Cross-regional delivery identification methods, devices, computer equipment, and storage media

Abstract

This application relates to the field of logistics distribution management technology, and provides a method, device, computer equipment, and storage medium for identifying cross-regional delivery. The method involves: filtering valid records of internal delivery from a waybill table, and removing full-truckload and internal component waybills; statistically analyzing waybills with the same trajectory from the waybill table by receiving area, with the receiving area having the most waybills being the primary delivery area; removing waybills from the primary delivery area from the waybill table, and determining whether the remaining waybills are cross-regional delivery issues; for each cross-regional delivery issue, obtaining an increment value based on the sum of the actual distances between the cross-regional delivery issue and its adjacent primary delivery area, and the straight-line distance between the cross-regional delivery issue and its corresponding adjacent primary delivery area, to generate an analysis result for each cross-regional delivery issue, thus completing the cross-regional delivery identification. This method is applicable to the intelligent scheduling and optimization of internal delivery operations in logistics enterprises, solving the problems of crude delivery identification and unreasonable resource allocation in existing technologies.

Description

Technical Field

[0001] This application relates to the field of logistics and distribution management technology, and in particular to a method, apparatus, computer equipment and storage medium for identifying cross-regional delivery. Background Technology

[0002] In the logistics and distribution sector, cross-regional delivery issues (i.e., goods being delivered to areas other than the primary area of ​​responsibility) lead to circuitous delivery routes, low vehicle loading efficiency, and increased delivery costs. Efficiently identifying cross-regional delivery issues and optimizing delivery strategies is a significant challenge for logistics scheduling and management. Traditional delivery identification methods typically rely on manual experience or simple rule-based filtering, such as rough classification based solely on the recipient's address and its assigned region, lacking systematic analysis of multi-dimensional data such as waybill trajectory, vehicle load, time, and distance.

[0003] In existing technologies, the determination of cross-regional waybills often relies on a single condition (such as address attribution), failing to form a comprehensive identification system that combines statistics of the main delivery area, vehicle weight utilization rate, and route distance optimization. This makes it difficult to accurately locate cross-regional issues and generate actionable consolidation delivery suggestions, leading to wasted logistics resources and low delivery efficiency. Therefore, there is an urgent need for a method that can integrate multi-dimensional data, systematically identify cross-regional delivery, and provide optimization strategies to solve the problems of crude delivery identification and unreasonable resource allocation in existing technologies. Summary of the Invention

[0004] This application provides a method, apparatus, computer equipment, and storage medium for cross-regional delivery identification, aiming to solve the problems of crude delivery identification and unreasonable resource allocation in the prior art.

[0005] Firstly, this application provides a method for identifying cross-regional delivery, including:

[0006] Filter the valid records in the waybill table that are delivered internally, and remove waybills for full vehicles and waybills for internal components.

[0007] The waybill table is used to count waybills with the same route by receiving area. The receiving area with the most shipments is the main delivery area. If there are shipments with the same number of shipments, the receiving area with the largest cargo weight is selected as the main delivery area.

[0008] Remove waybills from the main delivery area from the waybill table, and determine whether the remaining waybills are cross-regional problem tickets; among them, within the preset departure time range and straight-line distance range corresponding to the cross-regional problem tickets, find waybills whose vehicle weight ratio is less than or equal to the preset load capacity ratio, mark them as combinable tickets and generate a combined delivery suggestion.

[0009] For each out-of-area problem ticket, the distance increment value is obtained by combining the sum of the actual distances between the out-of-area problem ticket and the adjacent main delivery area with the straight-line distance between the out-of-area problem ticket and the adjacent main delivery area, so as to generate the analysis result corresponding to each out-of-area problem ticket and complete the out-of-area delivery identification.

[0010] In some embodiments, the step of filtering valid records with internal delivery as the delivery method from the waybill table and removing waybills of the full vehicle type and internal component type includes: performing a filtering operation on the waybill data in the waybill table to exclude waybills marked as full vehicle transportation or internal components in the transportation type field, and only retaining waybills with internal delivery as the delivery method field; for waybills with the same receipt date, delivery person number, and delivery vehicle trajectory number, selecting the unique record with the earliest delivery receipt time and status as receipt completed, and if the delivery receipt time field is missing, replacing it with the receipt time field value.

[0011] In some embodiments, the step of counting waybills with the same trajectory from the waybill table by receiving area and identifying the receiving area with the most waybills as the main delivery area includes: grouping waybills according to the delivery vehicle trajectory number; counting waybills in each group by the receiving area name field to generate the number of waybills for each receiving area; identifying the receiving area with the highest number of waybills as the main delivery area; and if multiple receiving areas have the same number of waybills, comparing the sum of the cargo weight fields of each receiving area and selecting the receiving area with the largest total cargo weight as the main delivery area.

[0012] In some embodiments, determining whether the remaining waybills are cross-regional issue tickets includes: for the remaining waybills after removing waybills from the main delivery area, extracting the delivery time field and the departure time field of the corresponding trajectory, and calculating the time difference; if the time difference is greater than 2 hours, and the sum of the actual driving distances from the latitude and longitude of the recipient address of the waybill to the latitude and longitude of the center of the adjacent main delivery area, minus the straight-line distance between the latitude and longitude of the center of the adjacent main delivery area, is greater than or equal to 10 kilometers, it is determined to be a cross-regional issue ticket; if waybills from the preceding and following main delivery areas are missing, the latitude and longitude registered at the delivery point are used as substitute coordinates for distance calculation.

[0013] In some embodiments, marking a ticket as a combinable ticket and generating a combined delivery suggestion includes: for each ticket with a cross-regional issue, searching for other waybills within a 2-hour period before and after the departure time of the corresponding trajectory, with a straight-line distance not exceeding 3 kilometers; obtaining the preset load capacity value corresponding to the delivery vehicle type, and calculating the ratio of the current vehicle weight to the preset load capacity value; if the ratio is less than or equal to 80%, marking the eligible waybill as a combinable ticket, and generating an optimal combined delivery route suggestion based on the delivery address of each waybill; wherein, the preset load capacity value corresponding to a van is 600 kg, the preset load capacity value corresponding to a medium-sized truck is 2 tons, the preset load capacity value corresponding to a 7.6-meter large truck is 8 tons, and the preset load capacity value corresponding to a 9.6-meter large truck is 10 tons.

[0014] In some embodiments, generating the analysis results corresponding to each cross-regional issue ticket to complete the cross-regional delivery identification includes: using the waybill receipt time field as the statistical time caliber and the delivery point number as the organizational affiliation unit to count the cross-regional issue tickets, calculating the number of cross-regional tickets, the proportion of tickets that can be combined, the average increment value, and the month-on-month change rate; associating and archiving the waybill number, increment value, approval mark, and optimization suggestions for each cross-regional issue ticket, and generating weekly or monthly statistical reports for use by the dispatch management system.

[0015] In some embodiments, the method further includes: constructing a historical cross-regional behavior dataset and inputting it into a machine learning model for training. The input features of the machine learning model include waybill weight, delivery time interval, driving distance difference, and vehicle load utilization rate. A dynamic judgment threshold is obtained through machine learning model training to automatically adjust preset parameters, including the difference between delivery time and departure time, distance difference, and vehicle load ratio. The machine learning model is incrementally updated based on real-time logistics data to optimize cross-regional delivery identification rules and improve the accuracy of abnormal behavior identification.

[0016] Secondly, this application provides a cross-regional delivery identification device, comprising:

[0017] The waybill rejection unit is used to filter valid records with internal delivery as the delivery method from the waybill table and reject waybills of the whole vehicle type and waybills of the internal component type.

