Order Anomaly Monitoring System Based on Enterprise Management
By constructing an order anomaly monitoring system with a feature space and adaptive parameter selection, the efficiency and accuracy issues of order anomaly monitoring in enterprise management are solved, enabling accurate identification and management of order anomalies.
Patent Information
- Application Number
- CN202310921273.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-07-26
- Publication Date
- 2026-11-17
- Estimated Expiration
- 2043-07-26
AI Technical Summary
Existing technologies for monitoring order anomalies in enterprise management suffer from insufficient monitoring efficiency and accuracy. Traditional manual review and threshold setting methods cannot identify complex anomalies, and data clustering algorithms are sensitive to parameters, leading to inaccurate results.
An order anomaly monitoring system based on enterprise management is adopted. Through data collection, feature space construction, optimal parameter acquisition, and data detection modules, clustering parameters are adaptively selected to achieve accurate clustering and anomaly monitoring of order information data.
It improves the accuracy and efficiency of order anomaly monitoring, can identify complex anomalies, realize personalized anomaly supervision and management, and optimizes the clustering effect of the supervision system.
Smart Images

Figure CN116739645B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data processing technology, and more specifically to an order anomaly monitoring system based on enterprise management. Background Technology
[0002] Order anomalies are a critical issue in enterprise management that requires close attention. They can lead to problems such as insufficient inventory, financial difficulties, and decreased customer satisfaction. Order anomalies can cause fluctuations in inventory levels, making inventory management difficult and increasing the risk of overstocking and stockouts. They can also lead to direct financial losses for the enterprise. Anomalies may even involve illegal activities, such as fraud, which could result in legal disputes, regulatory penalties, and compliance risks for the enterprise. To mitigate the harm and impact of order anomalies, enterprises typically establish internal review systems, traditionally using manual review or other methods to detect and warn of anomalies and take swift action. Currently, the traditional methods for monitoring order anomalies mainly involve manual review and threshold-based statistical analysis. Manual review is generally only suitable for small-scale enterprises, but as the number of orders increases, the efficiency and accuracy of this method cannot be guaranteed. The second common method involves setting thresholds for certain key indicators. When an order's indicator exceeds the threshold, the system automatically triggers an alert. However, this method may not be able to identify complex anomalies hidden within, i.e., it cannot analyze and judge specific anomalies. This type of threshold setting method can only monitor anomalies for one type of indicator without linking and analyzing the various indicators, thus failing to obtain the true degree of anomaly for each abnormal data.
[0003] Therefore, more and more companies are now using artificial intelligence and machine learning algorithms to monitor and manage order data anomalies; among which, data clustering algorithms are frequently used. The screening of outlier data is a density-based clustering algorithm that can effectively address the shortcomings of traditional monitoring methods. However, the construction of the feature space and the selection of parameter points are difficult challenges for the algorithm to choose and evaluate. Because the algorithm is highly sensitive to parameters, different parameter settings will lead to significantly different clustering results, which in turn will result in inaccurate screening results for outlier data. Summary of the Invention
[0004] This invention provides an order anomaly monitoring system based on enterprise management to solve existing problems.
[0005] The order anomaly monitoring system based on enterprise management of the present invention adopts the following technical solution: One embodiment of the present invention provides an order anomaly monitoring system based on enterprise management, the system comprising: Data acquisition module: Acquires order information data from enterprises; Feature space construction module: Analyze the various indicator data in the enterprise's order information data to construct a clustering feature space; The optimal parameter acquisition module: It obtains the values of partition nodes based on various indicator data in the order information data; it adaptively acquires partition regions based on the partition node values to construct the clustering feature space, resulting in multiple clustering feature space regions and their indices; it selects different neighboring data points based on the indices of the clustering feature space regions, and obtains the neighborhood radius in the optimal clustering parameters of any clustering feature space region based on the distance between neighboring data points; it obtains the minimum number of data points within the neighborhood radius region in the optimal clustering parameters of any clustering feature space region based on the indices of the clustering feature space regions. The data detection module performs clustering operations on the enterprise's order information data based on the neighborhood radius and the minimum number of data points within the neighborhood radius, which are the optimal clustering parameters of the clustering feature spatial region, to obtain abnormal data and realize the supervision and management of enterprise order anomalies.
