Hospital logistics material management method, device, equipment, medium and product

Abstract

The invention discloses a hospital logistics material management method, device and equipment, a medium and a product, and relates to the technical field of data acquisition and analysis. The method comprises the following steps of: reading n pieces of historical normal sensing data acquired by an environment sensor from a database, randomly extracting a plurality of sample data subsets from a reading result, constructing a plurality of isolated trees based on the data subsets and an isolated forest algorithm, calculating to obtain an abnormal score of the current sensing data based on the isolated trees, and calculating the abnormal score of the current sensing data according to the abnormal score of the current sensing data. And finally, when it is found that the abnormal score exceeds a preset threshold value, triggering execution of an abnormal condition alarm action for the hospital logistics materials, thereby realizing whole-process monitoring and fine management of the material storage environment, remarkably improving the management efficiency, and meanwhile, improving the safety of the logistics materials. In addition, potential abnormal conditions affecting safe storage of the materials can be accurately recognized, abnormal alarm is immediately triggered, then related personnel are timely notified to carry out corresponding processing, safety and supply continuity of the materials are ensured, and unnecessary loss of the materials is reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of data collection and analysis, and particularly relates to a hospital logistics material management method, device, equipment, medium and product. Background Art

[0002] With the rapid development of the medical industry, the scale of hospitals and the number of devices are constantly expanding, and the traditional logistics material management method has been difficult to meet the efficient operation requirements of modern hospitals. There is a wide variety of medical materials, including drugs, medical devices and disposable consumables, etc., and there are strict requirements for storage environment, supply timeliness and use safety. However, the traditional material management mode usually relies on manual records and operations, with problems such as low efficiency, high error rate and poor real-time performance, and it is difficult to meet the requirements of precise and intelligent management in modern hospitals.

[0003] The development of Internet of Things (IoT) technology and big data analysis technology provides new solutions for the logistics material management in smart hospitals. IoT technology can realize the real-time monitoring of key parameters such as temperature, humidity, location and quantity of materials by deploying devices such as sensors and Radio Frequency Identification (RFID), and transmit the data to the cloud or local server through wireless network for processing. Big data analysis technology uses the collected real-time data to build a virtual model of materials, making it possible to manage the entire life cycle of the material state. Therefore, how to combine IoT technology with big data analysis technology to provide an abnormal situation alarm scheme applied to hospital logistics material management, so as to realize the whole-process monitoring and refined management of the material storage environment, accurately identify potential abnormal situations affecting the safe storage of materials and immediately trigger an abnormal alarm, and then timely notify relevant personnel for corresponding processing, ensure the safety and supply continuity of materials, and reduce unnecessary losses of materials, is an urgent research topic for those skilled in the art. Summary of the Invention

[0004] The purpose of the present invention is to provide a hospital logistics material management method, device, computer equipment, computer-readable storage medium and computer program product, so as to solve the problems of low management efficiency, high error rate of abnormal alarm and poor real-time performance in the existing hospital logistics material management scheme in terms of the safe storage of materials.

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

[0006] In the first aspect, a hospital logistics material management method is provided, including:

[0007] Read n pieces of historical normal sensing data collected by environmental sensors from a database, where the environmental sensors are arranged in the storage environment of hospital logistics supplies, and n represents a positive integer greater than or equal to 5;

[0008] Randomly extract multiple sample data subsets from the n pieces of historical normal sensing data, where each sample data subset contains m pieces of historical normal sensing data, and m represents a positive integer greater than or equal to 3 and less than n;

[0009] For each sample data subset among the multiple sample data subsets, randomly select features and split points based on the isolation forest algorithm to recursively split the corresponding data set until the corresponding isolation tree is constructed;

[0010] Receive the current sensing data from the environmental sensors in real time;

[0011] For each isolation tree corresponding to each sample data subset, determine the height of the current sensing data on the corresponding tree;

[0012] According to all the heights, calculate the anomaly score s of the current sensing data according to the following anomaly score formula:

[0013]

[0014] In the formula, Eh represents the average height calculated based on all the heights, and c(n) represents a normalization factor positively correlated with n;

[0015] If it is found that the anomaly score s exceeds the preset threshold, trigger an anomaly alarm action for the hospital logistics supplies.

[0016] Based on the above invention content, a new solution for combining Internet of Things technology and big data analysis technology for hospital logistics supplies management is provided. That is, first read n pieces of historical normal sensing data collected by environmental sensors from a database, and randomly extract multiple sample data subsets from the reading results. Then, based on the data subsets and the isolation forest algorithm, construct multiple isolation trees. Then, calculate the anomaly score of the current sensing data based on the isolation trees. Finally, when it is found that the anomaly score exceeds the preset threshold, trigger an anomaly alarm action for the hospital logistics supplies. In this way, the whole process monitoring and refined management of the material storage environment can be realized. While significantly improving the management efficiency, it can also accurately identify potential abnormal situations affecting the safe storage of materials and immediately trigger an anomaly alarm, and then timely notify relevant personnel for corresponding handling to ensure the safety and supply continuity of materials, reduce unnecessary losses of materials, and facilitate practical application and promotion.

[0017] In a possible design, the environmental sensor includes a temperature sensor, a humidity sensor, a light intensity sensor, a PM2.5 sensor, and / or an odor sensor.