[0018] The waybill statistics unit is used to count waybills with the same trajectory from the waybill table by receiving area. The receiving area with the most counted waybills is the main delivery area. If there are waybills with the same number of waybills, the receiving area with the largest cargo weight is selected as the main delivery area.

[0019] The problem judgment unit is used to remove waybills from the main delivery area in the waybill table and determine whether the remaining waybills are cross-regional problem tickets. Among them, within the preset departure time range and straight-line distance range corresponding to the cross-regional problem tickets, the unit finds waybills whose vehicle weight ratio is less than or equal to the preset load capacity ratio, marks them as combinable tickets and generates a combined delivery suggestion.

[0020] The identification completion unit is used to obtain the distance increment value for each cross-regional problem ticket based on the sum of the actual distances between the cross-regional problem ticket and the adjacent main delivery area and the straight-line distance between the cross-regional problem ticket and the adjacent main delivery area, so as to generate the analysis result corresponding to each cross-regional problem ticket and complete the cross-regional delivery identification.

[0021] Thirdly, this application also provides a computer device, comprising:

[0022] Memory and processor;

[0023] The memory is used to store computer programs;

[0024] The processor is configured to execute the computer program and, when executing the computer program, implement the steps of the cross-regional delivery identification method as described in the first aspect above.

[0025] Fourthly, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to implement the steps of the cross-area delivery identification method described in the first aspect above.

[0026] This application provides a method, apparatus, computer equipment, and storage medium for identifying cross-regional delivery. The method performs dual statistical analysis of the number of parcels and cargo weight for waybills with the same trajectory, ensuring the scientific validity of the primary delivery area determination and avoiding the bias caused by a single indicator (such as only the number of parcels). This lays an accurate foundation of data for subsequent cross-regional delivery identification. By combining preset departure time ranges, straight-line distance ranges, and vehicle weight-to-load ratios, it can not only identify cross-regional parcels with abnormal address attributions but also simultaneously filter out waybills that can be merged for delivery, achieving the integration of "problem identification" and "optimization suggestions," thus improving the efficiency of delivery resource integration. By calculating the distance increment value and generating analysis results, it provides data support for scheduling management, facilitating targeted optimization of delivery routes. Simultaneously, the merging suggestions for parcels can effectively reduce vehicle empty-running rates and lower delivery costs. Through statistical analysis of cross-regional problem parcels (such as the number of cross-regional parcels and the proportion of parcels that can be merged), a systematic management report is generated, helping logistics companies achieve digital and refined control of the delivery process.

[0027] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0028] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 This is a schematic flowchart illustrating the steps of a cross-regional delivery identification method provided in an embodiment of this application;

[0030] Figure 2 This is a schematic diagram of the structure of a cross-regional delivery identification device provided in one embodiment of this application;

[0031] Figure 3This is a schematic block diagram of the structure of a computer device provided in an embodiment of this application.

[0032] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Detailed Implementation

[0033] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0034] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0035] It should be understood that, in order to clearly describe the technical solutions of the embodiments of the present invention, the terms "first" and "second" are used in the embodiments of the present invention to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and the terms "first" and "second" are not necessarily different.

[0036] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0037] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0038] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0039] In the logistics and distribution sector, cross-regional delivery issues (i.e., goods being delivered to areas other than the primary area of ​​responsibility) lead to circuitous delivery routes, low vehicle loading efficiency, and increased delivery costs. Efficiently identifying cross-regional delivery issues and optimizing delivery strategies is a significant challenge for logistics scheduling and management. Traditional delivery identification methods typically rely on manual experience or simple rule-based filtering, such as rough classification based solely on the recipient's address and its assigned region, lacking systematic analysis of multi-dimensional data such as waybill trajectory, vehicle load, time, and distance.

[0040] In existing technologies, the determination of cross-regional waybills often relies on a single condition (such as address attribution), failing to form a comprehensive identification system that combines statistics of the main delivery area, vehicle weight utilization rate, and route distance optimization. This makes it difficult to accurately locate cross-regional issues and generate actionable consolidation delivery suggestions, leading to wasted logistics resources and low delivery efficiency. Therefore, there is an urgent need for a method that can integrate multi-dimensional data, systematically identify cross-regional delivery, and provide optimization strategies to solve the problems of crude delivery identification and unreasonable resource allocation in existing technologies.

[0041] To resolve the above issues, please refer to [link / reference]. Figure 1 , Figure 1 This is a schematic flowchart illustrating a cross-regional delivery identification method according to an embodiment of this application. This cross-regional delivery identification method can be implemented using computer equipment, which can be deployed on a single server or a server cluster. It can also be deployed on handheld terminals, laptops, wearable devices, or robots, etc.

[0042] It should be noted that the acquisition of any information involved in the provided methods is in compliance with relevant regulations and is carried out with the user's consent, and will not infringe on the user's privacy or violate relevant laws and regulations.

[0043] Specifically, such as Figure 1 As shown, the provided cross-regional delivery identification method includes steps S101 to S104, which are detailed below:

[0044] Step S101. Filter the valid records with the delivery method of internal delivery from the waybill table, and remove the waybills of the whole vehicle type and the waybills of the internal parts type.

[0045] Specifically, the core of filtering valid records from the waybill table is to eliminate interfering data and focus on the target waybills to be analyzed. Specific filtering criteria include: Delivery method is "internal delivery": focusing only on waybills handled by the company's internal delivery team, excluding third-party delivery or self-pickup scenarios. Full truckload (FTL) waybills are excluded: FTL transportation typically uses dedicated vehicles, and its delivery logic differs from scattered order dispatches, requiring separate processing to avoid interfering with ordinary order dispatch analysis. Internal component waybills are excluded: Internal components (such as internally circulated documents and materials) do not require external delivery and do not involve cross-region issues, so they must be excluded to reduce data noise. The waybill table fields must include "Delivery method" (enumerated values: internal delivery / third-party delivery / self-pickup, etc.) and "Waybill type" (enumerated values: full truckload / less-than-truckload / internal component, etc.). Filtering can be performed using database SQL statements or data processing scripts (such as Python Pandas) to obtain the valid waybill dataset to be analyzed, containing key information such as waybill ID, trajectory number, recipient address, cargo weight, departure time, and recipient area code.

[0046] Step S102. Count the waybills with the same trajectory from the waybill table by receiving area, and select the receiving area with the most shipments as the main delivery area; if there are shipments with the same number, select the receiving area with the largest cargo weight as the main delivery area.

[0047] Specifically, for waybills with the same delivery trajectory (i.e., the same delivery vehicle or the same planned route), the number of waybills is counted according to the receiving area, and the main delivery area is determined by the rule of "number of waybills first + weight of goods as auxiliary", which serves as the benchmark for judging cross-regional delivery.

[0048] Statistics by trajectory grouping: Waybills under the same trajectory belong to the same route planning range and need to be analyzed together. The main area determination rules include: if a certain area has the most tickets, it is the main area; if multiple areas have the same number of tickets, the area with the largest total cargo weight is selected (cargo weight directly affects vehicle loading efficiency, and route planning for areas with heavy cargo is prioritized).

[0049] Group the waybills by "Track Number". Within each group, count the number of shipments and total weight by "Receiving Area". Sort the areas within each group in descending order of shipment count; if the number of shipments is the same, sort them in descending order of weight. Take the first area after sorting as the primary area. This can be achieved using SQL grouping aggregation functions (such as GROUP BY Track Number, Receiving Area, combined with COUNT(*) to count shipments and SUM(weight) to calculate total weight). Alternatively, sorting and filtering can be implemented using data frameworks (such as Pandas) with the groupby and sort_values ​​methods. Output: Each track corresponds to a primary delivery area, associated with all waybills under that track.