[0006] Preferably, the specific steps for constructing a clustering feature space from the various indicator data in the enterprise's order information data are as follows: The return rate of the indicator data is normalized and used as the vertical axis in the cluster feature space; the ratio of the total order amount to the number of orders is used as the horizontal axis in the cluster feature space; the vertical axis and the horizontal axis constitute the cluster feature space.
[0007] Preferably, the specific formula for obtaining the division node values based on various indicator data in the order information data is as follows: In the formula, These represent the values of the nodes that divide the cluster feature space along the horizontal and vertical axes, respectively. This represents the number of all order numbers in the company's order information data; This indicates the first item in the company's order information data. The return rate is a key characteristic indicator for each order number. This indicates the first item in the company's order information data. The number of orders corresponding to each order number; This indicates the first item in the company's order information data. The characteristic indicator data corresponding to each order number is the order transaction amount; It is a linear normalization function.
[0008] Preferably, the method for obtaining the plurality of clustering feature space regions and the index of each clustering feature space region is as follows: set up The index of the cluster feature spatial region. Values: ;set up Time represents the upper right region in the cluster feature space; set Time represents the lower right region in the cluster feature space; set Time represents the upper left region of the cluster feature space; set The region at time represents the lower left region in the cluster feature space; thus, four cluster feature space regions and their serial numbers are obtained.
[0009] Preferably, the specific steps of selecting different neighboring data points according to the index of the clustering feature space region, and obtaining the neighborhood radius in the optimal clustering parameters of any clustering feature space region based on the distance between neighboring data points, are as follows: Preset a parameter For any region in the cluster feature space Obtain the feature spatial region of this cluster. Inside each Data points To its nearest The Euclidean distance between the n nearest neighbor data points is denoted as the nth. Given a set of distances to a set of data points, sort the Euclidean distances of all data points in the set of distances to a set in ascending order. ; set From left to right, the values of adjacent elements are subtracted, and the maximum value of the subtraction is taken. The Euclidean distance corresponding to this maximum value is used as the feature space region of the cluster. Optimal clustering neighborhood radius parameter .
[0010] Preferably, the specific steps for obtaining the minimum number of data points within the neighborhood radius region in the optimal clustering parameters of any clustering feature space region based on the index of the clustering feature space region are as follows: The first in the cluster feature space The expression for calculating the minimum number of data points within the neighborhood radius of a region is: In the formula, Represents the first in the cluster feature space The minimum number of data points within the neighborhood radius of a given region; Represents the first in the cluster feature space One region; This represents the dimension of the cluster feature space.
[0011] The beneficial effects of the technical solution of the present invention are: the present invention is for The algorithm is improved to address the uncertainty of clustering results in enterprise order anomaly monitoring scenarios; parameters are analyzed and selected based on the scenario to adaptively obtain the optimal clustering parameters; ensuring... The clustering algorithm's clustering effect in the current order anomaly monitoring scenario aims to accurately detect order anomalies. The monitoring system will be optimized. Attached Figure Description
[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0013] Figure 1 This is a module framework diagram of the order anomaly monitoring system based on enterprise management of the present invention; Figure 2 This is a schematic diagram of the clustering feature space partitioning results of the order anomaly monitoring system based on enterprise management according to the present invention. Detailed Implementation
[0014] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of the order anomaly monitoring system based on enterprise management proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.
[0015] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.
[0016] The specific solution of the order anomaly monitoring system based on enterprise management provided by the present invention will be described in detail below with reference to the accompanying drawings.
[0017] Please see Figure 1 The diagram illustrates a module framework of an order anomaly monitoring system based on enterprise management, according to an embodiment of the present invention. The system includes the following modules: Data acquisition module: Acquires order information data from enterprises.
[0018] It should be noted that when retrieving order information forms stored in the enterprise system, these forms must contain important key data, such as order number and customer information. This includes data on metrics such as order amount, order quantity, and return rate, as well as metric information for each order number and customer. If the value of a certain metric corresponding to any customer's order number shows a significant anomaly, the customer's order is considered potentially abnormal. The degree of impact of anomalies varies depending on the specific metric data corresponding to a customer's order number.
