Transportation path optimization-based goods substitute customer portrait generation method

Abstract

The invention discloses a freight agent customer portrait generation method based on transportation path optimization, and relates to the technical field of freight agent customer portrait generation. Extracting a path offset direction consistency rate, a path reproduction frequency and a path staying inertia span based on the path adoption offset sequence, and combining to generate a path offset inertia parameter set for determining the stability degree of the customer path behavior; and mapping the path offset inertial parameter set to a freezing credibility interval, and judging whether a label freezing signal is generated or not for identifying that the customer path behavior enters a stable preference state and correspondingly triggering an update stop operation of the portrait label. According to the method, the problem that the portrait label is mistakenly updated due to the fact that the path behavior stability cannot be identified when the customer stably adopts the non-recommended transportation path for a long time in the existing goods substitute customer portrait is solved, and accurate freezing and dynamic release control of the customer path preference label is realized.

Description

Technical Field

[0001] This invention relates to the field of freight forwarding customer profile generation technology, specifically to a method for generating freight forwarding customer profiles based on transportation route optimization. Background Technology

[0002] Customer profiling based on transportation route optimization refers to the process of collecting and analyzing a large amount of transportation route data in freight forwarding (freight forwarding) business to uncover customer behavioral characteristics during the logistics transportation process, thereby constructing a customer profile that reflects multi-dimensional characteristics such as customer transportation preferences, business habits, and service needs. Existing technologies typically begin by collecting customer-related transportation route data through methods such as trajectory tracking, order records, and scheduling logs. Route optimization algorithms (such as shortest path algorithms, time window optimization, and cost function modeling) are then used to analyze and optimize transportation routes, thereby extracting customer behavioral characteristics in dimensions such as transportation time sensitivity, destination distribution, route selection preferences, and frequency patterns. This process usually includes four key steps: first, multi-source collection and normalization of transportation route data; second, route optimization analysis to construct optimal transportation models for individual customers in multiple scenarios; third, structured extraction of customer behavioral characteristics, including route preferences, service levels, and cost tendencies; and fourth, profile generation and classification modeling, using a tag system or vector space to achieve visual representation of customer groups and differentiated service support. Through the above process, freight forwarding companies can achieve accurate customer identification and personalized service optimization based on actual transportation behavior, thereby improving transportation efficiency and customer satisfaction.

[0003] The existing technology has the following shortcomings: In freight forwarding, when some high-frequency clients consistently use actual transportation routes that deviate from the optimized routes due to objective reasons such as carrier constraints, corporate compliance requirements, or confidentiality needs, this leads to a long-term deviation between their actual transportation behavior and the system's recommended routes. The high consistency of these clients' actual routes over a considerable period reflects a clear and stable business decision-making logic in their route selection, rather than an occasional deviation. However, existing systems, in generating client profiles based on route optimization, typically update route adoption tags in the client profile dynamically based on the consistency between recommended and actual routes within a fixed time window to improve profile update efficiency. In the aforementioned situation, because the system fails to distinguish whether the client's route deviation is due to occasional behavior or a long-term stable strategy, it continues to update the profile using the recommended routes as a reference, leading to erroneous tag changes. Existing freight forwarding customer profiling technologies based on transportation route optimization cannot determine whether profile tags should stop being dynamically updated based on the stability of customer route behavior when they consistently use non-recommended routes over a long period. This leads to drift in the tags reflecting route preferences in the customer profile, reducing the accuracy of the customer profile in depicting actual transportation behavior and further affecting the effectiveness of transportation scheduling decisions and service strategy matching based on the profile.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide a method for generating freight forwarding customer profiles based on transportation route optimization, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for generating freight forwarding customer profiles based on transportation route optimization, specifically including the following steps: S1. Map the recommended route and the actual route for each transportation task of the freight forwarding customer to a complete route. Generate a route offset label based on the overlap of route nodes, the trend of route offset direction, and the rejection of route selection. Form a route adoption offset sequence according to the order of task completion time to determine whether the customer has been using non-recommended routes for a long time. S2. When customers consistently use non-recommended paths over a long period, extract the path offset direction consistency rate, path recurrence frequency, and path dwell inertia span based on the path adoption offset sequence, and combine them to generate a path offset inertia parameter set to determine the stability of the customer's path behavior. S3. Map the path offset inertia parameter set to the frozen confidence interval, determine whether to generate a label freeze signal, which is used to indicate that the customer's path behavior has entered a stable preference state and trigger the corresponding profile label update stop operation. S4. After generating the tag freeze signal, establish behavior monitoring annotations, expand the record of the path adoption offset sequence generated by subsequent transportation tasks, continuously track the path behavior trend and calculate the degree of offset between the path and historical parameters. S5. Based on the updated path offset inertia parameter set in the behavior monitoring annotation, determine whether the freeze release condition is met. If the freeze release condition is met, release the label freeze state and resume the profile label update process to achieve dynamic control of customer path profile labels.