[0050] Step S103. Remove the waybills from the main delivery area in the waybill table, and determine whether the remaining waybills are cross-regional problem tickets; among them, within the preset departure time range and straight-line distance range corresponding to the cross-regional problem tickets, find the waybills whose vehicle weight ratio is less than or equal to the preset load capacity ratio, mark them as combined tickets and generate a combined delivery suggestion.

[0051] Specifically, after removing waybills from the main area, the remaining waybills belong to non-main areas, i.e., potential cross-regional tickets. The screening of cross-regional tickets involves searching for waybills with a low ratio of vehicle weight to preset load capacity (i.e., load utilization rate) within a preset time range (such as the same departure time) and straight-line distance range (such as a reasonable radius around the main area). If the ratio is ≤ preset value, such as 70%, it indicates that the vehicle still has loading space and can be combined for delivery to optimize the route. The generation of merging suggestions can combine combinable cross-regional tickets with main area routes, reducing duplicate dispatches or detours.

[0052] For each trajectory, records where "receiver area ≠ main area" are retained from the waybill and marked as candidate cross-regional tickets. Multi-dimensional filtering conditions include: Time range: a preset departure time fluctuation range (e.g., ±30 minutes) to ensure candidate tickets and main area waybills belong to the same delivery time period. Straight-line distance: calculating the straight-line distance between the candidate ticket and the center of the main area (or major distribution points within the main area) using the latitude and longitude of the recipient address, setting a threshold (e.g., ≤10 km) to exclude abnormal tickets that are too far away. Load utilization rate: calculating the ratio of the loaded weight to the vehicle's rated load weight (load weight / rated load weight ≤ preset ratio, e.g., 0.7) to determine if the vehicle has remaining space. Address latitude and longitude conversion can be achieved by converting the recipient address to coordinates using an address resolution interface (e.g., Baidu Maps API). Distance calculation can use the Haversine formula to calculate the spherical distance between two points. A merging suggestion is generated, suggesting that cross-regional tickets meeting the conditions be merged with main area waybills for loading, adjusting the delivery order to shorten the route. Mark problematic tickets with overlapping routes, associate them with combinable waybill groups, and generate "suggested merged delivery routes" (such as adding stopover points or adjusting route order).

[0053] Step S104. For each cross-regional problem ticket, obtain the increment value based on the sum of the actual distances between the cross-regional problem ticket and the adjacent main delivery area and the straight-line distance between the cross-regional problem ticket and the adjacent main delivery area, so as to generate the analysis result corresponding to each cross-regional problem ticket and complete the cross-regional delivery identification.

[0054] Specifically, by quantifying the increase in path length caused by cross-regional delivery (distance increment), the impact of the cross-regional delivery problem is assessed, providing a basis for optimization strategies. Actual distance includes the actual travel distance (or calculated based on the delivery network path) required for cross-regional tickets to be delivered separately from the main delivery area. Straight-line distance includes the shortest straight-line distance (the optimal distance under ideal conditions) between the cross-regional ticket recipient address and the adjacent main delivery area. Distance increment = Actual distance - Straight-line distance; a larger value indicates more severe detours and requires priority optimization.

[0055] Actual distance is calculated based on path data of the logistics and distribution network (such as road connectivity, restricted areas, etc.). A route planning algorithm (such as Dijkstra's algorithm) is used to calculate the round-trip distance (or detour distance) from the main area distribution point to the cross-regional ticket address. Straight-line distance is calculated by directly calculating the spherical distance between the cross-regional ticket address and the center of the main area. The increment value = actual travel distance - straight-line distance. Each cross-regional ticket is labeled with its increment value, associated trajectory, main area, and can be combined with ticket list information to generate a visual analysis report or data report for dispatchers to optimize delivery order or reassign areas. Actual road distances are obtained using GIS geographic information systems or route planning APIs. Data visualization tools (such as Tableau and Python Matplotlib) are used to display the increment value distribution and locate high-impact cross-regional issues.

[0056] In some embodiments, the step of filtering valid records with internal delivery as the delivery method from the waybill table and removing waybills of the full vehicle type and internal component type includes performing a filtering operation on the waybill data in the waybill table, excluding waybills marked as full vehicle transportation or internal components in the transportation type field, and only retaining waybills with internal delivery as the delivery method field; for waybills with the same receipt date, delivery person number and delivery vehicle trajectory number, selecting the unique record with the earliest delivery receipt time and status as receipt completed, if the delivery receipt time field is missing, it is replaced by the receipt time field value.

[0057] When filtering valid waybills, in addition to basic filtering conditions, deduplication rules and time field processing logic are added: Filtering conditions include excluding waybills with transportation type "full truckload" or "internal parts", and only retaining waybills with delivery method "internal delivery". Deduplication rules, for waybills with the same receipt date, delivery person number, and delivery vehicle trajectory number, retain only the unique record with the earliest delivery receipt time and status "received completed"; if "delivery receipt time" is missing, it is replaced by "receipt time".

[0058] Data filtering is performed using fields in the waybill table (transportation type, delivery method) to exclude irrelevant waybills. Deduplication is performed by grouping the filtered waybills by "receipt date, delivery person number, and delivery vehicle trajectory number". Within each group, records are sorted in ascending order by "delivery check-in time", prioritizing records with a status of "receipt completed"; if "delivery check-in time" is empty, then sorted by "receipt time". Window functions (such as SQL's ROW_NUMBER()) are used to assign a sequence number to each group of records, retaining the record with sequence number 1. This results in unique, valid waybill records after deduplication, ensuring the uniqueness and accuracy of subsequent data analysis.

[0059] In some embodiments, the step of counting waybills with the same trajectory from the waybill table by receiving area and identifying the receiving area with the most waybills as the main delivery area includes: grouping waybills according to the delivery vehicle trajectory number; counting waybills in each group by the receiving area name field to generate the number of waybills for each receiving area; identifying the receiving area with the highest number of waybills as the main delivery area; and if multiple receiving areas have the same number of waybills, comparing the sum of the cargo weight fields of each receiving area and selecting the receiving area with the largest total cargo weight as the main delivery area.

[0060] By clearly defining the statistical rules for the main delivery area, based on trajectory grouping and dual-dimensional sorting (ticket count first, cargo weight second): Trajectory grouping groups waybills by "delivery vehicle trajectory number," with each group corresponding to a set of waybills on the same delivery route. Area statistics generate the number of waybills for each area by counting waybills within each group according to "receiver area name." The determination of the main area includes: if an area has the highest number of tickets, it is designated as the main area; if the number of tickets is the same, the total cargo weight of each area is compared, and the area with the highest cargo weight is selected.

[0061] Group statistics utilize the data grouping and aggregation function to count the number of waybills and their weight by region within each trajectory group. Sorting and filtering are performed by first sorting the statistical results of each trajectory group in descending order by "Number of Waybills," and then by "Total Weight of Goods." The "Receiving Region Name" of the first record after sorting is taken as the primary delivery region. A unique primary delivery region corresponding to each trajectory is output as the basis for determining cross-regional issues.

[0062] In some embodiments, determining whether the remaining waybills are cross-regional issue tickets includes: for the remaining waybills after removing waybills from the main delivery area, extracting the delivery time field and the departure time field of the corresponding trajectory, and calculating the time difference; if the time difference is greater than 2 hours, and the sum of the actual driving distances from the latitude and longitude of the recipient address of the waybill to the latitude and longitude of the center of the adjacent main delivery area, minus the straight-line distance between the latitude and longitude of the center of the adjacent main delivery area, is greater than or equal to 10 kilometers, it is determined to be a cross-regional issue ticket; if waybills from the preceding and following main delivery areas are missing, the latitude and longitude registered at the delivery point are used as substitute coordinates for distance calculation.