[0019] Specifically, therefore, according to the time sequence of the order number, the data of each indicator corresponding to the order number in the enterprise's order information data are stored separately; that is, a system is established. The set is used to store the sequence of order amount data collected. , This indicates the first item in the company's order information data. The transaction amount corresponding to each order number This indicates the total number of order numbers in the company's order information data; [This is followed by a seemingly unrelated sentence: "Establish..."] The set is used to store the sequence of order quantity data collected. , This indicates the first item in the company's order information data. The number of orders corresponding to each order number This indicates the total number of order numbers in the company's order information data; this is used to collect and store the company's order information data.
[0020] This completes the collection of the company's order information data.
[0021] Feature space construction module: Analyze the various indicator data in the enterprise's order information data to construct a clustering feature space.
[0022] It should be noted that the construction of the feature space in the early stage of the clustering algorithm has a significant impact on the overall clustering result. In the traditional approach, it is difficult to guarantee the clustering effect by simply selecting any two correlated indicators to construct the horizontal and vertical coordinates in the feature space. Therefore, this embodiment needs to analyze the influence of the key indicator data of each order anomaly to comprehensively obtain the feature space construction parameters and adaptively obtain the optimal clustering parameters for the feature space. This allows anomalies with a small influence of the key indicator data of order anomalies to be classified as non-outliers, while anomalies with a large influence of the key indicator data of order anomalies can be classified as outliers. This results in a more accurate classification result, thereby achieving personalized and effective anomaly supervision and management.
[0023] Typically, constructing a cluster feature space requires determining the comparison parameters, i.e., the horizontal and vertical axis indicators. A generally effective approach is to select two feature indicators that are related to order anomalies; for example, the horizontal axis could be set as the order amount, and the vertical axis as the order quantity. In order to combine this implementation's anomaly monitoring scenario and obtain the optimal clustering effect under this scenario, it is necessary to analyze the various indicator data in the enterprise's order information data.
[0024] Specifically, the analysis is as follows: If there is a sudden increase or decrease in order volume among the order anomalies, it indicates abnormal fluctuations in order quantity, which may be caused by promotional activities, seasonal demand, or supply chain issues. This problem is generally not serious and is a relatively normal market change or difference in customer purchasing behavior. However, for anomalies in order amount, i.e., order amounts deviating significantly from the normal range, it may be due to incorrect pricing, improperly calculated discounts, or malicious tampering. Therefore, this type of order anomaly is more serious. Similarly, anomalies in return rate data are also a relatively serious order anomaly, indicating a sudden increase in the return rate of a particular product, suggesting product quality issues, inaccurate descriptions, or damage during logistics. Therefore, this embodiment needs to link the three types of characteristic indicator data together to comprehensively assess order anomalies.
[0025] This implementation takes the three feature index data as an example and uses the idea of dimensionality reduction to represent the three feature index data in a two-dimensional feature space. This can objectively and effectively reflect the abnormal attributes of the enterprise's order information data; thereby obtaining the clustering feature space parameters.
[0026] Specifically, the return rate of the characteristic indicator data is normalized and used as the vertical axis in the clustering feature space, that is... The ratio of total order amount to order quantity, representing the feature index data, is used as the horizontal axis in the clustering feature space. .
[0027] The Represents the vertical axis in the cluster feature space; The horizontal axis represents the cluster feature space. The return rate represents the characteristic indicator data. If the product return rate of any order is high, the vertical axis data of the data point marked for that order in the feature space will be larger, and it will be located at the top of the space. Conversely, if the return rate is low, it will be located at the bottom of the feature space. This represents the total order amount as a key performance indicator. This indicates the number of orders represented by the characteristic indicator data. This is the ratio of the two. It is a linear normalization function; it maps the result to the range {0~1}; the larger the value, the higher the individual amount of the order, and the larger the horizontal axis value of the data point marked by the order in the feature space, that is, it is located on the right side of the feature space; if the ratio is smaller, it means that the individual amount of the order is smaller, that is, the data point marked by the order in the feature space is located on the left side of the feature space.
[0028] It should be further noted that the method of constructing the cluster feature space in this embodiment is more effective than the traditional method of constructing the cluster feature space. It can better analyze the specific situation of the index anomaly by the distribution of data points in the feature space. Then, it can adaptively adjust the clustering rule parameters according to the degree of impact of different index anomalies to achieve better personalized clustering effect and accurately identify more serious order anomalies.