[0018] In a possible design, after calculating the anomaly score s of the current sensed data, the method further includes:

[0019] If it is found that the anomaly score s does not exceed the preset threshold, the current sensed data is stored in the database as a piece of normal sensed data;

[0020] Periodically read n pieces of historical normal sensed data stored in the database within the most recent unit time period;

[0021] Randomly re-sample from the n pieces of historical normal sensed data stored in the most recent unit time period to obtain multiple new subsets of sample data, where each new subset of sample data contains m pieces of historical normal sensed data;

[0022] For each new subset of sample data in the multiple new subsets of sample data, randomly re-select features and split points based on the Isolation Forest algorithm to recursively split the corresponding data set until the corresponding isolation tree is constructed.

[0023] In a possible design, the method further includes:

[0024] Dynamically adjust the preset threshold according to any one or any combination of the following methods (A) to (D):

[0025] (A) Periodically adjust the preset threshold: Collect multiple pieces of historical sensed data collected by the environmental sensor within the time period of the current year; for each piece of historical sensed data in the multiple pieces of historical sensed data, calculate the corresponding anomaly score based on the respective isolation trees and the anomaly score formula; determine the preset threshold applicable within the time period of the current year according to the anomaly score distribution of each piece of historical sensed data;

[0026] (B)Perform feedback adjustment on the preset threshold: Collect multiple anomaly scores calculated in the most recent unit time period; if it is found that the total number of anomaly scores exceeding the current preset threshold among the multiple anomaly scores is too large and at least k anomaly scores belong to the interval (Th, Th + Tk], then increase the current preset threshold; and if it is found that the total number of anomaly scores exceeding the current preset threshold among the multiple anomaly scores is too small and at least k anomaly scores belong to the interval [Th - Tk, Th), then decrease the current preset threshold, where k represents a positive integer, Th represents the current preset threshold, and Tk represents the absolute value of the difference between a value close to the current preset threshold and the current preset threshold;

[0027] (C)Perform adaptive adjustment on the preset threshold: Read n pieces of historical normal sensing data stored in the most recent unit time period from the database; randomly re - extract multiple new subsets of sample data from the n pieces of historical normal sensing data stored in the most recent unit time period, where each new subset of sample data contains m pieces of historical normal sensing data; for each new subset of sample data among the multiple new subsets of sample data, randomly re - select features and split points based on the isolation forest algorithm to recursively split the corresponding data set until the corresponding isolation tree is constructed; collect multiple pieces of historical sensing data collected by the environmental sensor in the most recent unit time period; for each piece of historical sensing data among the multiple pieces of historical sensing data, calculate the corresponding anomaly score based on each isolation tree corresponding one - to - one with each new subset of sample data and the anomaly score formula; if it is found that the proportion of anomaly scores exceeding the current preset threshold among the multiple anomaly scores corresponding one - to - one with the multiple pieces of historical sensing data is too high, then increase the current preset threshold; and if it is found that the proportion of anomaly scores exceeding the current preset threshold among the multiple anomaly scores corresponding one - to - one with the multiple pieces of historical sensing data is too low, then decrease the current preset threshold;

[0028] (D)Perform combined adjustment on the preset threshold: Combine the relevance between multi - feature data and environmental sensing data to dynamically optimize the preset threshold, where the multi - feature data includes aging determination data of the environmental sensor, sensitivity change determination data of the environmental sensor, calibration data of the environmental sensor, and / or non - environmental sensor data.

[0029] In a possible design, if it is found that the anomaly score s exceeds the preset threshold, then trigger an abnormal situation alarm action for the hospital logistics supplies, including:

[0030] If it is found that the abnormal score s exceeds a preset threshold and belongs to the range of abnormal score intervals corresponding to a certain warning level, an abnormal situation alarm action corresponding to the certain warning level is triggered for the hospital logistics supplies.

[0031] In a possible design, the method further includes:

[0032] The current sensing data, the abnormal score, and / or the abnormal detection result are displayed in real time through a chart and / or a dashboard.

[0033] In a second aspect, a hospital logistics supply management device is provided, including a sensing data reading unit, a data subset extraction unit, an isolation tree construction unit, a sensing data receiving unit, a tree height determination unit, an abnormal score calculation unit, and an alarm action trigger unit;

[0034] The sensing data reading unit is configured to read n pieces of historical normal sensing data collected by an environmental sensor from a database, where the environmental sensor is arranged in the storage environment of hospital logistics supplies, and n represents a positive integer greater than or equal to 5;

[0035] The data subset extraction unit is communicatively connected to the sensing data reading unit and is configured to randomly extract multiple sample data subsets from the n pieces of historical normal sensing data, where each sample data subset includes m pieces of historical normal sensing data, and m represents a positive integer greater than or equal to 3 and less than n;

[0036] The isolation tree construction unit is communicatively connected to the data subset extraction unit and is configured to, for each sample data subset in the multiple sample data subsets, randomly select features and split points based on the isolation forest algorithm to recursively split the corresponding data set until the corresponding isolation tree is constructed;

[0037] The sensing data receiving unit is configured to receive the current sensing data from the environmental sensor in real time;

[0038] The tree height determination unit is communicatively connected to the isolation tree construction unit and the sensing data receiving unit respectively, and is configured to determine the height of the current sensing data on the corresponding tree for each isolation tree corresponding to each sample data subset one by one;

[0039] The abnormal score calculation unit is communicatively connected to the tree height determination unit and is configured to calculate the abnormal score s of the current sensing data according to all the heights according to the following abnormal score formula:

[0040]

[0041] Wherein, Eh represents the average height calculated based on all the heights, and c(n) represents a normalization factor that is positively correlated with n;

[0042] The alarm action triggering unit is communicatively connected to the abnormal score calculation unit, and is configured to trigger an abnormal situation alarm action for the hospital logistics materials when it is found that the abnormal score s exceeds a preset threshold.