[0063] By refining the criteria for identifying cross-regional issues with waybills, a dual threshold of time difference and distance difference is introduced, including: Time condition: A difference of more than 2 hours between the waybill delivery time and the vehicle departure time on the tracking line is considered a time mismatch. Distance condition: If the actual driving distance from the recipient address to the center of the adjacent main area, minus the straight-line distance between the centers of the main areas, is ≥10 kilometers, it is considered a severely circuitous route. Handling missing coordinates: If the coordinates of a waybill within a main area are missing, the latitude and longitude registered at the delivery point are used instead.

[0064] The time difference is calculated by extracting the "delivery time" of the waybill and the "departure time" of the trajectory, and calculating the time difference (unit: hours). If the difference is greater than 2 hours, the distance determination is initiated.

[0065] Distance calculation includes: Coordinate acquisition: Under normal circumstances: using the latitude and longitude of the main area center (such as the average coordinates of all receiving addresses within the area) and the latitude and longitude of the receiving address of the cross-regional ticket. Missing coordinates: using the fixed latitude and longitude registered at the delivery point (such as the coordinates resolved from the point address). Actual driving distance: obtaining the road distance between two points through a route planning API (such as Amap). Straight-line distance: calculating the spherical distance using the Haversine formula. Distance difference = actual driving distance - straight-line distance. If ≥ 10 km, it is judged as a cross-regional ticket. Judgment logic: if time difference > 2 hours and distance difference ≥ 10 km: marked as a cross-regional ticket.

[0066] In some embodiments, marking a ticket as a combinable ticket and generating a combined delivery suggestion includes: for each ticket with a cross-regional issue, searching for other waybills within a 2-hour period before and after the departure time of the corresponding trajectory, with a straight-line distance not exceeding 3 kilometers; obtaining the preset load capacity value corresponding to the delivery vehicle type, and calculating the ratio of the current vehicle weight to the preset load capacity value; if the ratio is less than or equal to 80%, marking the eligible waybill as a combinable ticket, and generating an optimal combined delivery route suggestion based on the delivery address of each waybill; wherein, the preset load capacity value corresponding to a van is 600 kg, the preset load capacity value corresponding to a medium-sized truck is 2 tons, the preset load capacity value corresponding to a 7.6-meter large truck is 8 tons, and the preset load capacity value corresponding to a 9.6-meter large truck is 10 tons.

[0067] By clearly defining the screening criteria and merging strategies for invoices that can be combined, based on vehicle type and load capacity, time window, distance range, and load utilization rate, the following criteria are defined: Search Scope: Waybills within 2 hours before and after the departure time with a straight-line distance ≤ 3 kilometers. Vehicle Type and Load Capacity: Preset load caps for different vehicle types (600kg for vans, 2 tons for medium-sized trucks, 8 / 10 tons for large trucks). Load Utilization Rate: When the current vehicle weight / preset load capacity ≤ 80%, it is determined that the invoices can be merged. Merging Suggestions: Generate the optimal delivery route based on the recipient address.

[0068] Conditional retrieval retrieves a preset load capacity value based on the waybill's "vehicle type" and calculates the ratio of the current load (total loaded cargo weight) to the load capacity value. Waybills with a departure time ±2 hours, a straight-line distance ≤3 kilometers, and a load capacity ratio ≤0.8 are filtered. Route optimization uses the TSP (Traveling Salesman Problem) algorithm or route planning API to generate the optimal delivery sequence based on all recipient address coordinates, reducing the total travel distance. Tagging and suggestions are provided by adding a "combinable ticket" tag to eligible waybills and generating route adjustment suggestions (such as adding stop order).

[0069] In some embodiments, generating the analysis results corresponding to each cross-regional issue ticket to complete the cross-regional delivery identification includes: using the waybill receipt time field as the statistical time caliber and the delivery point number as the organizational affiliation unit to count the cross-regional issue tickets, calculating the number of cross-regional tickets, the proportion of tickets that can be combined, the average increment value, and the month-on-month change rate; associating and archiving the waybill number, increment value, approval mark, and optimization suggestions for each cross-regional issue ticket, and generating weekly or monthly statistical reports for use by the dispatch management system.

[0070] By defining the statistical dimensions and output format of the analysis results, based on time caliber, organizational units, and quantitative indicators, the following are included: Statistical Dimensions: "Waybill receipt time" is used as the time caliber, and "delivery point number" is used as the organizational unit. Statistical Indicators: Number of cross-regional tickets, percentage of tickets that can be combined (combinable tickets / total number of cross-regional tickets), average distance increase value, and month-on-month change rate (compared to last week / month). Data Archiving and Reporting: Linking waybill numbers, distance increase values, and optimization suggestions, periodic reports (weekly / monthly) are generated.

[0071] Data aggregation is grouped by "delivery point number" and "receipt time (week / month)" to calculate various indicators. Report generation involves storing the results in a database table and then periodically generating PDF / Excel reports using ETL tools or the dispatch system, including trend charts, percentage analysis, and other visualizations. Data archiving involves recording "approval flag" (a field for manual confirmation by the dispatcher) and "optimization suggestions" for each cross-region ticket, forming historical traceability data.

[0072] In some embodiments, the method further includes: constructing a historical cross-regional behavior dataset and inputting it into a machine learning model for training. The input features of the machine learning model include waybill weight, delivery time interval, driving distance difference, and vehicle load utilization rate. A dynamic judgment threshold is obtained through machine learning model training to automatically adjust preset parameters, including the difference between delivery time and departure time, distance difference, and vehicle load ratio. The machine learning model is incrementally updated based on real-time logistics data to optimize cross-regional delivery identification rules and improve the accuracy of abnormal behavior identification.

[0073] By introducing a machine learning model to dynamically adjust preset thresholds, the accuracy of identification is improved. This includes: Dataset construction: collecting historical cross-regional behavior data and labeling whether it is a genuine cross-regional problem ticket. Input features: waybill weight, delivery time interval, travel distance difference, and vehicle load utilization rate. Model training: using classification algorithms (such as random forest, XGBoost) to train the dynamic judgment threshold. Incremental updates: continuously optimizing the model based on real-time data and automatically adjusting preset parameters such as time, distance, and load.

[0074] Feature engineering extracts four types of features: numerical features: waybill weight (kg), travel distance difference (actual distance - straight distance, km), and load utilization rate (vehicle weight / load value); and temporal features: the difference between delivery time and departure time (hours).

[0075] Model training involves dividing the model into training and testing sets and using supervised learning algorithms to train the model, with the goal of determining whether a ticket is a cross-regional ticket. Dynamic thresholds (e.g., adjusting the time difference threshold from a fixed 2-hour period to a dynamic value that varies with core utilization) are derived from the probability values ​​or feature importance output by the model.

[0076] Threshold updates include periodically (e.g., daily) fine-tuning the thresholds based on new data. For example, when the load utilization rate is >70%, the time difference threshold is relaxed to 3 hours; when the load utilization rate is ≤50%, the time difference threshold is tightened to 1.5 hours. Model deployment involves integrating the trained model into the recognition system to receive logistics data in real time and output dynamic parameters, replacing traditional fixed threshold rules.

[0077] By constructing a graph structure using geographic spatiotemporal data and capturing the spatiotemporal correlation features between areas through a graph convolutional network (GCN), pattern recognition and prediction for cross-region problems can be achieved.

