Logistics order abnormity processing system and method

Abstract

The invention belongs to the technical field of logistics management, and provides a logistics order exception handling system and method, and the method comprises the steps: carrying out the clustering through the historical occurrence frequency and historical average tracing time of an exception source link, obtaining a plurality of exception source class groups, constructing a corresponding key clue library for each exception source class group, the method comprises the following steps of: tracing a current logistics abnormal order according to an abnormal source category group, setting a monitoring period, monitoring a tracing process in real time, dynamically distributing human resources in the tracing process, and calculating an abnormal key element proportion and a relation chain matching degree of the abnormal source category group; according to the method, the height anomaly category group is identified according to the proportion of the abnormal key elements, the target relation chains are positioned according to the relation chain matching degree, secondary allocation of human resources is performed, each target relation chain is verified, and the final abnormal relation chain is determined, so that the efficiency and accuracy of logistics order anomaly processing are improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of logistics management, in particular to a logistics order exception processing system and method. BACKGROUND

[0002] With the rapid development of e-commerce, the business volume of the logistics industry has increased dramatically, and various abnormal situations inevitably occur in the circulation process of logistics orders, such as damage to goods, delayed delivery, and information errors. The traditional logistics order exception processing method relies on manual investigation, which has the problems of low processing efficiency, unreasonable resource allocation, inaccurate abnormal source positioning, etc.

[0003] The existing technology for processing logistics order exceptions often analyzes independently for a single abnormal link, lacks systematic classification of abnormal source links, leads to repeated processing of similar problems, wastes a lot of manpower and resources, and in the abnormal tracing process, the resource allocation method is fixed and cannot be dynamically adjusted according to the tracing progress, resulting in waste of resources in low-potential tracing directions and insufficient resources in high-potential directions, affecting the efficiency of abnormal processing. In addition, there is a lack of quantitative standard for identifying the cause-and-effect relationship chain, making it difficult to quickly and accurately locate the final abnormal relationship chain.

[0004] Therefore, the present application provides a logistics order exception processing system and method. SUMMARY

[0005] In order to make up for the shortcomings of the prior art and solve at least one technical problem raised in the background art.

[0006] The technical scheme adopted by the present application to solve its technical problems is: a logistics order exception processing method, comprising: Step one: clustering the abnormal source links according to the historical occurrence frequency and historical average tracing time of the abnormal source links, to obtain a plurality of abnormal source category groups; Step two: extracting the key elements of all detail relationship chains in each abnormal source category group, and constructing a corresponding key clue library; Step three: tracing the current logistics abnormal order according to the abnormal source category group, and monitoring the tracing process in real time, and dynamically allocating human resources to the abnormal source category group during the monitoring process; Step four: after each monitoring period ends, calculating the abnormal key element proportion and relationship chain matching degree of the abnormal source category group, identifying the high abnormality category group and locating the target relationship chain; Step five: according to the relationship chain matching degree, secondarily allocating human resources to each target relationship chain, and verifying each target relationship chain to determine the final abnormal relationship chain.

[0007] Further, the process of clustering the abnormal source links to obtain a plurality of abnormal source category groups is: Based on any one abnormal source link: The historical abnormal occurrence frequency and the historical average tracing time are standardized as features to construct a feature vector; Determine the number of clusters K according to the elbow method; Randomly select K feature values of the links as initial cluster centers, calculate the Euclidean distance between each abnormal source link and the cluster center, and assign the abnormal source link to the cluster with the nearest Euclidean distance; After all the abnormal source links are assigned to the cluster, the average value of the feature values of all the abnormal source links in each cluster is recalculated as the new cluster center; Repeat the assignment of the abnormal source links and the update of the cluster centers, record the change value of each cluster center, and compare it with the preset threshold value, when the change value is less than the preset threshold value, the clustering is completed, and each cluster is an abnormal source category group.

[0008] Further, the process of constructing the corresponding key clue library includes: Extract all the nodes in each abnormal source category group, i.e. key elements; The structure of the key clue library is: {cluster label: {root cause element: [e0], transmission element: [e1, e2,..., e m ], result element: [e m+1 ]}}; Wherein, e0 is the initial node triggering the abnormality, e1 to e m are intermediate causal nodes, and e m+1 is the final abnormality.

[0009] Further, the process of dynamically allocating human resources to the abnormal source category group in the monitoring process is: According to the corresponding tracing mode of the current abnormal type, the tracing of each abnormal source category group is started synchronously, and the initial human resources are allocated to each abnormal source category group; Set the monitoring period to monitor the tracing process of each abnormal source category group in real time, and record the abnormal key elements traced; After each monitoring period, the number of abnormal key elements of each abnormal source category group is counted, and the amount of human resources allocated to the abnormal source category group after dynamic allocation adjustment is determined according to the number of abnormal key elements of the abnormal source category group.