[0043] In a third aspect, the present invention provides a computer device, including a memory, a processor, and a transceiver that are communicatively connected in sequence. Among them, the memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program and execute the hospital logistics material management method as described in the first aspect or any possible design in the first aspect.

[0044] In a fourth aspect, the present invention provides a computer-readable storage medium, on which instructions are stored. When the instructions are run on a computer, the hospital logistics material management method as described in the first aspect or any possible design in the first aspect is executed.

[0045] In a fifth aspect, the present invention provides a computer program product, including a computer program or instructions, and the computer program or the instructions, when executed by a computer, implement the hospital logistics material management method as described in the first aspect or any possible design in the first aspect.

[0046] Beneficial effects of the above solutions:

[0047] (1) The present invention creatively provides a new solution that combines Internet of Things technology and big data analysis technology for hospital logistics material management. That is, first read n historical normal sensing data collected by environmental sensors from a database, and randomly extract multiple sample data subsets from the reading results. Then, based on the data subsets and the isolation forest algorithm, construct multiple isolation trees. Then, calculate the abnormal score of the current sensing data based on the isolation trees. Finally, when it is found that the abnormal score exceeds a preset threshold, trigger an abnormal situation alarm action for the hospital logistics materials. In this way, the whole process monitoring and refined management of the material storage environment can be realized. While significantly improving the management efficiency, it can also accurately identify potential abnormal situations that affect the safe storage of materials and immediately trigger an abnormal alarm, and then timely notify relevant personnel for corresponding processing, ensuring the safety and supply continuity of materials, reducing unnecessary losses of materials, and being convenient for practical application and promotion;

[0048] (2) It is also possible to regularly reconstruct the isolation forest using the latest normal environment data to optimize the normal environment distribution characteristics and ensure the accuracy and robustness of the detection results;

[0049] (3) Compared with traditional methods, it can significantly reduce manual intervention and false alarm rate, and improve the accuracy and management efficiency of anomaly detection;

[0050] (4) Through dynamic threshold update, it can adapt to seasonal changes and equipment performance fluctuations, ensuring long-term stable operation;

[0051] (5) It provides a data visualization function, which can provide real-time monitoring, historical records and trend analysis for managers, helping with scientific decision-making and optimized management;

[0052] (6) It is especially suitable for cold chain materials with strict environmental requirements (such as vaccines and blood products), can effectively reduce material losses and operating costs, realize the intelligent, automated and refined management of hospital logistics materials, and significantly improve hospital operation efficiency and service quality. Description of the Drawings

[0053] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0054] Figure 1 It is a schematic flowchart of the hospital logistics material management method provided by the embodiment of the present application.

[0055] Figure 2 It is a schematic structural diagram of the hospital logistics material management system provided by the embodiment of the present application.

[0056] Figure 3 It is a schematic flowchart of regularly reconstructing the isolation forest in the hospital logistics material management method provided by the embodiment of the present application.

[0057] Figure 4 It is a schematic structural diagram of the hospital logistics material management device provided by the embodiment of the present application.

[0058] Figure 5 It is a schematic structural diagram of the computer device provided by the embodiment of the present application. Detailed Embodiments

[0059] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the present invention will be briefly introduced below in combination with the accompanying drawings and the description of the embodiments or the prior art. Obviously, the following description of the structures of the accompanying drawings is only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other embodiments can be obtained based on these embodiments. It should be noted here that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation on the present invention.

[0060] It should be understood that although terms such as first and second etc. may be used herein to describe various objects, these objects should not be limited by these terms. These terms are only used to distinguish one object from another. For example, the first object can be called the second object, and similarly the second object can be called the first object, without departing from the scope of the exemplary embodiments of the present invention.

[0061] It should be understood that for the term "and / or" that may appear in this article, it is only a description of the association relationship of associated objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, B exists alone, or A and B exist simultaneously, etc. For another example, A, B and / or C can mean any one of A, B and C or any combination of them; for the term " / and" that may appear in this article, it is a description of another association object relationship, indicating that two relationships can exist. For example, A / and B can mean: A exists alone or A and B exist simultaneously, etc. In addition, for the character " / " that may appear in this article, generally it means that the associated objects before and after are in an "or" relationship.

[0062] Embodiment

[0063] As Figures 1 - 2 shown, the hospital logistics material management method provided in the first aspect of this embodiment can, but is not limited to, be executed by a computer device having certain computing resources and respectively communicatively connected to an environmental sensor, an RFID tag, and a database. For example, it can be executed by an electronic device such as a cloud server, a personal computer (Personal Computer, PC, referring to a multi-purpose computer suitable for personal use in terms of size, price, and performance; desktop computers, laptop computers, small laptop computers, tablet computers, and ultrabooks, etc. all belong to personal computers), a smart phone, a personal digital assistant (Personal Digital Assistant, PDA), or a wearable device. As Figure 1 shown, the hospital logistics material management method can, but is not limited to, include the following steps S1 to S7.

