sales volume and sales trend of the controlled object; the replenishment cycle includes the warehousing time, outbound time, inventory and replenishment cycle interval of the controlled object; and the other factors include seasonal marks, holiday marks, number of patients and high-frequency diagnoses.

Images