[0010] Further, the acquisition method of the amount of human resources is: Count the number of abnormal key elements of each abnormal source category group and sum them up to get the total number of abnormal key elements; Obtaining the proportion of the number of abnormal key elements of the abnormal source category group to the total number of abnormal key elements, to obtain the distribution coefficient of the abnormal source category group; Multiplying the distribution coefficient of each abnormal source category group with the total amount of human resources to obtain the amount of human resources after dynamic distribution adjustment of the abnormal source category group.

[0011] Further, the way of identifying the high abnormality category group is After each monitoring period, the proportion of the number of abnormal key elements of each abnormal source category group in the corresponding key clue library is calculated to obtain the proportion of abnormal key elements; Comparing the proportion of abnormal key elements with the preset proportion, if the proportion of abnormal key elements is greater than the preset proportion, the category group is marked as a high abnormality category group.

[0012] Further, the process of positioning the target relationship chain is: Based on any one detail relationship chain, the key elements contained in the detail relationship chain are summarized to obtain the key element library corresponding to the detail relationship chain; Comparing all abnormal key elements with the key element library corresponding to the detail relationship chain to obtain the relationship chain matching degree; Comparing the relationship chain matching degree with the preset matching degree, extracting the detail relationship chain with the relationship chain matching degree greater than the preset relationship chain matching degree as the target relationship chain.

[0013] Further, the way of obtaining the relationship chain matching degree is: Based on any one key element in the key element library corresponding to the detail relationship chain; In all abnormal key elements, if there is an abnormal key element of the same type as the key element, the key element is marked as a coincident key element; In all abnormal key elements, if there is no abnormal key element of the same type as the key element, the key element is marked as a non-coincident key element; Statistical the proportion of the number of coincident key elements in the key element library corresponding to the detail relationship chain to obtain the relationship chain matching degree.

[0014] Further, the way of secondary distribution of human resources according to the relationship chain matching degree of each target relationship chain is: Summing up the relationship chain matching degrees of all target relationship chains to obtain the total relationship chain matching degree; For any one target relationship chain, the relationship chain matching degree is proportionally calculated with the total relationship chain matching degree to obtain the relationship chain matching degree proportion; The amount of human resources allocated to each target relationship chain is the product of the total amount of human resources and the relationship chain matching degree proportion.

[0015] A logistics order exception processing system comprises the following modules: A clustering module: clustering abnormal source links according to historical occurrence frequency and historical average tracing time of the abnormal source links, to obtain a plurality of abnormal source category groups; A key clue library construction module: extracting key elements of all detail relationship chains in each abnormal source category group, and constructing a corresponding key clue library; A tracing and resource dynamic allocation module: tracing the current logistics abnormal order according to the abnormal source category group, and monitoring the tracing process in real time, and dynamically allocating human resources to the abnormal source category group during the monitoring process; A high abnormality identification and target positioning module: after each monitoring period ends, calculating the abnormal key element proportion and relationship chain matching degree of the abnormal source category group, identifying a high abnormality category group and positioning a target relationship chain; A secondary allocation and verification module: secondary allocating human resources to each target relationship chain according to the relationship chain matching degree, and verifying each target relationship chain to determine a final abnormal relationship chain.

[0016] The beneficial effects of the present application are as follows: By clustering abnormal source links, the dispersed abnormal source links are classified according to characteristics, avoiding the redundancy of managing each link separately, realizing the same problem and the same processing, and providing a scientific basis for resource allocation, facilitating the reasonable tilt of initial resources, integrating the dispersed causal chain nodes into structured information, avoiding starting from zero to comb the logic each time, and recording the belonging link and the associated causal chain of the key elements, providing support for the rapid matching of elements in real-time tracing, improving the tracing efficiency, dynamically allocating human resources in the tracing process, being able to timely adjust the tracing strategy, investing more resources into the promising tracing direction, adapting to the changes of tracing difficulty and progress of different category groups, ensuring efficient tracing, identifying a high abnormality category group through the abnormal key element proportion, positioning a target relationship chain through the relationship chain matching degree, and the progressive determination of the double threshold value ensures the accuracy of the tracing direction, accelerates the final positioning through resource concentration, shortens the abnormal processing period, secondary allocates human resources according to the relationship chain matching degree, preferentially supports the verification of the target relationship chain with high matching degree, improves the verification efficiency, determines the final abnormal relationship chain based on the verification result, ensures the accuracy of the result, forms a complete closed loop of clustering-tracing-positioning-verification, and improves the efficiency and accuracy of logistics order exception processing. BRIEF DESCRIPTION OF DRAWINGS

[0017] The present application will be further described below with reference to the accompanying drawings.