[0078] Graph structure modeling: Each delivery area is regarded as a graph node. Node features include historical cross-region frequency, average cargo weight, geographical coordinates, etc.; edge features include historical shared order volume between areas, actual driving distance, road travel time, etc.

[0079] Spatiotemporal feature fusion: Based on GCN, a time dimension (such as time sliding window) is introduced to construct a spatiotemporal GCN model and learn the evolution pattern of cross-region behavior in time series.

[0080] Anomaly detection: Identify abnormal cross-region patterns by using the node embedding vectors output by the model (such as a connection weight between a certain region and a non-main region that is significantly higher than the normal level).

[0081] Data preprocessing: The coordinates of the area are converted into graph nodes. The real-time road distance between nodes is calculated using the Dijkstra algorithm as the edge weight. The total number of freight orders carried between areas is statistically analyzed on a weekly / monthly basis as the edge feature. A time series feature matrix is ​​constructed, with each time slice (e.g., 1 hour) containing dynamic data such as the distribution of freight weight and the time of departure for each node.

[0082] The model is constructed by using two GCN layers to extract spatial features, combined with gated recurrent units (GRUs) to process time series, and finally outputting serial region probability values ​​through a fully connected layer.

[0083] Model Training and Application: Using historical cross-region labels as supervision signals, the model is trained using the cross-entropy loss function to identify whether real-time waybills belong to abnormal cross-region patterns. The output includes a visual report containing a heatmap of region associations, assisting dispatchers in optimizing long-term region allocation strategies.

[0084] In some embodiments, a reinforcement learning (RL) agent is constructed to dynamically generate optimal merging and delivery decisions based on real-time waybill data, replacing traditional fixed threshold rules.

[0085] The state space definition includes the current vehicle weight, remaining vehicle capacity, set of planned route nodes, time window / distance range for cross-zone tickets, vehicle type capacity parameters, etc.

[0086] The action space design includes discrete actions such as "merging current zone tickets", "splitting to subsequent trains", and "adjusting the main area range". The reward function design uses the reduction in total distance increase after merging, the improvement in delivery time, and vehicle load utilization as positive rewards, and the increase in path detour as a negative reward.

[0087] Environmental modeling uses a logistics simulation platform to simulate the delivery process, inputting real-time waybill data and road network data (such as real-time traffic API) as the environmental state.

[0088] The agent training employs a deep Q-network (DQN) or a policy gradient algorithm (such as PPO), and gradually optimizes the merging strategy through offline historical data simulation training.

[0089] A simplified reward function example includes: `def calculate_reward(distance increase before merging, distance increase after merging, load utilization rate): Distance reduction = distance increase before merging - distance increase after merging; Load reward = 0.5 if 0.7 ≤ load utilization rate ≤ 0.9 else 0.2; return distance reduction * 0.1 + load reward`. The online decision-making application receives new ticket data in real time, and the agent outputs the optimal action (such as whether to merge and route adjustment schemes) based on the current state, which is then executed directly through the scheduling system via API.

[0090] In some embodiments, for historical unlabeled waybill data, self-supervised learning is used to construct feature representations of normal delivery patterns, and unlabeled cross-regional abnormal tickets are identified through reconstruction errors. Self-supervised task design: Pre-training tasks such as "address mask prediction" and "trajectory sequence shuffling recovery" are designed to force the model to learn the spatiotemporal correlation features of normal delivery. Feature reconstruction and anomaly scoring: Waybill features are extracted using an autoencoder (AE) or a contrastive learning model (such as SimCLR), and the reconstruction error of abnormal tickets is calculated. If the error exceeds a threshold, the ticket is determined to be cross-regional.

[0091] Pre-training data construction involves data augmentation of normal waybill data (such as randomly masking the latitude and longitude of the recipient address and shuffling the order of waybills within the trajectory) to construct self-supervised training samples.

[0092] The model architecture uses a multi-layer Transformer encoder as a feature extractor, and the input is a vector sequence containing features such as waybill time, cargo weight, address coordinates, and trajectory number.

[0093] Anomaly detection involves generating feature representations for new waybill data using a pre-trained model, calculating the cosine distance or reconstruction error with the normal pattern, and setting dynamic thresholds (such as mean + 3 times standard deviation) to identify cross-section anomalies.

[0094] In some embodiments, a heterogeneous graph of "waybill-region-vehicle" is constructed, and a graph neural network is used to predict the merging potential of cross-regional tickets and main region waybills, replacing the traditional single rule of distance / load weight. Heterogeneous graph modeling: Node types include waybills (including cargo weight and time), regions (including coordinates and historical co-loading frequency), and vehicles (including vehicle type and load capacity). Edge types include "waybill belongs to region", "vehicle loads waybill", and "region adjacency relationship". Graph attention mechanism: The interaction weights between different types of nodes are learned through GAT (Graph Attention Network) to capture the comprehensive impact of waybill merging on path efficiency (such as time window matching degree and load complementarity).

[0095] Heterogeneous graph construction includes: using knowledge graph tools (such as Neo4j) to store three types of nodes and their relationships, such as: waybill node attributes: waybill ID, cargo weight, delivery time, address coordinates; area node attributes: area ID, center coordinates, historical average departure time; vehicle node attributes: vehicle ID, vehicle type, maximum load capacity, and current load.

[0096] The model is trained using historical mergeable votes as positive samples and non-mergeable votes as negative samples to train a GNN classification model, which outputs the mergeable probability value.

[0097] Application scenarios include: for newly identified cross-region tickets, the probability of merging is predicted using a GNN model. If the probability is higher than 70%, a merging suggestion is automatically generated; if it is lower than 30%, it is marked as "requiring manual review".

[0098] In some embodiments, to address the problem of data sparsity in newly developed areas, historical cross-regional models from mature areas are used for transfer learning to quickly build localized identification models.

[0099] The source and target domain division includes: selecting mature regions with abundant data as the source domain (such as Beijing, Shanghai, and Guangzhou), and new regions as the target domain (such as third- and fourth-tier cities). Feature transfer strategy: retaining the low-level feature extraction layers of the source domain model (such as geocoding and temporal feature processing), and only fine-tuning the parameters of the top-level classification layer. Domain adaptation techniques: using adversarial domain adaptation (such as DANN) to reduce the distribution differences between the source and target domains and improve the model's generalization ability.

[0100] Source domain model pre-training learns common cross-region features (such as time difference, distance difference, and load ratio) by training a base model (such as random forest or neural network) on mature region data.

[0101] Target domain fine-tuning includes: collecting a small amount of labeled data from the new region, freezing the first 90% of the network layers in the base model, and updating the parameters only for the last fully connected layer. A domain adversarial loss function is introduced to prevent the model from distinguishing between input data from the source and target domains. Localization optimization combines specific rules for the new region (such as rural road restrictions and regional delivery time windows) with a custom rule layer added to the transfer model, achieving a hybrid "algorithm + rule" recognition.

[0102] In some embodiments, the root cause of cross-region problems is analyzed using causal inference models (such as structural causal models, SCM), replacing traditional correlation analysis, and causal relationships such as "mismatched vehicle models" and "unreasonable division of regions" are identified.

[0103] Cause-and-effect graph construction: Define potential causal variables (such as vehicle type, distance between areas, and departure time), construct a directed acyclic graph (DAG), and label the causal relationships between variables (such as "vehicle type too small → insufficient load → forced cross-area crossing"). Intervention analysis: Simulate the impact of intervention measures (such as changing vehicle type or adjusting area boundaries) on the cross-area crossing rate through "do-calculus" to quantify the causal effect.