[0007] Preferably, S1 specifically includes the following steps: S101. Compare the path node sequence corresponding to the recommended path in each transportation task of the freight forwarding customer with the path node sequence corresponding to the actual path node, establish a mapping relationship between the recommended path node and the actual path node based on the spatial location correspondence of the path node, and form a complete path mapping covering the entire recommended path and the actual path based on the mapping relationship. S102. Calculate the path node overlap between the recommended path and the actual path based on the complete path mapping, extract the overall path offset direction trend of the actual path relative to the recommended path, and determine the road segment selection rejection degree based on the actual path's avoidance of specific road segments in multiple transportation tasks. Combine the path node overlap degree, path offset direction trend and road segment selection rejection degree to generate the path offset label corresponding to each transportation task. S103. Arrange the path offset labels according to the order of completion time of the transportation task to form a path adoption offset sequence. Divide the path adoption offset sequence into segments based on whether adjacent path offset labels are consecutive non-recommended path types. When the number of non-recommended path offset labels in a consecutive segment exceeds a preset threshold and the trend of path offset direction does not reverse within the consecutive segment, it is determined that the customer has been using non-recommended paths for a long time.

[0008] Preferably, S102 is as follows: Based on the complete path mapping, the path node sequence corresponding to the recommended path and the path node sequence corresponding to the actual path are extracted. By statistically analyzing the ratio of the number of spatially matched nodes in the two sets of path node sequences to the total number of path nodes, the path node overlap between the recommended path and the actual path is calculated. Based on the spatial arrangement order of each path node in the complete path mapping, the overall offset direction change trajectory of the actual path relative to the recommended path is calculated, and by performing consistency analysis on the offset direction of continuous path nodes, the overall path offset direction trend reflecting the actual path relative to the recommended path is extracted. Based on the complete path mapping results of multiple transportation tasks, the set of road segments that were repeatedly detoured or not used in the actual path is counted. The frequency of occurrence of each road segment being avoided in multiple transportation tasks is calculated to determine the road segment selection rejection degree. The path node overlap degree, path offset direction trend and road segment selection rejection degree are combined and mapped to generate the path offset label corresponding to each transportation task.

[0009] Preferably, S2 specifically includes the following steps: S201. When customers continue to use non-recommended paths for a long time, based on the non-recommended path offset labels that appear consecutively in the path adoption offset sequence, count the path offset direction corresponding to each offset label, and calculate the proportion of the number of times the same offset direction appears in consecutive offset labels to the total number of consecutive offset labels, so as to extract the path offset direction consistency rate. Based on the number of times the path offset label corresponding to the same actual path appears in the path adoption offset sequence, count the number of path recurrences. S202. Based on the continuous distribution of path offset labels corresponding to the same actual path in the path adoption offset sequence in the time dimension, determine the continuous duration of the customer in the state of keeping the same actual path unchanged, extract the path dwell inertia span with the continuous duration, and combine the path offset direction consistency rate, path recurrence times and path dwell inertia span after uniform dimension processing to generate a path offset inertia parameter set. S203. Based on the value distribution of each parameter in the path offset inertia parameter set in the path adoption offset sequence, construct a parameter determination interval to distinguish different stability levels. When the entire path offset inertia parameter set falls into the corresponding stability interval, determine the stability level of the customer's path behavior.