[0018] Figure 1is a step flow chart of a logistics order exception processing method according to Embodiment 1 of the present application; Figure 2 is a logic judgment chart of a logistics order exception processing method according to Embodiment 1 of the present application; Figure 3 is a program block diagram of a logistics order exception processing system according to Embodiment 2 of the present application. DETAILED DESCRIPTION

[0019] In order to make the technical means, creative features, purposes and effects of the present application easy to understand, the present application will be further described below in conjunction with specific embodiments.

[0020] Embodiment 1: Please refer to Figure 1 The logistics order exception processing method according to the present application comprises the following steps: Step 1: Clustering the abnormal source links according to the historical occurrence frequency and historical average tracing time of the abnormal source links, to obtain a plurality of abnormal source category groups; In Step 1, the process of clustering the abnormal source links comprises: The abnormal source links refer to the abnormal source links of historical same-type exceptions in a historical period; Obtaining the abnormal occurrence frequency and historical average tracing time of each abnormal source link in the occurrence of historical same-type exceptions; Based on the historical abnormal occurrence frequency and historical average tracing time of each abnormal source link, using the K-Means clustering algorithm to cluster the abnormal source links, the specific process comprises: Based on any one abnormal source link: After standardizing the historical abnormal occurrence frequency and historical average tracing time, taking them as features to construct a feature vector; Determining the clustering number K according to the elbow method, which specifically comprises: taking K from 1, calculating the error sum of squares SSE corresponding to each K value, the smaller the SSE, the more concentrated the same links, and taking K as the horizontal axis and SSE as the vertical axis to draw a graph, the inflection point at which the curve slows down is the optimal K value; According to the determined clustering number K, randomly selecting the feature values of K links as the initial cluster centers, calculating the Euclidean distance between each abnormal source link and the cluster center, and distributing the abnormal source link to the nearest cluster; After all the abnormal source links are distributed to the clusters, the average value of the feature values of all the abnormal source links in each cluster is recalculated as the new cluster center; Repeating the distribution of the abnormal source links and the update of the cluster centers, recording the change value of each cluster center, and comparing it with the preset threshold value, when the change value is less than the preset threshold value, the clustering is completed; Each cluster is a source of abnormal category group, and each source of abnormal category group contains several abnormal source links; For example, assume that the data related to the handling of "goods damage" exceptions by a certain logistics company in the past year is collected, involving 7 abnormal source links: | Abnormal source link | Historical frequency | Historical average traceability time | | Transportation link | 45 | 2.0 | | Sorting link | 43 | 1.5 | | Distribution link | 22 | 0.5 | | Warehouse link | 8 | 3.0 | | Order review link | 3 | 0.2 | | Packaging link | 15 | 1.0 | | Information input link | 5 | 0.8 | First, the original data is standardized by Z-score to obtain the feature vector: Transportation link: (1.50, 0.80), sorting link: (1.38, 0.28), distribution link: (0.14, -0.76), warehouse link: (-0.70, 1.84), order review link: (-0.99, -1.07), packaging link: (-0.28, -0.24), information input link: (-0.87, -0.45); K takes values from 1 to 5, respectively calculates the corresponding error sum of squares SSE, K = 1, SSE = 18.72, K = 2, SSE = 8.25, K = 3, SSE = 3.47, K = 4, SSE = 2.15, K = 5, SSE = 1.58; After drawing the K-SSE curve, when K = 3, an inflection point appears, so K = 3 is selected as the best clustering number; Assume that the following 3 links are randomly selected as initial cluster centers: Cluster 1 center: transportation link (1.50, 0.80); Cluster 2 center: distribution link (0.14, -0.76); Cluster 3 center: warehouse link (-0.70, 1.84); Calculate the Euclidean distance of each abnormal source link to the three cluster centers respectively, and assign it to the cluster with the closest Euclidean distance: the sorting link is assigned to cluster 1, the distribution link is assigned to cluster 2, the order review link is assigned to cluster 2, the packaging link is assigned to cluster 2, and the information input link is assigned to cluster 2; The first round of assignment results: cluster 1: transportation, sorting, cluster 2: distribution, order review, packaging, information input, cluster 3: warehouse; Compute the average of the features of all links in each cluster as the new cluster center: Cluster 1 new center: (1.44, 0.54), Cluster 2 new center: (-0.50, -0.63), Cluster 3 only warehouse link, center unchanged: (-0.70, 1.84); Reassign all links using the new cluster centers, find that the cluster attribution of all links is unchanged, and the cluster center change value is less than the preset threshold, and the clustering ends; Standardize the obtained cluster centers to obtain the cluster features in the original data space: |Cluster number|Contains links|Average abnormal frequency|Average traceability time (hours)|; |Cluster 1|Transportation, sorting|44|1.75|; |Cluster 2|Distribution, packaging, information entry, order review|11.25|0.63|; |Cluster 3|Warehouse|8|3.0|; According to the historical occurrence frequency and the historical average traceability time of the abnormal source link, the role of clustering the abnormal source link is: Role 1: Cluster the dispersed abnormal source links (such as transportation, sorting, and warehousing) according to the characteristics of “high incidence / low incidence” and “easy traceability / difficult traceability”, avoid the redundancy of managing each link separately, and realize the same problem with the same processing; Role 2: Provide a basis for resource allocation, and the average abnormal frequency and average traceability time of the cluster group can directly guide the initial resource tilt (such as reserving more resources for clusters with high frequency and long time-consuming); Step two: Based on the abnormal source category group, extract the key elements of all detail relationship chains in each abnormal source category group, and construct a corresponding key clue library; In step two, the construction process of the key clue library includes: Extract the nodes of all detail relationship chains in each abnormal source category group, that is, the key elements, and count the number of key elements of each category, wherein the detail relationship chain refers to the potential causal chain of each link abnormal, for example: For the transportation link, there are two detail relationship chains: Relationship chain 1: Driver temporary leave -〉vehicle scheduling delay -〉transportation departure time late -〉transportation time delay; Relationship chain 2: Weather warning -〉highway closure -〉transportation route change -〉transportation time extension -〉transportation delay; Summarize the key elements contained in all detail relationship chains in each abnormal source category group, and record the detail relationship chain and link to which each key element belongs; Hierarchically classify the key elements according to the causal role: Root element: The initial node e0 that triggers the abnormality; Conducting elements: intermediate causal nodes e1, e2,..., e m ; Resulting elements: final abnormal performance e m+1 ; After layering, the clue library structure is {cluster label: {root cause element: [e0], conducting elements: [e1, e2,..., e m ], resulting elements: [e m+1 ]}} for subsequent matching degree calculation; The detail relationship chain is: e0->e1->e2->->e m ->e m+1 ; It should be noted that the key element is the core node in the abnormal causal chain, and is the specific investigation object during tracing, and the detail relationship chain refers to the complete causal chain from the source to the result of the abnormality of a certain link, reflecting the propagation path of the abnormality; The role of constructing the key clue