[0104] The definition of variables and data preparation include: defining independent variables: vehicle type (X1), distance from the center of the area (X2), and fluctuation of departure time (X3); dependent variable: cross-area occurrence (Y); and control variables: cargo weight distribution (Z1) and road congestion index (Z2).

[0105] Cause-effect graph learning uses PC or GES algorithms to learn the causal structure from historical data and combines it with logistics expert knowledge to correct the cause-effect graph.

[0106] Causal effect calculation uses propensity score matching (PSM) to calculate the average treatment effect (ATE) of vehicle type on cross-regional areas. For example: ATE = E[Y|do(vehicle type = medium truck)] - E[Y|do(vehicle type = van)]. Strategy optimization is carried out by formulating optimization schemes for variables with high causal effect. For example, if it is found that "the distance between the center of the area is >15 kilometers" is the main reason for cross-regional areas, the area re-division process is triggered.

[0107] This application provides a method, apparatus, computer equipment, and storage medium for identifying cross-regional delivery. The method performs dual statistical analysis of the number of parcels and cargo weight for waybills with the same trajectory, ensuring the scientific validity of the primary delivery area determination and avoiding the bias caused by a single indicator (such as only the number of parcels). This lays an accurate foundation of data for subsequent cross-regional delivery identification. By combining preset departure time ranges, straight-line distance ranges, and vehicle weight-to-load ratios, it can not only identify cross-regional parcels with abnormal address attributions but also simultaneously filter out waybills that can be merged for delivery, achieving the integration of "problem identification" and "optimization suggestions," thus improving the efficiency of delivery resource integration. By calculating the distance increment value and generating analysis results, it provides data support for scheduling management, facilitating targeted optimization of delivery routes. Simultaneously, the merging suggestions for parcels can effectively reduce vehicle empty-running rates and lower delivery costs. Through statistical analysis of cross-regional problem parcels (such as the number of cross-regional parcels and the proportion of parcels that can be merged), a systematic management report is generated, helping logistics companies achieve digital and refined control of the delivery process.

[0108] Please see Figure 2 As shown, Figure 2 This is a schematic diagram of the cross-regional delivery identification device 200 provided in this application embodiment. The cross-regional delivery identification device 200 is used to execute the steps of the cross-regional delivery identification method shown in the above embodiments. The cross-regional delivery identification device 200 can be a single server or a server cluster, or it can be a terminal, such as a handheld terminal, a laptop computer, a wearable device, or a robot.

[0109] like Figure 2 As shown, the cross-regional delivery identification device 200 includes:

[0110] The waybill rejection unit 201 is used to filter valid records with internal delivery method from the waybill table and reject waybills of the whole vehicle type and waybills of the internal component type.

[0111] The waybill statistics unit 202 is used to count waybills with the same trajectory from the waybill table by receiving area, and the receiving area with the most counted tickets is the main delivery area; if there are the same number of tickets, the receiving area with the largest cargo weight is selected as the main delivery area.

[0112] Problem judgment unit 203 is used to remove waybills from the main delivery area in the waybill table and determine whether the remaining waybills are cross-regional problem tickets; wherein, within the preset departure time range and straight-line distance range corresponding to the cross-regional problem tickets, the waybills whose vehicle weight ratio to preset load capacity ratio is less than or equal to the preset ratio are found, marked as combinable tickets and a combined delivery suggestion is generated.

[0113] The identification completion unit 204 is used to obtain an increment value for each cross-regional problem ticket based on the sum of the actual distances between the cross-regional problem ticket and the adjacent main delivery area and the straight-line distance between the cross-regional problem ticket and the adjacent main delivery area, so as to generate the analysis result corresponding to each cross-regional problem ticket and complete the cross-regional delivery identification.

[0114] In some embodiments, the step of filtering valid records with internal delivery as the delivery method from the waybill table and removing waybills of the full vehicle type and internal component type includes: performing a filtering operation on the waybill data in the waybill table to exclude waybills marked as full vehicle transportation or internal components in the transportation type field, and only retaining waybills with internal delivery as the delivery method field; for waybills with the same receipt date, delivery person number, and delivery vehicle trajectory number, selecting the unique record with the earliest delivery receipt time and status as receipt completed, and if the delivery receipt time field is missing, replacing it with the receipt time field value.

[0115] In some embodiments, the step of counting waybills with the same trajectory from the waybill table by receiving area and identifying the receiving area with the most waybills as the main delivery area includes: grouping waybills according to the delivery vehicle trajectory number; counting waybills in each group by the receiving area name field to generate the number of waybills for each receiving area; identifying the receiving area with the highest number of waybills as the main delivery area; and if multiple receiving areas have the same number of waybills, comparing the sum of the cargo weight fields of each receiving area and selecting the receiving area with the largest total cargo weight as the main delivery area.

[0116] In some embodiments, determining whether the remaining waybills are cross-regional issue tickets includes: for the remaining waybills after removing waybills from the main delivery area, extracting the delivery time field and the departure time field of the corresponding trajectory, and calculating the time difference; if the time difference is greater than 2 hours, and the sum of the actual driving distances from the latitude and longitude of the recipient address of the waybill to the latitude and longitude of the center of the adjacent main delivery area, minus the straight-line distance between the latitude and longitude of the center of the adjacent main delivery area, is greater than or equal to 10 kilometers, it is determined to be a cross-regional issue ticket; if waybills from the preceding and following main delivery areas are missing, the latitude and longitude registered at the delivery point are used as substitute coordinates for distance calculation.

[0117] In some embodiments, marking a ticket as a combinable ticket and generating a combined delivery suggestion includes: for each ticket with a cross-regional issue, searching for other waybills within a 2-hour period before and after the departure time of the corresponding trajectory, with a straight-line distance not exceeding 3 kilometers; obtaining the preset load capacity value corresponding to the delivery vehicle type, and calculating the ratio of the current vehicle weight to the preset load capacity value; if the ratio is less than or equal to 80%, marking the eligible waybill as a combinable ticket, and generating an optimal combined delivery route suggestion based on the delivery address of each waybill; wherein, the preset load capacity value corresponding to a van is 600 kg, the preset load capacity value corresponding to a medium-sized truck is 2 tons, the preset load capacity value corresponding to a 7.6-meter large truck is 8 tons, and the preset load capacity value corresponding to a 9.6-meter large truck is 10 tons.

[0118] In some embodiments, generating the analysis results corresponding to each cross-regional issue ticket to complete the cross-regional delivery identification includes: using the waybill receipt time field as the statistical time caliber and the delivery point number as the organizational affiliation unit to count the cross-regional issue tickets, calculating the number of cross-regional tickets, the proportion of tickets that can be combined, the average increment value, and the month-on-month change rate; associating and archiving the waybill number, increment value, approval mark, and optimization suggestions for each cross-regional issue ticket, and generating weekly or monthly statistical reports for use by the dispatch management system.

[0119] In some embodiments, the method further includes: constructing a historical cross-regional behavior dataset and inputting it into a machine learning model for training. The input features of the machine learning model include waybill weight, delivery time interval, driving distance difference, and vehicle load utilization rate. A dynamic judgment threshold is obtained through machine learning model training to automatically adjust preset parameters, including the difference between delivery time and departure time, distance difference, and vehicle load ratio. The machine learning model is incrementally updated based on real-time logistics data to optimize cross-regional delivery identification rules and improve the accuracy of abnormal behavior identification.

[0120] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the cross-regional delivery identification device and its modules described above can be referred to the corresponding process in the cross-regional delivery identification method embodiments described above, and will not be repeated here.