[0010] Preferably, S203 is as follows: Based on the path adoption offset sequence, the historical value distribution ranges of path offset direction consistency rate, path recurrence frequency and path dwell inertia span in multiple customer transportation tasks are extracted respectively. Based on the variation range and central tendency of each parameter, the single parameter value threshold range used to represent stable behavioral characteristics is determined. After setting stable judgment segments for path offset direction consistency rate, path recurrence count and path dwell inertial span, the current value of each parameter in the path offset inertial parameter set is compared with the corresponding stable judgment segment to identify whether each parameter falls into the corresponding stable judgment segment, thereby constructing the stable hit status of the current path offset inertial parameter set. Based on the stability hit statistics of the path offset inertial parameter set, the number of parameters falling into the stability judgment segment is counted. When the current value of all parameters falls into their corresponding stability judgment segment at the same time, it is determined that the stability of the customer's path behavior has reached a stable state, which is used to support the status control operation of the profile tag.

[0011] Preferably, S3 is as follows: After normalizing the path offset direction consistency rate, path recurrence count, and path dwell inertia span in the path offset inertia parameter set, a joint evaluation vector is constructed. Based on the stability distribution of the joint evaluation vector in historical task data, the path offset inertia parameter set is mapped to a frozen confidence interval to reflect the confidence level achieved by the offset inertia characteristics of customer path behavior. Set the freezing critical range in the freezing confidence interval, and compare the mapping result of the path offset inertial parameter set with the freezing critical range. When the entire path offset inertial parameter set falls into the freezing critical range, generate a label freezing signal to indicate that the customer's path behavior has entered a stable preference state. After generating the tag freeze signal, the tag freeze signal is associated with the path adoption class tag in the customer profile, and the signal is used as the control trigger to execute the update stop operation of the path adoption class tag, so that the path adoption class tag no longer changes with the subsequent path adoption offset sequence.

[0012] Preferably, S4 is as follows: After generating the tag freeze signal, establish a behavior monitoring annotation, write the path adoption offset sequence and path offset inertia parameter set at the time the tag freeze signal is generated into the behavior monitoring annotation, and bind the customer identifier for continuous comparison of subsequent path behavior trend changes. After each subsequent transportation task is completed, the newly generated path offset labels are added to the path adoption offset sequence in the order of the transportation task completion time, forming an extended record that includes path behavior before and after freezing, and the extended path adoption offset sequence is written into the behavior monitoring annotation; Based on the path adoption offset sequence after extended recording, the path offset direction consistency rate, path recurrence count, and path dwell inertia span are re-extracted to calculate the current path offset inertia parameter set. This set is then compared at the parameter level with the historical path offset inertia parameter set recorded in the behavior monitoring annotations to obtain the degree of offset between the current and historical data, which is used to continuously track path behavior trends.

[0013] Preferably, S5 is as follows: Based on the updated path offset inertial parameter set in the behavior monitoring annotation, the current path offset direction consistency rate, path recurrence count, and path dwell inertial span are compared with the corresponding parameters in the historical path offset inertial parameter set recorded when the tag freeze signal is generated. This results in three sets of parameter offset amplitude data. The offset amplitude of each parameter is then compared with the corresponding release judgment threshold. When the offset amplitude of all three sets of parameters exceeds their corresponding release judgment threshold, it is determined that the freeze release condition is met. If the conditions for freezing and releasing are met, the freeze signal of the tag associated with the customer identifier will be set to invalid, the tag will be unfrozen, and the current state switch time and state change result will be recorded in the behavior monitoring annotation. After the label freeze is lifted, the expanded path adoption offset sequence is reintegrated into the profile label update process, and path adoption class labels are calculated based on the path adoption offset sequence to restore the dynamic control of customer path profile labels.

[0014] The technical effects and advantages provided by the present invention in the above technical solution are as follows: 1. This invention enhances the ability of customer profiling systems to identify the stability of transportation behavior by constructing a dynamic adjustment mechanism for profile tags based on path behavior characteristics. By introducing the construction of path adoption offset sequences, the generation of path offset tags, and the extraction of path offset inertia parameter sets, the system can accurately identify whether the deviation behavior possesses stability and business decision-making intent when a customer deviates from the recommended path for an extended period. Furthermore, by combining the generation of frozen confidence intervals and tag freezing signals, conditional static update control of path adoption tags is achieved, effectively avoiding erroneous changes in profile tags caused by the inability to distinguish between occasional deviations and stable strategies. This solution overcomes the deficiency of traditional customer profiling technologies in lacking stability judgment and response mechanisms when dealing with path deviation behavior.