library is: Role 1: Integrating scattered causal chain nodes into a structured clue library to avoid starting from scratch to sort out the logic each time; Role 2: Recording the belonging link and associated causal chain of the key element to provide support for the rapid matching of elements in real-time tracing; Step three: tracing the current logistics abnormal order according to the abnormal source category group, and real-time monitoring the tracing process, and dynamically allocating human resources to the abnormal source category group in the monitoring process; In step three, the process of dynamically allocating human resources to the abnormal source category group in the monitoring process includes: Statistical current available for abnormal tracing human resource quantity; Among them, human resources include but are not limited to tracing personnel, system permissions, data interface, tracing personnel is responsible for querying data, verifying information, system permissions refer to the permission of calling sorting record, transportation GPS data, etc., data interface refers to the interface of warehousing system, personnel management system, etc.; According to the current abnormal type, adopt corresponding tracing mode, start tracing for each abnormal source category group at the same time, initially allocate human resources to each abnormal source category group on an average basis, and ensure that each abnormal source category group starts tracing at the same time; Set the monitoring period, real-time monitor the tracing process of each abnormal source category group, and record the number of abnormal key elements traced by each abnormal source category group in this monitoring period; Statistical abnormal key element number of each abnormal source category group, and sum to get the total number of abnormal key elements; Obtain the proportion of the number of abnormal key elements of the abnormal source category group and the total number of abnormal key elements, and obtain the allocation coefficient of the abnormal source category group; The allocation coefficient of each abnormal source category group is multiplied by the total amount of human resources to obtain the amount of human resources allocated dynamically after adjustment of the abnormal source category group; The role of dynamically allocating human resources in the tracing process includes: Role 1: By dynamically allocating human resources, the tracing strategy can be adjusted in a timely manner, and more resources can be invested in promising tracing directions to increase the likelihood of finding the root cause of the anomaly; Role 2: During the tracing process, the tracing difficulty and progress of different category groups may change, and dynamic resource allocation can better adapt to these changes to ensure that the tracing work remains efficient; Step four: After the end of each monitoring period, calculate the proportion of abnormal key elements and the matching degree of the relationship chain for the abnormal source category group, identify the highly abnormal category group, and locate the target relationship chain; Please refer to Figure 2 In step four, the identification process of the highly abnormal category group includes: After the end of each monitoring period: Calculate the proportion of the number of abnormal key elements in each category group in the corresponding key clue library to obtain the proportion of abnormal key elements; Compare the proportion of abnormal key elements with the preset proportion. If the proportion of abnormal key elements is greater than the preset proportion, mark the category group as a highly abnormal category group; Stop the tracing of other category groups and invest all human resources in the tracing of the category group marked as a highly abnormal category group; In step four, the positioning method of the target relationship chain includes: Based on any one detail relationship chain, aggregate the key elements contained in the detail relationship chain to obtain a key element library corresponding to the detail relationship chain; Based on any one key element in the key element library corresponding to the detail relationship chain; Among all abnormal key elements, if there is an abnormal key element of the same type as the key element, mark the key element as a coincident key element; Among all abnormal key elements, if there is no abnormal key element of the same type as the key element, mark the key element as a non-coincident key element; Calculate the proportion of the number of coincident key elements in the key element library corresponding to the detail relationship chain to obtain the relationship chain matching degree; Compare the relationship chain matching degree with the preset relationship chain matching degree, and extract the detail relationship chain with a relationship chain matching degree greater than the preset relationship chain matching degree as the target relationship chain; It can be understood that the physical meaning of the relationship chain matching degree is: by calculating the proportion of the current abnormal key elements in the detail relationship chain, the close degree of the relationship chain and the actual abnormality is quantified. The higher the matching degree, the more the key elements contained in the relationship chain coincide with the key elements found in the current abnormality, and the greater the possibility of reflecting the actual abnormal causal path; The role of identifying the high abnormality category group and locating the target relationship chain is: Role 1: Focus first through the key element quantity threshold to avoid wasting resources in the low potential category group; Role 2: Take the relationship chain matching degree as the termination standard to ensure the integrity and reliability of the target relationship chain; Role 3: The progressive determination of the double threshold ensures the accuracy of the tracing direction and accelerates the final positioning through resource concentration, which can shorten the abnormality processing period; Step five: according to the relationship chain matching degree of each target relationship chain, the human resources are secondarily allocated, and each target relationship chain is verified to determine the final abnormal relationship chain; In step five, the process of secondarily allocating human resources according to the relationship chain matching degree of each target relationship chain includes: Sum the relationship chain matching degrees of all target relationship chains to obtain the total relationship chain matching degree; For any target relationship chain, the relationship chain matching degree is proportionally calculated with the total relationship chain matching degree to obtain the relationship chain matching degree proportion; The amount of human resources allocated to each target relationship chain is the product of the total amount of human resources and the relationship chain matching degree proportion; According to the allocated human resources, the key elements in the target relationship chain that have not been traced are verified; If all the untraced elements in the target relationship chain are verified as abnormal, the target relationship chain is directly determined as the final abnormal relationship chain; If there are untraced elements in the target relationship chain that are verified as non-abnormal, the element is deleted and the relationship chain with the highest relationship chain matching degree is selected as the final abnormal relationship chain; It should be noted that the role of the relationship chain matching degree is: Role 1: As the basis for secondary allocation of resources, by calculating the relationship chain matching degree proportion of each target relationship chain, the human resources are tilted towards the relationship chain with higher matching degree, and the relationship chain with closer association with the actual abnormality is preferentially supported for verification, improving the verification efficiency; Role 2: As a reference for determining the final relationship chain, during the verification process, the relationship chain with a continuously high matching degree is more likely to completely reflect the causal logic of the abnormality. Finally, based on the matching degree and the verification result, the relationship chain with the highest matching degree and the verification result is determined as the final abnormal relationship chain, ensuring the accuracy of the result.