[0064] S1. Read n pieces of historical normal sensing data collected by environmental sensors from the database, where the environmental sensors are arranged in the storage environment of hospital logistics supplies, and n represents a positive integer greater than or equal to 5.

[0065] In step S1, the hospital logistics supplies are the objects to be supervised, specifically but not limited to including key supplies such as drugs and / or reagents, and their corresponding storage environment is the existing environment, such as a cold chain warehouse or a drug storage room, etc. Specifically, the environmental sensors include but are not limited to temperature sensors, humidity sensors, light intensity sensors, PM2.5 sensors, and / or odor sensors, etc. The sensing data collected in this way will include but not be limited to environmental temperature, environmental humidity, environmental light intensity, environmental PM2.5 value, and / or environmental odor information, etc. The historical normal sensing data refers to the sensing data that has been collected historically and has been determined to be in a normal situation. The determination methods can include but are not limited to the following: (1) Combine expert knowledge and manual annotation to confirm the normality of the sensing data. For example, based on all the historically collected sensing data, use statistical methods (such as quantile analysis, etc.) to dynamically define the normal range (such as the range within which 95% of the sensing data falls), and regard the historically collected sensing data that falls within this normal range as historical normal sensing data; (2) According to the alarm records of the method in this embodiment, automatically exclude the historically collected sensing data during the abnormal alarm time period, and regard the remaining historically collected sensing data as historical normal sensing data. The historical normal sensing data can specifically include but is not limited to multiple different sensing data (such as temperature sensors, humidity sensors, light intensity sensors, PM2.5 sensors, and / or odor sensors, etc.) collected historically at the same acquisition moment and / or the same sensing data collected historically at different acquisition moments that are continuous in time series. The database can adopt but is not limited to a MySQL database (i.e., a relational database management system) or a time series database, etc., in order to support efficient data reading and querying. In addition, it is necessary to preprocess the aforementioned sensing data by means such as cleaning, denoising, standardization processing, and / or using a sliding window or median filtering algorithm to smooth data fluctuations to exclude outliers and ensure data quality.

[0066] S2. Randomly select multiple sample data subsets from the n pieces of historical normal sensing data, where each sample data subset contains m pieces of historical normal sensing data, and m represents a positive integer greater than or equal to 3 and less than n.

[0067] S3. For each sample data subset in the multiple sample data subsets, randomly select features and split points based on the isolation forest algorithm to recursively split the corresponding data set until the corresponding isolation tree is constructed.

[0068] In the step S3, the Isolation Forest (iForest) algorithm is a machine learning algorithm for anomaly detection, proposed by Professor Zhou Zhihua of Nanjing University and others in 2008. This algorithm is proposed based on the following intuition: Anomaly points are the minority in the data, and their distribution in the feature space is different from that of normal data points, usually manifested as being farther away from most data points. The Isolation Forest algorithm specifically realizes the detection of anomaly points by constructing multiple Isolation Trees. The basic steps are as follows: Randomly select a feature, and then randomly select a splitting point (also called a segmentation point) between the minimum and maximum values of this feature; According to the splitting point, the data set is split into two subsets; Repeat the above steps for each subset (i.e., perform recursive splitting) until the stopping condition is met, such as the number of data points in the subset is less than a certain threshold; Through the above steps, multiple Isolation Trees can be constructed, and each tree is a random splitting process. Based on the above algorithm steps, corresponding Isolation Trees can be constructed for each of the sample data subsets respectively.

[0069] S4. Real-time receive the current sensing data from the environmental sensor.

[0070] In the step S4, the current sensing data needs to have the same data structure as the historical normal sensing data so that the height of the current sensing data on the Isolation Tree can be determined subsequently and the current sensing data can also be used as historical normal sensing data relative to the subsequent sensing data. Specifically, the current sensing data can be transmitted from the environmental sensor to the local device through a conventional wireless IoT network (such as a WiFi network, a LoRa network, or an NB-IoT network, etc.). In addition, the basic information of the materials from the RFID tags bound to the hospital logistics materials can also be received to indicate the unique identifier, name, batch number, storage requirements, etc. of the hospital logistics materials.

[0071] S5. For each of the Isolation Trees corresponding to each of the sample data subsets, determine the height of the current sensing data on the corresponding tree.

[0072] In the step S5, the height of the current sensing data on the Isolation Tree is also called the path length, which can be determined conventionally based on the recursive splitting rule of the Isolation Tree: If more splitting times are required to isolate the current sensing data (i.e., divide it into a leaf node of the Isolation Tree), the higher the height, and vice versa.

[0073] S6. According to all the above heights, calculate the anomaly score s of the current sensing data according to the following anomaly score formula:

[0074]

[0075] In the formula, Eh represents the average height calculated based on all the said heights, and c(n) represents a normalization factor that is positively correlated with n.

[0076] In the step S6, the average height Eh is used to measure the separability of the sample to be measured (i.e., the current sensing data) in the isolation forest (i.e., composed of the respective isolation trees). The lower the height, the more abnormal the sample to be measured. The normalization factor c(n) is used to normalize the average height Eh. Since when the number of samples is large, the density of sample points is higher and a higher height is required to isolate a certain sample point, it is necessary to make it positively correlated with n. The specific formula is as follows: The anomaly score s is used to comprehensively evaluate the anomaly of the sample to be measured: when the average height Eh is significantly less than the normalization factor c(n), the anomaly score s approaches 1 and is determined as an anomaly point; while when the average height Eh is close to the normalization factor c(n), the anomaly score s is close to 0.5 and is determined as a normal point.