[0121] The aforementioned cross-regional delivery identification method can be implemented as a computer program, which can be used in, for example... Figure 2 It runs on the device shown.

[0122] Please see Figure 3 , Figure 3 This is a schematic block diagram of the structure of a computer device provided in an embodiment of this application. The computer device includes a processor, a memory, and a network interface connected via a device bus, wherein the memory may include a storage medium and internal memory.

[0123] The storage medium may store operating devices and computer programs. The computer program includes program instructions that, when executed, cause the processor to perform any serial area delivery identification method.

[0124] The processor provides computing and control capabilities, supporting the operation of the entire computer device.

[0125] Internal memory provides an environment for the execution of computer programs in non-volatile storage media. When the computer program is executed by the processor, it enables the processor to perform any serial area delivery identification method.

[0126] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the terminal to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0127] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0128] In one embodiment, the processor is configured to run a computer program stored in memory to perform the following steps:

[0129] Filter the valid records in the waybill table that are delivered internally, and remove waybills for full vehicles and waybills for internal components.

[0130] The waybill table is used to count waybills with the same route by receiving area. The receiving area with the most shipments is the main delivery area. If there are shipments with the same number of shipments, the receiving area with the largest cargo weight is selected as the main delivery area.

[0131] Remove waybills from the main delivery area from the waybill table, and determine whether the remaining waybills are cross-regional problem tickets; among them, within the preset departure time range and straight-line distance range corresponding to the cross-regional problem tickets, find waybills whose vehicle weight ratio is less than or equal to the preset load capacity ratio, mark them as combinable tickets and generate a combined delivery suggestion.

[0132] For each out-of-area problem ticket, the distance increment value is obtained by combining the sum of the actual distances between the out-of-area problem ticket and the adjacent main delivery area with the straight-line distance between the out-of-area problem ticket and the adjacent main delivery area, so as to generate the analysis result corresponding to each out-of-area problem ticket and complete the out-of-area delivery identification.

[0133] In some embodiments, the step of filtering valid records with internal delivery as the delivery method from the waybill table and removing waybills of the full vehicle type and internal component type includes: performing a filtering operation on the waybill data in the waybill table to exclude waybills marked as full vehicle transportation or internal components in the transportation type field, and only retaining waybills with internal delivery as the delivery method field; for waybills with the same receipt date, delivery person number, and delivery vehicle trajectory number, selecting the unique record with the earliest delivery receipt time and status as receipt completed, and if the delivery receipt time field is missing, replacing it with the receipt time field value.

[0134] In some embodiments, the step of counting waybills with the same trajectory from the waybill table by receiving area and identifying the receiving area with the most waybills as the main delivery area includes: grouping waybills according to the delivery vehicle trajectory number; counting waybills in each group by the receiving area name field to generate the number of waybills for each receiving area; identifying the receiving area with the highest number of waybills as the main delivery area; and if multiple receiving areas have the same number of waybills, comparing the sum of the cargo weight fields of each receiving area and selecting the receiving area with the largest total cargo weight as the main delivery area.

[0135] In some embodiments, determining whether the remaining waybills are cross-regional issue tickets includes: for the remaining waybills after removing waybills from the main delivery area, extracting the delivery time field and the departure time field of the corresponding trajectory, and calculating the time difference; if the time difference is greater than 2 hours, and the sum of the actual driving distances from the latitude and longitude of the recipient address of the waybill to the latitude and longitude of the center of the adjacent main delivery area, minus the straight-line distance between the latitude and longitude of the center of the adjacent main delivery area, is greater than or equal to 10 kilometers, it is determined to be a cross-regional issue ticket; if waybills from the preceding and following main delivery areas are missing, the latitude and longitude registered at the delivery point are used as substitute coordinates for distance calculation.

[0136] In some embodiments, marking a ticket as a combinable ticket and generating a combined delivery suggestion includes: for each ticket with a cross-regional issue, searching for other waybills within a 2-hour period before and after the departure time of the corresponding trajectory, with a straight-line distance not exceeding 3 kilometers; obtaining the preset load capacity value corresponding to the delivery vehicle type, and calculating the ratio of the current vehicle weight to the preset load capacity value; if the ratio is less than or equal to 80%, marking the eligible waybill as a combinable ticket, and generating an optimal combined delivery route suggestion based on the delivery address of each waybill; wherein, the preset load capacity value corresponding to a van is 600 kg, the preset load capacity value corresponding to a medium-sized truck is 2 tons, the preset load capacity value corresponding to a 7.6-meter large truck is 8 tons, and the preset load capacity value corresponding to a 9.6-meter large truck is 10 tons.

[0137] In some embodiments, generating the analysis results corresponding to each cross-regional issue ticket to complete the cross-regional delivery identification includes: using the waybill receipt time field as the statistical time caliber and the delivery point number as the organizational affiliation unit to count the cross-regional issue tickets, calculating the number of cross-regional tickets, the proportion of tickets that can be combined, the average increment value, and the month-on-month change rate; associating and archiving the waybill number, increment value, approval mark, and optimization suggestions for each cross-regional issue ticket, and generating weekly or monthly statistical reports for use by the dispatch management system.

[0138] In some embodiments, the method further includes: constructing a historical cross-regional behavior dataset and inputting it into a machine learning model for training. The input features of the machine learning model include waybill weight, delivery time interval, driving distance difference, and vehicle load utilization rate. A dynamic judgment threshold is obtained through machine learning model training to automatically adjust preset parameters, including the difference between delivery time and departure time, distance difference, and vehicle load ratio. The machine learning model is incrementally updated based on real-time logistics data to optimize cross-regional delivery identification rules and improve the accuracy of abnormal behavior identification.

[0139] The embodiments of this application also provide a computer-readable storage medium storing a computer program, the computer program including program instructions, and the processor executing the program instructions to implement the steps of the cross-regional delivery identification method provided in the above embodiments of this application.

[0140] The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiments, such as the hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, SmartMedia Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the computer device.

[0141] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A string zone shipment identification method, characterized by, The method comprises the following steps: Screening valid records with internal delivery mode from the delivery table, and excluding whole vehicle type delivery and internal part type delivery; Counting the delivery tickets of the same track in the delivery table according to the receiving area, and taking the receiving area with the most tickets as the main delivery area; if there are the same number of tickets, the receiving area with the heaviest goods is selected as the main delivery area; Excluding the delivery tickets of the main delivery area from the delivery table, and judging whether the remaining delivery tickets are serial area problem tickets; within the preset departure time range and straight line distance range corresponding to the serial area problem tickets, searching for the delivery tickets with a vehicle load ratio less than or equal to a preset ratio, marking them as combinable tickets, and generating a merging delivery suggestion; For each serial area problem ticket, the sum of the actual distance of the serial area problem ticket and the adjacent main delivery area and the straight line distance of the adjacent main delivery area corresponding to the serial area problem ticket are obtained to obtain a distance increase value, so as to generate an analysis result corresponding to each serial area problem ticket, and complete the serial area delivery identification; the method further comprises: constructing a historical serial area behavior data set, inputting it into a machine learning model for training, and the input features corresponding to the machine learning model include delivery ticket weight, delivery time interval, driving distance difference and vehicle load utilization rate; a dynamic judgment threshold is obtained through machine learning model training, which is used to automatically adjust the preset parameters, including the delivery time and departure time difference, distance difference and vehicle load ratio; based on real-time logistics data, the machine learning model is incrementally updated, the serial area delivery identification rule is optimized, and the abnormal behavior recognition accuracy is improved; a graph structure is constructed by using geographical spatio-temporal data, the spatio-temporal correlation features between the areas are captured through a graph convolution network, the pattern recognition and prediction of the serial area problem are realized, the graph structure modeling regards each delivery area as a graph node, the node features include historical serial area frequency, average goods weight and geographical coordinates; the edge features include the historical co-loading ticket quantity between the areas, the actual driving distance and the road passing time, the spatio-temporal feature fusion constructs a spatio-temporal GCN model by introducing a time sliding window in the time dimension, learns the serial area behavior evolution law of the area in the time sequence, and detects the abnormal serial area mode through the node embedding vector output by the model; the data preprocessing converts the area coordinates into graph nodes, calculates the real-time road distance between the nodes as the edge weight through the Dijkstra algorithm, and calculates the co-loading ticket quantity between the areas as the edge feature, constructs a time sequence feature matrix, each time slice contains the goods weight distribution of the node and the dynamic data of the departure time period, the model construction extracts the spatial features through two GCN layers, combines the gate recurrent unit to process the time sequence, and finally outputs the serial area probability value through the full connection layer, the model training and application use the historical serial area label as the supervision signal, train the model using the cross-entropy loss function, identify whether the real-time delivery ticket belongs to the abnormal serial area mode, and output a visual report containing the area correlation heat map to assist the dispatcher in optimizing the long-term area division strategy.