[0021] The technical scheme and advantages of the embodiments of the present application are as follows: According to the historical occurrence frequency and the historical average tracing time of the abnormal source link, the abnormal source link is clustered to obtain a plurality of abnormal source category groups, key elements of all detail relationship chains in each abnormal source category group are extracted, a corresponding key clue library is constructed, the current logistics abnormal order is traced according to the abnormal source category group, and the tracing process is monitored in real time, the human resources of the abnormal source category group are dynamically allocated in the monitoring process, the abnormal key element proportion calculation and the relationship chain matching degree calculation of the abnormal source category group are performed after each monitoring period ends, the highly abnormal category group is identified and the target relationship chain is located, the human resources are secondarily allocated according to the relationship chain matching degree of each target relationship chain, each target relationship chain is verified, and the final abnormal relationship chain is determined. The historical occurrence frequency and the historical average tracing time of the source link are clustered to obtain a plurality of abnormal source category groups, a corresponding key clue library is constructed for each abnormal source category group, the current logistics abnormal order is traced according to the abnormal source category group, and a monitoring period is set. The tracing process is monitored in real time, the human resources are dynamically allocated in the tracing process, the abnormal key element proportion calculation and the relationship chain matching degree calculation of the abnormal source category group are performed, the highly abnormal category group is identified according to the abnormal key element proportion, the target relationship chain is located according to the relationship chain matching degree and the human resources are secondarily allocated, each target relationship chain is verified, and the final abnormal relationship chain is determined. The efficiency and accuracy of the logistics order abnormal processing are improved.