[0077] S7. If it is found that the anomaly score s exceeds a preset threshold, then an alarm action for an abnormal situation is triggered for the hospital logistics supplies.

[0078] In the step S7, the preset threshold can be fixedly configured in the storage requirement or can be dynamically adjusted to adapt to seasonal fluctuations of the environment or changes in equipment performance. Specifically, the method further includes but is not limited to dynamically adjusting the preset threshold in any one of the following manners (A) to (D) or any combination thereof.

[0079] (A) Periodically adjust the preset threshold: Collect multiple historical sensing data collected by the environment sensor in the time period within the current year; for each piece of historical sensing data in the multiple historical sensing data, calculate the corresponding anomaly score based on the respective isolation trees and the anomaly score formula; determine the preset threshold applicable in the time period within the current year according to the anomaly score distribution of the respective historical sensing data. The aforementioned time period within the year can be but is not limited to being in units of seasons or months. For example, if the current time is 14:03 on January 17th, then the time period within the current year is January. Thus, it is possible to but is not limited to collect multiple historical sensing data collected by the environment sensor in January of the previous year, January of last year, and January of this year. The aforementioned process of calculating the anomaly score can be obtained by conventional derivation referring to the aforementioned steps S5 to S6 and will not be elaborated herein. In addition, determining the preset threshold applicable in the time period within the current year according to the anomaly score distribution of the respective historical sensing data can be but is not limited to being conventionally determined based on the normal distribution result of the anomaly score.

[0080] (B) Perform feedback adjustment on the preset threshold: collect multiple anomaly scores calculated in the most recent unit time period; if it is found that the total number of anomaly scores exceeding the current preset threshold among the multiple anomaly scores is too large and at least k anomaly scores belong to the interval (Th, Th + Tk], then increase the current preset threshold; and if it is found that the total number of anomaly scores exceeding the current preset threshold among the multiple anomaly scores is too small and at least k anomaly scores belong to the interval [Th - Tk, Th), then decrease the current preset threshold, where k represents a positive integer, Th represents the current preset threshold, and Tk represents the absolute value of the difference between the value close to the current preset threshold and the current preset threshold. The aforementioned unit time period can be but is not limited to seasons, months, or weeks, etc. For example, collect multiple anomaly scores calculated in the most recent month. In addition, the determination process of whether the total number is too large or too small can be routinely determined based on the comparison result of the total number and the threshold.

[0081] (C) Adaptive adjustment of the preset threshold: Read n pieces of historical normal sensing data stored in the database in the most recent unit time period; randomly re-sample from the n pieces of historical normal sensing data stored in the most recent unit time period to obtain multiple new subsets of sample data, where each new subset of sample data contains m pieces of historical normal sensing data; for each new subset of sample data in the multiple new subsets of sample data, randomly re-select features and split points based on the Isolation Forest algorithm to recursively split the corresponding data set until the corresponding isolation tree is constructed; collect multiple pieces of historical sensing data collected by the environmental sensor in the most recent unit time period; for each piece of historical sensing data in the multiple pieces of historical sensing data, calculate the corresponding anomaly score based on each isolation tree corresponding to each new subset of sample data and the anomaly score formula; if it is found that the proportion of anomaly scores exceeding the current preset threshold in the multiple anomaly scores corresponding to the multiple pieces of historical sensing data is too high, increase the current preset threshold; if it is found that the proportion of anomaly scores exceeding the current preset threshold in the multiple anomaly scores corresponding to the multiple pieces of historical sensing data is too low, decrease the current preset threshold. The aforementioned unit time period can be, but is not limited to, seasons, months, weeks, etc. For example, read n pieces of historical normal sensing data stored in the database in the most recent month. The aforementioned isolation tree construction process and anomaly score calculation process can be specifically derived from the conventional derivation of the aforementioned steps S1 to S6 and will not be elaborated here. In addition, the aforementioned process of judging whether the proportion is too high or too low can also be conventionally determined based on the comparison result of the total number and the threshold; for example, if the current preset threshold is 0.8 and the proportion of anomaly scores exceeding the current preset threshold is greater than 10%, increase the current preset threshold to 0.85; if the current preset threshold is 0.8 and the proportion of anomaly scores exceeding the current preset threshold is less than 1%, decrease the current preset threshold to 0.75.

[0082] (D)Jointly adjust the preset threshold: Combine the relevance between multi-feature data and environmental sensing data to dynamically optimize the preset threshold, where the multi-feature data includes but is not limited to the aging determination data of the environmental sensor, the sensitivity change determination data of the environmental sensor, the calibration data of the environmental sensor, and / or non-environmental sensor data, etc. The specific process of the foregoing joint adjustment can be but is not limited to being conventionally implemented based on artificial intelligence algorithms. For example, import the multi-feature data into an anomaly score threshold estimation model (the specific training process of which is prior art) pre-trained based on an artificial intelligence algorithm (which is a core algorithm that specifically studies how a computer simulates or implements human learning behaviors to acquire new knowledge or skills and reorganize the existing knowledge structure to continuously improve its own performance, and is the fundamental way to make a computer intelligent; specifically, it can be but is not limited to using machine learning algorithms based on support vector machines, stochastic gradient descent methods, multivariate linear regression, multi-layer perceptrons, decision trees, backpropagation neural networks, or radial basis function networks), and output an anomaly score estimation threshold as the preset threshold.