[0029] This completes the construction of the clustering feature space.
[0030] The module for obtaining optimal parameters adaptively partitions the clustering feature space and sets the optimal clustering parameter rules to obtain the optimal clustering parameters.
[0031] 1. Adaptively partition the cluster feature space.
[0032] It should be noted that since abnormal order information data accounts for a small proportion of the overall order information data, the average value of the horizontal axis indicator data and the average value of the vertical axis indicator data of all orders are used as the division criteria. The more the node deviates from the division value, the more likely the order indicator data is to be an abnormal data point. Thus, the cluster feature space is divided into regions.
[0033] Specifically, based on the above-mentioned clustering feature space construction principle, the partitioned regions are adaptively obtained; the threshold for partitioning numerical nodes is obtained: In the formula, These represent the values of the nodes that divide the cluster feature space along the horizontal and vertical axes, respectively. This represents the number of all order numbers in the company's order information data; This indicates the first item in the company's order information data. The return rate is a key characteristic indicator for each order number. This indicates the first item in the company's order information data. The number of orders corresponding to each order number; This indicates the first item in the company's order information data. The characteristic indicator data corresponding to each order number is the order transaction amount; It is a linear normalization function.
[0034] Based on the clustering feature space partitioning nodes obtained above, the clustering feature space is partitioned, and the schematic diagram of the clustering feature space partitioning result is shown below. Figure 2 As shown in the figure; To divide the cluster feature space into nodes, This refers to the index of the spatial region of the cluster feature.
[0035] At this point, the adaptive partitioning of the cluster feature space is complete, resulting in four regions of the cluster feature space and their sequence numbers.
[0036] 2. Obtain the neighborhood radius from the optimal clustering parameters for any region in the clustering feature space.
[0037] It should be noted that, due to the division of the cluster feature space into regions with varying degrees of influence, the... This represents the partitioned cluster feature space regions, with different cluster feature space regions corresponding to... The magnitude of the value varies, and its influence on the corresponding cluster feature space region also varies.
[0038] set up For the divided areas, Values: Its different values represent different spatial regions, and are set... The time represents the upper right region in the clustering feature space. This region has a high return rate and high order value. Anomalies in data points in this region have a significant impact on the company's order information data; (Setting...) The time represents the lower right region in the cluster feature space; this region has higher order amounts and its impact is less significant compared to the outlier data points in the upper right region; [Setting...] The term represents the upper left region of the cluster feature space; this region has a high return rate but relatively normal order amounts, and the impact of outlier data points in this region is relatively moderate. The time represents the lower left region in the cluster feature space; this region is where the order amount is relatively normal and the return rate is low. Abnormal data points in this region do not have a significant impact on the company's order information data.
[0039] Preset a parameter In this embodiment, The example is described, but this embodiment does not impose specific limitations. It depends on the specific implementation.
[0040] Specifically, optimal clustering parameters are set for each region in the clustering feature space; the neighborhood radius setting rule is as follows: For any region in the cluster feature space Obtain the feature spatial region of this cluster. Inside each Data points To its nearest The Euclidean distance between the n nearest neighbor data points is denoted as the nth. Given a set of distances to a set of data points, sort the Euclidean distances of all data points in the set of distances to a set in ascending order. ; set From left to right, the values of adjacent elements are subtracted, and the maximum value of the subtraction is taken. The Euclidean distance corresponding to this maximum value is used as the feature space region of the cluster. Optimal clustering neighborhood radius parameter .
[0041] The purpose of the above operations is as follows: If the presence of outlier data points in a certain cluster feature space region has a significant impact on the company's order information data, then the clustering algorithm for that cluster feature space region needs to maintain sensitivity to outliers, that is, set a smaller neighborhood decision radius. For spatial regions with a smaller impact, the clustering algorithm can appropriately maintain insensitivity to outliers, that is, expand the neighborhood radius and classify some data points that are discrete in principle but not very highly discrete as core points or boundary points within clusters.
[0042] Thus, the neighborhood radius in the optimal clustering parameters for any region of the clustering feature space is obtained.
[0043] 3. Obtain the minimum number of data points within the neighborhood radius region of the optimal clustering parameters for any region in the clustering feature space.