[0022] Embodiment 2: Please refer to Figure 3 The logistics order abnormal processing system provided by the embodiments of the present application includes the following modules: The clustering module clusters the abnormal source link according to the historical occurrence frequency and the historical average tracing time of the abnormal source link to obtain a plurality of abnormal source category groups; The process of clustering the abnormal source link includes: The abnormal source link refers to the abnormal source link of the historical same type of abnormality in the historical period; The occurrence frequency of each abnormal source link in the historical same type of abnormality and the historical average tracing time are obtained; Based on the historical abnormal occurrence frequency and the historical average tracing time of each abnormal source link, the K-Means clustering algorithm is used to cluster the abnormal source link, and the specific process includes: Based on any one abnormal source link: The historical abnormal occurrence frequency and the historical average tracing time are standardized as features to construct a feature vector; The elbow method is used to determine the number of clusters K, specifically including: K is taken from 1, the sum of squared errors SSE corresponding to each K value is calculated, the smaller the SSE, the more concentrated the same type of link, and a plot is drawn with K as the horizontal axis and SSE as the vertical axis, and the inflection point at which the curve slows down is the optimal K value; According to the determined number of clusters K, the characteristic values of K links are randomly selected as initial cluster centers, the Euclidean distance between each abnormal source link and the cluster center is calculated, and the abnormal source link is assigned to the nearest cluster; After all the abnormal source links are assigned to the cluster, the average value of the characteristic values of all the abnormal source links in each cluster is recalculated as the new cluster center; Repeat the assignment of the abnormal source link and the update of the cluster center, record the change value of each cluster center, and compare it with the preset threshold value, when the change value is less than the preset threshold value, the clustering is completed; Each cluster is an abnormal source category group, and each abnormal source category group contains a plurality of abnormal source links; The key clue library construction module: based on the abnormal source category group, the key elements of all detail relationship chains in each abnormal source category group are extracted, and the corresponding key clue library is constructed; The construction process of the key clue library includes: Extract the nodes of all detail relationship chains in each abnormal source category group, that is, the key elements, and count the number of key elements of each category, wherein the detail relationship chain refers to the potential causal chain of each link anomaly, for example: For the transportation link, there are two detail relationship chains: Relationship chain 1: driver temporary leave -〉vehicle dispatch delay -〉transportation delivery time late -〉transportation time delay; Relationship chain 2: weather warning -〉highway closure -〉transportation route change -〉transportation time extension -〉transportation delay; All key elements contained in all detail relationship chains in each abnormal source category group are summarized, and the detail relationship chain and the link to which each key element belongs are recorded; The key elements are layered according to the causal role: Root element: initial node e0 triggering the anomaly; Conductive element: intermediate causal nodes e1, e2,...,e m ; Result element: final abnormal performance e m+1 ; After layering, the clue library structure is {cluster label: {root element: [e0], conductive element: [e1, e2,...,e m ], result element: [e m+1 ]}}, which is convenient for subsequent matching degree calculation; The detail relationship chain is: e0->e1->e2->->e m ->e m+1 ; It should be noted that the key element is the core node in the abnormal causal chain, and is the specific investigation object during the tracing, the detail relationship chain refers to the complete causal chain from the source to the result of the abnormal link, and reflects the propagation path of the abnormality; The tracing and resource dynamic allocation module: tracing the current logistics abnormal order according to the abnormal source category group, and monitoring the tracing process in real time, and dynamically allocating human resources to the abnormal source category group in the monitoring process; The process of dynamically allocating human resources to the abnormal source category group in the monitoring process includes: Statistical current available for abnormal tracing human resource quantity; Among them, human resources include but are not limited to tracing personnel, system permissions, data interface, tracing personnel are responsible for querying data, verifying information, system permissions refer to the permissions of calling sorting records, transportation GPS data, etc., and data interface refers to the interface of warehouse system, personnel management system, etc.; According to the current abnormal type, the corresponding tracing mode is adopted, each abnormal source category group starts tracing synchronously, and the human resources are initially allocated to each abnormal source category group, so as to ensure that each abnormal source category group starts tracing synchronously; Set the monitoring period, monitor the tracing process of each abnormal source category group in real time, and record the number of abnormal key elements traced by each abnormal source category group in this monitoring period; Statistical abnormal key element number of each abnormal source category group, and sum to get the total number of abnormal key elements; Obtain the proportion of the number of abnormal key elements of the abnormal source category group and the total number of abnormal key elements, and obtain the allocation coefficient of the abnormal source category group; Multiply the allocation coefficient of each abnormal source category group with the total amount of human resources respectively to obtain the amount of human resources allocated dynamically to the abnormal source category group after adjustment; High abnormality identification and target positioning module: after the end of each monitoring period, the proportion of abnormal key elements of the abnormal source category group is calculated, and the matching degree of the relationship chain is calculated, the high abnormality category group is identified, and the target relationship chain is positioned; Please refer to Figure 2 The identification process of the high abnormality category group includes: After the end of each monitoring period: Statistical proportion of the number of abnormal key elements of each category group in the corresponding key clue library to obtain the proportion of abnormal key elements; The proportion of the abnormal key element is compared with the preset proportion. If the proportion of the abnormal key element is greater than the preset proportion, the category group is marked as a highly abnormal category group. The tracing of other category groups is stopped, and all human resources are invested in tracing the category group marked as the highly abnormal category group. The positioning mode of the target relationship chain comprises: Based on any one detail relationship chain, the key elements contained in the detail relationship chain are summarized to obtain a key element library corresponding to the detail relationship chain. Based on any one key element in the key element library corresponding to the detail relationship chain. Among all the abnormal key elements, if there is an abnormal key element of the same type as the key element, the key element is marked as a coincident key element. Among all the abnormal key elements, if there is no abnormal key element of the same type as the key element, the key element is marked as a non-coincident key element. The number proportion of the coincident key element in the key element library corresponding to the detail relationship chain is counted to obtain a relationship chain matching degree. The relationship chain matching degree is compared with a preset relationship chain matching degree, and a detail relationship chain with a relationship chain matching degree greater than the preset relationship chain matching degree is extracted as a target relationship chain. The secondary allocation and verification module: according to the relationship chain matching degree, the human resources are secondarily allocated to each target relationship chain, and each target relationship chain is verified to determine a final abnormal relationship chain. The secondary allocation process of human resources according to the relationship chain matching degree of each target relationship chain comprises: The relationship chain matching degrees of all target relationship chains are summed to obtain a total relationship chain matching degree. For any one target relationship chain, the relationship chain matching degree and the total relationship chain matching degree are proportionally calculated to obtain a relationship chain matching degree proportion. The amount of human resources allocated to each target relationship chain is the product of the total amount of human resources and the relationship chain matching degree proportion. According to the allocated human resources, the key elements in the target relationship chain that have not been traced are verified. If all the untraced elements in the target relationship chain are verified as abnormal, the target relationship chain is directly determined as a final abnormal relationship chain. If there is an untraced element in the target relationship chain that is verified as non-abnormal, the element is deleted and the relationship chain with the highest relationship chain matching degree is selected as the final abnormal relationship chain.