[0083] In step S7, the specific manners of the abnormal situation alarm action include but are not limited to multi-channel notification manners such as SMS notification, APP push, and / or sound and light alarm. In order to trigger different levels of alarm actions according to the high or low abnormal scores, preferably, if it is found that the abnormal score s exceeds the preset threshold, then an abnormal situation alarm action is triggered for the hospital logistics supplies, including but not limited to: if it is found that the abnormal score s exceeds the preset threshold and belongs to the abnormal score interval range corresponding to a certain warning level, then an abnormal situation alarm action corresponding to the certain warning level is triggered for the hospital logistics supplies. For example, if the certain warning level is a low warning level, the management personnel can be reminded to check the storage environment; and if the certain warning level is a high warning level (the corresponding abnormal score interval range does not overlap with and is higher than that of the low warning level), then an emergency alarm is triggered and relevant personnel are notified to take emergency measures. When performing the multi-channel notification manner, the basic information of the supplies can be carried in the notification content to indicate the hospital logistics supplies. In addition, in order to provide a traceability analysis function for abnormal events to help management personnel locate the root cause of the problem, the method further includes: real-time displaying the current sensing data, the abnormal score, and / or the abnormal detection result, etc. through charts and / or dashboards.

[0084] Based on the hospital logistics material management method described in the foregoing steps S1 to S7, a new solution for combining Internet of Things technology and big data analysis technology for hospital logistics material management is provided. That is, first, n historical normal sensing data collected by environmental sensors are read from the database, and multiple sample data subsets are randomly selected from the read results. Then, multiple isolation trees are constructed based on the data subsets and the isolation forest algorithm. Then, the anomaly score of the current sensing data is calculated based on the isolation trees. Finally, when it is found that the anomaly score exceeds the preset threshold, an anomaly alarm action is triggered for the hospital logistics materials. In this way, the whole process monitoring and refined management of the material storage environment can be realized. While significantly improving the management efficiency, it can also accurately identify potential abnormal situations affecting the safe storage of materials and immediately trigger an anomaly alarm, and then timely notify relevant personnel for corresponding processing to ensure the safety and supply continuity of materials, reduce unnecessary losses of materials, and facilitate practical application and promotion.

[0085] Based on the technical solution of the foregoing first aspect, this embodiment also provides a possible design for how to periodically reconstruct the isolation forest, that is, as Figure 3 shown, after calculating the anomaly score s of the current sensing data, the method further includes but is not limited to the following steps S81 to S84.

[0086] S81. If it is found that the anomaly score s does not exceed the preset threshold, the current sensing data is stored in the database as a normal sensing data.

[0087] S82. Periodically read n historical normal sensing data stored in the database in the most recent unit period.

[0088] In step S82, the above unit period can be but is not limited to seasons, months, weeks, etc. For example, n historical normal sensing data stored in the database in the most recent month are read. In addition, the period can be but is not limited to every unit period. For example, every month, n historical normal sensing data stored in the database in the most recent month are read.

[0089] S83. Randomly re-select multiple new sample data subsets from the n historical normal sensing data stored in the most recent unit period, where each new sample data subset contains m historical normal sensing data.

[0090] S84. For each new sample data subset in the multiple new sample data subsets, based on the isolation forest algorithm, randomly re-select features and split points to recursively split the corresponding data set until the corresponding isolation tree is constructed.

[0091] The specific processes of the above steps S83 to S84 can be obtained by conventional derivation based on the foregoing S2 to S3, and will not be elaborated herein.

[0092] Based on the above possible design 1, it is also possible to regularly use the latest normal environment data to reconstruct the isolation forest, optimize the normal environment distribution characteristics, and ensure the accuracy and robustness of the detection results.

[0093] As Figure 4 shown, in the second aspect of this embodiment, a virtual device for implementing the hospital logistics material management method described in the first aspect or any possible design in the first aspect is provided, including a sensing data reading unit, a data subset extraction unit, an isolation tree construction unit, a sensing data receiving unit, a tree height determination unit, an anomaly score calculation unit, and an alarm action trigger unit;

[0094] The sensing data reading unit is configured to read n pieces of historical normal sensing data collected by environmental sensors from a database, where the environmental sensors are arranged in the storage environment of hospital logistics materials, and n represents a positive integer greater than or equal to 5;

[0095] The data subset extraction unit is communicatively connected to the sensing data reading unit and is configured to randomly extract multiple sample data subsets from the n pieces of historical normal sensing data, where each sample data subset contains m pieces of historical normal sensing data, and m represents a positive integer greater than or equal to 3 and less than n;

[0096] The isolation tree construction unit is communicatively connected to the data subset extraction unit and is configured to, for each sample data subset in the multiple sample data subsets, randomly select features and split points based on the isolation forest algorithm to recursively split the corresponding data set until the corresponding isolation tree is constructed;

[0097] The sensing data receiving unit is configured to receive current sensing data from the environmental sensors in real time;

[0098] The tree height determination unit is communicatively connected to the isolation tree construction unit and the sensing data receiving unit respectively, and is configured to determine the height of the current sensing data on the corresponding tree for each isolation tree corresponding to each sample data subset one by one;

[0099] The anomaly score calculation unit is communicatively connected to the tree height determination unit and is configured to calculate the anomaly score s of the current sensing data according to all the heights according to the following anomaly score formula:

[0100]

[0101] Wherein, Eh represents the average height calculated based on all the said heights, and c(n) represents a normalization factor that is positively correlated with n;

[0102] The alarm action triggering unit, communicatively connected to the anomaly score calculation unit, is configured to trigger an alarm action for abnormal situations for the hospital logistics supplies when it is found that the anomaly score s exceeds a preset threshold.