[0044] Specifically, optimal clustering parameters are set for each region in the clustering feature space; among them, the parameter for setting the minimum number of data points within the neighborhood radius region is: In the formula, Represents the first in the cluster feature space The minimum number of data points within the neighborhood radius of a given region; Represents the first in the cluster feature space One region; The dimension of the clustering feature space is represented in this embodiment. .
[0045] Thus, the neighborhood radius and the minimum number of data points within the neighborhood radius region are obtained as the optimal clustering parameters for any region in the clustering feature space.
[0046] Data detection module: Clusters the enterprise's order information data according to the optimal clustering parameters to identify abnormal data.
[0047] Based on the above, any region of the clustering feature space is obtained. Optimal clustering parameters and Perform clustering on each region of the obtained feature space partitioning Clustering operation, and the final Outliers obtained from clustering are statistically analyzed, and the corresponding abnormal orders in the enterprise's order information data are monitored in real time; while for The core points and boundary points obtained by the clustering algorithm are identified as non-abnormal data points and therefore left unprocessed; thus, accurate personalized order anomaly monitoring and management can be achieved.
[0048] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. An order abnormality monitoring system based on business management, characterized by, The system includes the following modules: Data acquisition module: Acquires order information data from enterprises; Feature space construction module: Analyze the various indicator data in the enterprise's order information data to construct a clustering feature space; The optimal parameter acquisition module: It obtains the values of partition nodes based on various indicator data in the order information data; it adaptively acquires partition regions based on the partition node values to construct the clustering feature space, resulting in multiple clustering feature space regions and their indices; it selects different neighboring data points based on the indices of the clustering feature space regions, and obtains the neighborhood radius in the optimal clustering parameters of any clustering feature space region based on the distance between neighboring data points; it obtains the minimum number of data points within the neighborhood radius region in the optimal clustering parameters of any clustering feature space region based on the indices of the clustering feature space regions. Data detection module: Based on the neighborhood radius and the minimum number of data points within the neighborhood radius in the optimal clustering parameters of the clustering feature spatial region, the module performs clustering operations on the enterprise's order information data to obtain abnormal data, thereby enabling the monitoring and management of enterprise order anomalies. In this cluster feature space, the return rate of the indicator data is normalized and used as the vertical axis; the ratio of the total order amount to the number of orders of the indicator data is used as the horizontal axis; the vertical axis and the horizontal axis constitute the cluster feature space. wherein, setting the serial number of the clustering feature space region, value: ; setting indicates the upper right region in the clustering feature space; setting indicates the lower right region in the clustering feature space; setting indicates the upper left region in the clustering feature space; setting indicates the lower left region in the clustering feature space; thus obtaining four clustering feature space regions and the serial numbers of the clustering feature space regions thereof; Among them, a preset parameter For any region in the cluster feature space Obtain the feature spatial region of this cluster. Inside each Data points To its nearest The Euclidean distance between the n nearest neighbor data points is denoted as the nth. Given a set of distances to a set of data points, sort the Euclidean distances of all data points in the set of distances to a set in ascending order. ; set From left to right, the values of adjacent elements are subtracted, and the maximum value of the subtraction is taken. The Euclidean distance corresponding to this maximum value is used as the feature space region of the cluster. Optimal clustering neighborhood radius parameter ; Among them, the first in the cluster feature space The expression for calculating the minimum number of data points within the neighborhood radius of a region is: In the formula, Represents the first in the cluster feature space The minimum number of data points within the neighborhood radius of a given region; Represents the first in the cluster feature space One region; This represents the dimension of the cluster feature space.
2. The order anomaly monitoring system based on enterprise management according to claim 1, characterized in that, The specific formula for obtaining the division node values based on various indicator data in the order information data is as follows: In the formula, These represent the values of the nodes that divide the cluster feature space along the horizontal and vertical axes, respectively. This represents the number of all order numbers in the company's order information data; This indicates the first item in the company's order information data. The return rate is a key characteristic indicator for each order number. This indicates the first item in the company's order information data. The number of orders corresponding to each order number; This indicates the first item in the company's order information data. The characteristic indicator data corresponding to each order number is the order transaction amount; It is a linear normalization function.
Citation Information
Patent Citations
Taxi abnormal order identification method based on density clustering method
CN113673571A