[0023] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above-mentioned embodiments, and the above-mentioned embodiments and descriptions in the specification are only to illustrate the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for handling logistics order exceptions, characterized by: include: Step 1: Cluster the abnormal source links according to their historical occurrence frequency and historical average tracing time to obtain multiple abnormal source category groups; Step 2: Extract the key elements of all detail relationship chains within each abnormal source category group and build the corresponding key clue library; Step 3: Trace the current logistics abnormal orders according to the abnormal source category group, and monitor the tracing process in real time. During the monitoring process, dynamically allocate human resources to the abnormal source category group; Step 4: After each monitoring cycle, calculate the abnormal key element ratio and relationship chain matching degree for the abnormal source category group, identify highly abnormal category groups and locate the target relationship chain; Step 5: Perform secondary allocation of human resources for each target relationship chain based on the relationship chain matching degree, verify each target relationship chain, and determine the final abnormal relationship chain.

2. A logistics order exception handling method according to claim 1, characterized in that: The process of clustering the abnormal source links to obtain multiple abnormal source category groups is as follows: Based on any abnormal source link: The historical anomaly frequency and the historical average tracing time are standardized as features to construct a feature vector; Determine the number of clusters K according to the elbow method; Randomly select the characteristic values ​​of K links as the initial cluster center, calculate the Euclidean distance between each abnormal source link and the cluster center, and assign the abnormal source link to the cluster with the closest Euclidean distance; When all abnormal source links are assigned to clusters, the average value of the characteristic values ​​of all abnormal source links in each cluster is recalculated as the new cluster center; Repeat the allocation of abnormal source links and the update of cluster centers, record the change value of each cluster center, and compare it with the preset threshold. When the change value is less than the preset threshold, clustering is completed, and each cluster is an abnormal source category group.