[0015] 2. This invention, through the design of a behavior monitoring and annotation mechanism, continuously tracks the evolution trend of customer path behavior in a frozen state. It also establishes a freeze release judgment and dynamic recovery channel to ensure that the customer profile can respond promptly when actual behavior undergoes a stable change, automatically unfreezing and resuming the tag update process. This achieves synchronous adaptation between the profile system and changes in actual customer behavior. The overall solution possesses significant advantages in accurate behavior recognition, closed-loop control logic, and adjustable update mechanisms. While ensuring the stability of the customer profile, it also considers timely feedback on dynamic changes in individual behavior, enhancing the profile system's intelligent decision support capabilities in scenarios such as transportation scheduling, route recommendation, and customer strategy matching. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0017] Figure 1 This is a flowchart illustrating the freight forwarding customer profile generation method based on transportation route optimization according to the present invention. Detailed Implementation

[0018] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0019] This invention provides, for example Figure 1 The method for generating freight forwarder customer profiles based on transportation route optimization, as shown, specifically includes the following steps: S1. Map the recommended route and the actual route for each transportation task of the freight forwarding customer to a complete route. Generate a route offset label based on the overlap of route nodes, the trend of route offset direction, and the rejection of route selection. Form a route adoption offset sequence according to the order of task completion time to determine whether the customer has been using non-recommended routes for a long time. In this embodiment, S1 specifically includes the following steps: S101. Compare the path node sequence corresponding to the recommended path in each transportation task of the freight forwarding customer with the path node sequence corresponding to the actual path node, establish a mapping relationship between the recommended path node and the actual path node based on the spatial location correspondence of the path node, and form a complete path mapping covering the entire recommended path and the actual path based on the mapping relationship. In freight transportation, to accurately model the offset behavior between the customer's recommended route and the actual route, it is necessary to first establish a complete mapping relationship between the recommended and actual routes. The first step in this operation is to extract the path node sequences from the recommended and actual routes. Each path node can include points with clear spatial identification, such as road intersections, highway toll stations, transfer stations, or key turning coordinates. By traversing the recommended path node sequence, each node is compared one-to-one with the nodes in the actual path node sequence, and a node correspondence is established when the spatial distance is within an allowable threshold range. The comparison method can use Euclidean distance, GPS coordinate spherical distance, or spatial projection similarity judgment. When multiple consecutive nodes in the recommended route can be found in the actual route, it can be considered as forming a spatial mapping of a local path segment. By accumulating the mapping relationships of multiple local path segments, a complete path mapping covering the entire recommended and actual routes is finally obtained. This mapping result can not only be used to calculate the overlap of path nodes, but also provide a structured data foundation for analyzing the offset direction trend and route selection rejection between the two paths, thereby improving the modeling accuracy of path behavior in customer profiles.

[0020] The recommended path node sequence refers to the set of all key nodes in the ideal transportation route output by the path planning algorithm, arranged sequentially. The actual path node sequence, on the other hand, originates from the set of continuous trajectory points recorded by positioning devices during the vehicle's actual journey. Node-by-node comparison involves selecting each node sequentially from the recommended path node sequence and attempting to find the spatially closest node in the actual path node sequence, then judging their spatial distance and orientation angle to determine if a valid match exists. The spatial correspondence of path nodes is a spatial consistency relationship between point pairs, representing the degree of proximity of the geographical locations of recommended and actual path nodes in two-dimensional or three-dimensional space. The mapping relationship between recommended and actual path nodes refers to a set of correspondence relationships composed of several sets of spatially matching nodes, used to describe the correspondence pattern of the two paths at the key node level. A complete path mapping covering the entire recommended and actual paths means that the mapping relationship encompasses the entire route from the starting point to the end point, allowing subsequent calculations to perform consistency analysis across the entire path without being affected by local path offsets or interruptions, thus ensuring the global accuracy of the analysis results.