[0103] For the working process, working details and technical effects of the foregoing device provided in the second aspect of this embodiment, reference may be made to the hospital logistics supplies management method described in the first aspect or any possible design in the first aspect, which will not be elaborated herein.

[0104] As Figure 5 shown, the third aspect of this embodiment provides a computer device for executing the hospital logistics supplies management method described in the first aspect or any possible design in the first aspect, including a memory, a processor, and a transceiver that are communicatively connected in sequence. Among them, the memory is used to store computer programs, the transceiver is used to send and receive messages, and the processor is used to read the computer programs and execute the hospital logistics supplies management method described in the first aspect or any possible design in the first aspect. Specifically, by way of example, the memory may include, but is not limited to, random access memory (RAM), read-only memory (ROM), flash memory, first input first output (FIFO), and / or first input last output (FILO), etc.; the processor may include, but is not limited to, a microprocessor of the STM32F105 series. In addition, the computer device may further include, but is not limited to, a power module, a display screen, and other necessary components.

[0105] For the working process, working details and technical effects of the foregoing computer device provided in the third aspect of this embodiment, reference may be made to the hospital logistics supplies management method described in the first aspect or any possible design in the first aspect, which will not be elaborated herein.

[0106] In the fourth aspect of this embodiment, there is provided a computer-readable storage medium storing instructions for the hospital logistics material management method as described in the first aspect or any possible design in the first aspect, that is, instructions are stored on the computer-readable storage medium, and when the instructions are run on a computer, the hospital logistics material management method as described in the first aspect or any possible design in the first aspect is executed. Among them, the computer-readable storage medium refers to a carrier for storing data, and may include, but is not limited to, computer-readable storage media such as floppy disks, optical discs, hard disks, flash memories, USB flash drives, and / or Memory Sticks. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0107] For the working process, working details, and technical effects of the aforementioned computer-readable storage medium provided in the fourth aspect of this embodiment, reference may be made to the hospital logistics material management method as described in the first aspect or any possible design in the first aspect, and details will not be elaborated here.

[0108] In the fifth aspect of this embodiment, there is provided a computer program product including a computer program or instructions, and when the computer program or the instructions are executed by a computer, the hospital logistics material management method as described in the first aspect or any possible design in the first aspect is implemented. Among them, the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0109] Finally, it should be noted that the above are only preferred embodiments of the present invention and are not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A hospital logistics material management method, characterized in that: include: Reading n pieces of historical normal sensor data collected by environmental sensors from a database, wherein the environmental sensors are arranged in a storage environment of hospital logistics materials, and n represents a positive integer greater than or equal to 5; A plurality of sample data subsets are randomly extracted from the n pieces of historical normal sensor data, wherein the sample data subsets include m pieces of historical normal sensor data, where m represents a positive integer greater than or equal to 3 and less than n; For each sample data subset in the multiple sample data subsets, randomly selecting features and splitting points based on the isolation forest algorithm to recursively split the corresponding data set until a corresponding isolation tree is constructed; Receiving current sensing data from the environmental sensor in real time; For each isolated tree corresponding to each sample data subset, determining the height of the current sensor data on the corresponding tree; According to all the heights, the anomaly score s of the current sensor data is calculated according to the following anomaly score formula: Wherein, Eh represents the average height calculated based on all the heights, and c(n) represents a normalization factor positively correlated with n; If it is found that the abnormal score s exceeds a preset threshold, an abnormal situation alarm action is triggered for the hospital logistics materials.

2. The hospital logistics material management method according to claim 1, characterized in that: The environmental sensors include a temperature sensor, a humidity sensor, a light intensity sensor, a PM2.5 sensor and / or an odor sensor.

3. The hospital logistics material management method according to claim 1, characterized in that: After calculating the abnormal score s of the current sensor data, the method further includes: If it is found that the abnormal score s does not exceed the preset threshold, the current sensor data is stored in the database as a normal sensor data; Periodically reading n copies of historical normal sensor data stored in the most recent unit time period from the database; Re-randomly extracting a plurality of new subsets of sample data from the n pieces of historical normal sensor data stored in the most recent unit time period, wherein the new subsets of sample data include m pieces of historical normal sensor data; For each new subset of sample data in the multiple new subsets of sample data, features and splitting points are randomly reselected based on the isolation forest algorithm to recursively split the corresponding data set until a corresponding isolation tree is constructed.