2. The method of claim 1, wherein, The screening of valid records with internal delivery mode from the delivery table, and the exclusion of whole vehicle type delivery and internal part type delivery, comprises: Filtering operation is performed on the waybill data in the waybill table, and the waybills marked as whole vehicle transportation or internal parts in the transportation type field are excluded, and only the waybills with the delivery mode field as internal delivery are retained; For the waybills with the same signing date, delivery person number and delivery vehicle track number, the unique record with the earliest delivery and signing time and the status of completed signing is selected, and if the delivery and signing time field is missing, the value of the signing time field is replaced.

3. The method of claim 1, wherein, The waybills of the same track in the waybill table are counted according to the receiving area, and the receiving area with the most tickets is the main delivery area, including: According to the delivery vehicle track number, the waybills are grouped, and the number of waybills in each group is counted according to the receiving area name field to generate the number of waybills in each receiving area; The receiving area with the highest number of waybills is determined as the main delivery area; if there are multiple receiving areas with the same number of waybills, the total weight of each receiving area is compared, and the receiving area with the largest total weight is selected as the main delivery area.

4. The method of claim 1, wherein, The remaining waybills are judged whether they are string area problem tickets, including: For the remaining waybills after excluding the main delivery area waybills, the delivery time field and the corresponding track departure time field are extracted, and the time difference is calculated; If the time difference is greater than 2 hours, and the sum of the actual driving distance from the receiving address latitude and longitude of the waybill to the center latitude and longitude of the adjacent main delivery area is greater than or equal to 10 kilometers minus the straight line distance between the center latitude and longitude of the adjacent main delivery area, it is determined as a string area problem ticket; If the main delivery area waybills before and after are missing, the latitude and longitude registered by the delivery point department is used as the replacement coordinates for distance calculation.

5. The method of claim 1, wherein, The string area problem tickets are marked as combinable tickets and the merging delivery suggestion is generated, including: For each string area problem ticket, within 2 hours before and after the departure time of the corresponding track, other waybills with a straight line distance of not more than 3 kilometers are searched; The preset load value corresponding to the delivery vehicle type is obtained, and the ratio of the current vehicle load weight to the preset load value is calculated; if the ratio is less than or equal to 80%, the waybills meeting the conditions are marked as combinable tickets, and the optimal merging delivery path suggestion is generated based on the receiving addresses of each waybill; Wherein, the preset load value corresponding to the minivan is 600 kg, the preset load value corresponding to the medium truck is 2 tons, the preset load value corresponding to the 7.6-meter large truck is 8 tons, and the preset load value corresponding to the 9.6-meter large truck is 10 tons.

6. The method of claim 1, wherein, The analysis result corresponding to each string area problem ticket is generated to complete the string area delivery identification, including: Taking the waybill signing time field as the statistical time range and the delivery point department number as the organizational attribution unit, the string area problem tickets are counted to calculate the number of string area tickets, the proportion of combinable tickets, the average distance increase value and the year-on-year change rate; The waybill number, distance increase value, whether to approve the mark and optimization suggestion of each string area problem ticket are associated and archived to generate weekly or monthly statistical reports for the dispatch management system to call.

7. A string zone shipment identification device, characterized by, Including: The waybill exclusion unit is used to filter the valid records with the delivery mode as internal delivery from the waybill table, and to exclude the whole vehicle type waybills and internal part type waybills; The waybill statistics unit is configured to count the waybills of the same track by the receiving area in the waybill table, and determine the receiving area with the most waybills as the main delivery area; if the number of waybills is the same, the receiving area with the heaviest goods is selected as the main delivery area; The problem judgment unit is configured to exclude the waybills of the main delivery area from the waybill table, and determine whether the remaining waybills are serial area problem tickets; if the serial area problem ticket corresponds to a preset departure time range and a straight line distance range, the waybills with a vehicle load and a preset load ratio less than or equal to a preset ratio are searched, marked as combinable tickets, and a combined delivery suggestion is generated; The recognition completion unit is configured to obtain an increased distance value for each serial area problem ticket according to the sum of the actual distances of the serial area problem ticket and the adjacent main delivery area and the straight line distance of the serial area problem ticket and the adjacent main delivery area, to generate an analysis result corresponding to each serial area problem ticket, and to complete the serial area delivery identification; further comprising: constructing a historical serial area behavior data set, inputting into a machine learning model for training, the input features of the machine learning model including waybill weight, delivery time interval, driving distance difference, and vehicle load utilization rate; obtaining a dynamic determination threshold value through machine learning model training, which is used to automatically adjust preset parameters, the preset parameters including delivery time and departure time difference, distance difference, and vehicle load ratio; based on real-time logistics data, the machine learning model is incrementally updated to optimize the serial area delivery identification rule and improve the accuracy of abnormal behavior identification; a graph structure is constructed using geographic spatio-temporal data, and the spatio-temporal correlation features between the areas are captured through a graph convolution network to realize pattern recognition and prediction of the serial area problem; the graph structure modeling regards each delivery area as a graph node, and the node features include historical serial area frequency, average freight weight, and geographic coordinates; the edge features include the number of historical co-loaded waybills between the areas, the actual driving distance, and the road travel time; the spatio-temporal feature fusion constructs a spatio-temporal GCN model by introducing a time sliding window in the time dimension to learn the evolution rule of the serial area behavior of the areas in the time sequence; the abnormality detection identifies abnormal serial area patterns through the node embedding vector output by the model; the data preprocessing converts the area coordinates into graph nodes, calculates the real-time road distance between the nodes as the edge weight through the Dijkstra algorithm, and constructs a time series feature matrix by counting the number of co-loaded waybills between the areas as the edge features; each time slice contains the freight weight distribution of the nodes and the dynamic data of the departure time period; the model construction extracts spatial features using two GCN layers, processes the time sequence using a gated recurrent unit, and finally outputs the serial area probability value through a fully connected layer; the model training and application use the historical serial area label as a supervision signal, train the model using a cross-entropy loss function, identify whether the real-time waybill belongs to an abnormal serial area pattern, and output a visual report containing an area correlation heat map to assist dispatchers in optimizing long-term area division strategies.

8. A computer device, comprising: The computer device includes a memory and a processor; The memory is configured to store a computer program; The processor is configured to execute the computer program and implement the method of any one of claims 1 to 6 when executing the computer program.

9. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program, when executed by a processor, causes the processor to implement the method in any one of claims 1 to 6.

Images