[0021] S102. Calculate the path node overlap between the recommended path and the actual path based on the complete path mapping, extract the overall path offset direction trend of the actual path relative to the recommended path, and determine the road segment selection rejection degree based on the actual path's avoidance of specific road segments in multiple transportation tasks. Combine the path node overlap degree, path offset direction trend and road segment selection rejection degree to generate the path offset label corresponding to each transportation task. S103. Arrange the path offset labels according to the order of completion time of the transportation task to form a path adoption offset sequence. Divide the path adoption offset sequence into segments based on whether adjacent path offset labels are consecutive non-recommended path types. When the number of non-recommended path offset labels in a consecutive segment exceeds a preset threshold and the trend of path offset direction does not reverse within the consecutive segment, it is determined that the customer has been using non-recommended paths for a long time.

[0022] To determine whether a customer consistently uses non-recommended routes in their transportation behavior, a path adoption offset sequence can be constructed to analyze the persistence and stability of their path deviation. First, the path offset labels generated for each transportation task are arranged chronologically according to the task completion time, forming a path adoption offset sequence. Each element in this sequence represents the path adoption situation within a single task. Then, the path adoption offset sequence is traversed, checking if adjacent path offset labels are consecutively non-recommended path types. If consecutive segments of non-recommended path labels appear, these segments are marked and extracted. Within these marked consecutive segments, the number of non-recommended path labels is further calculated and compared to a preset threshold. When the number exceeds this threshold, it indicates that the customer exhibits a significant preference for non-recommended paths during that time period. Simultaneously, the trend of path offset direction is compared within these consecutive segments to determine if the trend remains consistent. If there is no significant reversal, i.e., no sudden shift from one direction to the opposite direction, the customer's offset behavior can be considered stable, thus indicating that the customer consistently uses non-recommended paths. This judgment mechanism not only avoids misjudgment based on single behaviors but also effectively identifies customer behavior patterns involving strategic path selection.

[0023] The route adoption offset sequence is a time series composed of multiple transportation task route offset labels arranged chronologically, used to observe the evolution trend of route adoption behavior. Non-recommended route types refer to route labels where customers deviate from the recommended route during actual transportation and exhibit clear offset characteristics. Segment segmentation involves identifying consecutively occurring non-recommended route type label segments within the route adoption offset sequence to assess behavioral persistence. A preset quantity threshold is a crucial parameter for measuring behavioral frequency; its setting is based on statistical experience or industry standards and defines the minimum requirement for "continuous behavior." Behavior persistence is considered only when the number of non-recommended route labels in a consecutive segment exceeds this threshold. No reversal occurs when the route offset direction maintains a consistent trend within the consecutive segment, without sudden shifts from southeast to northwest, for example. This analysis is a crucial condition for judging behavioral stability; only when the directional trend is stable can it be reasonably concluded that the customer's transportation path has a long-term tendency to deviate from the recommended route. Through the above judgment mechanism, customer route selection preferences can be efficiently identified without introducing complex models, supporting subsequent profile label freezing strategies.

[0024] In this embodiment, S102 specifically refers to: Based on the complete path mapping, the path node sequence corresponding to the recommended path and the path node sequence corresponding to the actual path are extracted. By statistically analyzing the ratio of the number of spatially matched nodes in the two sets of path node sequences to the total number of path nodes, the path node overlap between the recommended path and the actual path is calculated. To measure the degree of overlap between the recommended path and the actual path at the node level, the path node overlap degree can be generated by calculating the ratio between the number of spatially matched nodes of the two paths and the total number of nodes, based on the path node sequence extracted from the complete path mapping. In the specific implementation, firstly, the path node sequences of the recommended path and the actual path are extracted, with each node corresponding to a spatial coordinate point. Then, according to the order of the nodes in the recommended path, each node is spatially compared with the nodes in the actual path, and the geographical distance between the two nodes is calculated to determine whether they meet the matching conditions. The matching conditions can be based on a preset spatial distance threshold (e.g., within 50 meters) or a more precise spatial overlap judgment method, such as considering the consistency of road direction or the similarity of path shape at the nodes. When a node in the recommended path finds a matching node in the actual path that meets the conditions, it is counted as a matched node. The ratio of all successfully matched nodes to the total number of nodes in the recommended path is calculated to obtain the path node overlap degree. A higher value indicates a higher degree of actual use of the recommended path by the customer, while a lower value indicates a significant deviation between customer behavior and the recommended path. Path node overlap is a key parameter reflecting path consistency, quantitatively describing whether customers adopt recommended paths. It forms a crucial foundation for generating path offset labels and analyzing transportation behavior preferences. For example, if a recommended path contains 20 nodes, and 15 nodes in the actual path spatially match those in the recommended path, the path node overlap is 75%, indicating high consistency in path execution for the transportation task. This calculation not only provides quantitative support for customer profiling but also offers a core metric for identifying non-adoption behaviors and assessing behavioral stability.