4. The hospital logistics material management method according to claim 1, characterized in that: The method further comprises: The preset threshold is dynamically adjusted according to any one of the following methods (A) to (D) or any combination thereof: (A) periodically adjusting the preset threshold: collecting a plurality of historical sensor data collected by the environmental sensor in the current annual period; calculating a corresponding anomaly score for each of the plurality of historical sensor data based on each isolated tree and the anomaly score formula; and determining the preset threshold applicable to the current annual period according to the distribution of the anomaly scores of each of the historical sensor data; (B) Feedback adjustment of the preset threshold: collect multiple anomaly scores calculated in the most recent unit time period; if it is found that the total number of anomaly scores exceeding the current preset threshold among the multiple anomaly scores is too large and at least k anomaly scores belong to the interval (Th, Th+Tk]), then increase the current preset threshold; and if it is found that the total number of anomaly scores exceeding the current preset threshold among the multiple anomaly scores is too small and at least k anomaly scores belong to the interval [Th-Tk, Th), then reduce the current preset threshold, wherein k represents a positive integer, Th represents the current preset threshold, and Tk represents the absolute value of the difference between the value close to the current preset threshold and the current preset threshold; (C) Adaptively adjusting the preset threshold: reading n pieces of historical normal sensor data stored in the most recent unit time period from the database; re-randomly extracting multiple new sample data subsets from the n pieces of historical normal sensor data stored in the most recent unit time period, wherein the new sample data subsets contain m pieces of historical normal sensor data; for each new sample data subset in the multiple new sample data subsets, re-randomly selecting features and segmentation points based on the isolation forest algorithm to recursively split the corresponding data set until a corresponding isolation tree is constructed; collecting the samples of the samples collected by the environmental sensor in the most recent unit time period; A plurality of historical sensor data collected within a unit time period; for each of the plurality of historical sensor data, based on each isolated tree corresponding to each new subset of the sample data and the anomaly score formula, a corresponding anomaly score is calculated; if it is found that the proportion of anomaly scores exceeding the current preset threshold among the plurality of anomaly scores corresponding to the plurality of historical sensor data is too high, the current preset threshold is increased; and if it is found that the proportion of anomaly scores exceeding the current preset threshold among the plurality of anomaly scores corresponding to the plurality of historical sensor data is too low, the current preset threshold is reduced; (D) Jointly adjusting the preset threshold: dynamically optimizing the preset threshold based on the correlation between the multi-feature data and the environmental sensor data, wherein the multi-feature data contains aging measurement data of the environmental sensor, sensitivity change measurement data of the environmental sensor, calibration data of the environmental sensor and / or non-environmental sensor data.

5. The hospital logistics material management method according to claim 1, characterized in that: If it is found that the abnormal score s exceeds the preset threshold, an abnormal situation alarm action is triggered for the hospital logistics materials, including: If it is found that the abnormal score s exceeds the preset threshold and belongs to the abnormal score interval corresponding to a certain warning level, an abnormal situation alarm action corresponding to the certain warning level is triggered for the hospital logistics materials.

6. The hospital logistics material management method according to claim 1, characterized in that: The method further comprises: The current sensor data, the anomaly score and / or anomaly detection result are displayed in real time through charts and / or dashboards.

7. A hospital logistics material management device, characterized in that: It includes a sensor data reading unit, a data subset extraction unit, an isolated tree construction unit, a sensor data receiving unit, a tree height determination unit, an abnormal score calculation unit and an alarm action triggering unit; The sensor data reading unit is used to read n copies of historical normal sensor data collected by environmental sensors from a database, wherein the environmental sensors are arranged in a storage environment of hospital logistics materials, and n represents a positive integer greater than or equal to 5; The data subset extraction unit is communicatively connected to the sensor data reading unit, and is used to randomly extract a plurality of sample data subsets from the n pieces of historical normal sensor data, wherein the sample data subsets include m pieces of historical normal sensor data, where m represents a positive integer greater than or equal to 3 and less than n; The isolation tree construction unit is communicatively connected to the data subset extraction unit, and is used to randomly select features and split points based on the isolation forest algorithm for each sample data subset in the multiple sample data subsets to recursively split the corresponding data set until a corresponding isolation tree is constructed; The sensor data receiving unit is used to receive the current sensor data from the environmental sensor in real time; The tree height determination unit is respectively connected to the isolated tree construction unit and the sensor data receiving unit for communication, and is used to determine the height of the current sensor data on the corresponding tree for each isolated tree corresponding to each sample data subset; The anomaly score calculation unit is communicatively connected to the tree height determination unit, and is used to calculate the anomaly score s of the current sensor data according to all the heights according to the following anomaly score formula: Wherein, Eh represents the average height calculated based on all the heights, and c(n) represents a normalization factor positively correlated with n; The alarm action triggering unit is communicatively connected to the abnormality score calculating unit, and is used to trigger an abnormal situation alarm action for the hospital logistics materials when it is found that the abnormality score s exceeds a preset threshold.

8. A computer device, characterized in that: It includes a memory, a processor and a transceiver which are communicatively connected in sequence, wherein the memory is used to store computer programs, the transceiver is used to send and receive messages, and the processor is used to read the computer program to execute the hospital logistics material management method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores instructions, and when the instructions are executed on the computer, the hospital logistics material management method as described in any one of claims 1 to 6 is executed.

10. A computer program product comprising a computer program or instructions, characterized in that The computer program or the instruction, when executed by a computer, implements the hospital logistics material management method as described in any one of claims 1 to 6.

Images