3. A logistics order exception handling method according to claim 1, characterized in that: The process of constructing the corresponding key clue library includes: Extract all nodes of detail relationship chains in each abnormal source category group, i.e. key elements; The structure of the key clue library is: {cluster label: {root element: [e0], transmission element: [e1,e2,...,e m ], resulting element: [e m+1 ]}}; Among them, e0 is the initial node that triggers the exception, e1 to e m is the intermediate causal node, e m+1 The final abnormal performance.

4. A logistics order exception handling method according to claim 1, characterized in that: The process of dynamically allocating human resources to abnormal source category groups during the monitoring process is as follows: Adopt the corresponding tracing method based on the current exception type, and simultaneously start tracing for each exception source category group. Initially, human resources are evenly distributed for each exception source category group. Set a monitoring cycle, conduct real-time monitoring of the tracing process for each abnormal source category group, and record the key elements of the abnormalities traced; After each monitoring cycle, the number of abnormal key elements in each abnormal source category group is counted, and the amount of human resources after dynamic allocation adjustment of the abnormal source category group is determined based on the number of abnormal key elements in the abnormal source category group.

5. A logistics order exception handling method according to claim 4, characterized in that: The method for obtaining the human resources is as follows: Count the number of abnormal key elements in each abnormal source category group and sum them up to get the total number of abnormal key elements; Calculate the ratio of the number of abnormal key elements in the abnormal source category group to the total number of abnormal key elements to obtain the distribution coefficient of the abnormal source category group; The allocation coefficient of each abnormal source category group is multiplied by the total human resources to obtain the human resources after dynamic allocation adjustment of the abnormal source category group.

6. A logistics order exception handling method according to claim 1, characterized in that: The method of identifying highly abnormal category groups is as follows: After each monitoring cycle, the proportion of abnormal key elements in each abnormal source category group in the corresponding key clue library is counted to obtain the abnormal key element ratio; The proportion of abnormal key elements is compared with the preset proportion. If the proportion of abnormal key elements is greater than the preset proportion, the abnormal source category group is marked as a highly abnormal category group.

7. A logistics order exception handling method according to claim 1, characterized in that: The process of locating the target relationship chain is as follows: Based on any detail relationship chain, the key elements contained in the detail relationship chain are summarized to obtain the key element library corresponding to the detail relationship chain; Compare all abnormal key elements with the key element library corresponding to the detail relationship chain to obtain the relationship chain matching degree; The relationship chain matching degree is compared with the preset relationship chain matching degree, and the detail relationship chain whose relationship chain matching degree is greater than the preset relationship chain matching degree is extracted as the target relationship chain.

8. A logistics order exception handling method according to claim 7, characterized in that: The relationship chain matching degree is obtained as follows: Any key element in the key element library corresponding to the detail relationship chain; Among all abnormal key elements, if there is an abnormal key element of the same type as the key element, the key element will be marked as a coincident key element; Among all abnormal key elements, if there is no abnormal key element of the same type as the key element, the key element will be marked as a non-overlapping key element; The ratio of the number of overlapping key elements in the key element library corresponding to the detail relationship chain is counted to obtain the relationship chain matching degree.

9. A logistics order exception handling method according to claim 1, characterized in that: The secondary allocation method of human resources according to the relationship chain matching degree of each target relationship chain is: Sum up the relationship chain matching degrees of all target relationship chains to obtain the total relationship chain matching degree; For any target relationship chain, calculate the ratio of the relationship chain matching degree to the total relationship chain matching degree to obtain the relationship chain matching degree ratio; The amount of human resources allocated to each target relationship chain is the product of the total human resources and the relationship chain matching ratio.

10. A logistics order exception processing system includes the following modules: Clustering module: Clusters abnormal source links based on their historical occurrence frequency and historical average tracing time to obtain multiple abnormal source category groups; Key clue library construction module: extracts the key elements of all detail relationship chains within each abnormal source category group and constructs the corresponding key clue library; Tracing and dynamic resource allocation module: Tracing current logistics abnormal orders according to abnormal source category groups, and monitoring the tracing process in real time. During the monitoring process, dynamic allocation of human resources is performed for abnormal source category groups. High Anomaly Identification and Target Positioning Module: After each monitoring cycle, the module calculates the abnormal key element ratio and relationship chain matching degree of the abnormal source category group, identifies the highly abnormal category group, and locates the target relationship chain; Secondary allocation and verification module: Secondary allocation of human resources to each target relationship chain according to the relationship chain matching degree, and verification of each target relationship chain to determine the final abnormal relationship chain.

Images