[0025] Based on the spatial arrangement order of each path node in the complete path mapping, the overall offset direction change trajectory of the actual path relative to the recommended path is calculated, and by performing consistency analysis on the offset direction of continuous path nodes, the overall path offset direction trend reflecting the actual path relative to the recommended path is extracted. To identify stable deviation characteristics in customer route selection, the overall deviation trajectory of the actual path relative to the recommended path can be calculated based on the spatial arrangement order of path nodes in the complete path mapping. Consistency analysis is then performed on the deviation directions of consecutive path nodes to extract the overall deviation trend of the actual path relative to the recommended path. Specifically, spatial coordinate pairs of the recommended and actual paths under the same node index are first obtained from the complete path mapping. The deviation direction of each node is calculated based on the relative spatial position between each pair of coordinates, for example, determining whether the actual path node is located in the east, south, west, north, or northeast direction of the recommended path node. Then, according to the order of the recommended path nodes, all deviation directions are arranged sequentially, forming a trajectory curve representing the process of deviation direction change. To determine whether this deviation has a directional trend, a consistency analysis is further performed on this trajectory, i.e., detecting the proportion of consecutive segments with the same deviation direction. If the proportion is significant and the deviation direction does not change frequently, it indicates that the customer exhibits a stable spatial deviation preference in route selection. For example, if the actual route nodes shift more than ten nodes southeast of the recommended route nodes, and the shift direction shows no significant change throughout the entire route segment, it indicates that the customer has a long-term preference for using the southeast alternative route. This overall route shift trend reflects the customer's strategic route adjustments, rather than occasional detours, providing a crucial basis for further identifying the stability of customer transportation behavior and developing a profile label freezing strategy. This method can capture spatial preference characteristics in route selection, making customer profiles closer to their actual transportation habits and avoiding misjudgments of customer behavior due to single route differences.

[0026] Based on the complete path mapping results of multiple transportation tasks, the set of road segments that were repeatedly detoured or not used in the actual path is counted. The frequency of occurrence of each road segment being avoided in multiple transportation tasks is calculated to determine the road segment selection rejection degree. The path node overlap degree, path offset direction trend and road segment selection rejection degree are combined and mapped to generate the path offset label corresponding to each transportation task.

[0027] In constructing route profiles for freight behavior, to identify customers' tendency to avoid specific routes, it is necessary to statistically analyze recommended route segments that are repeatedly detoured or consistently not used in the actual routes, based on the complete route mapping results generated from multiple transportation tasks. This allows for the extraction of representative route avoidance behavior features. Specifically, the recommended routes are first divided into several continuous route segment units, and it is then counted whether these segments appear in the actual route mapping for each transportation task. If a route segment is repeatedly not covered by the actual route, it is considered that the customer is actively detouring from that route segment. These segments are then aggregated to form a set of repeatedly detoured or unused route segments. The frequency at which each route segment is avoided by customers in all transportation tasks is calculated, i.e., the ratio between the number of times a segment does not match and the total number of transportation tasks. This ratio represents the route selection aversion degree. A higher route selection aversion degree indicates that the customer is more inclined to avoid that route segment, reflecting their clear preference for route selection. Subsequently, the overlap of path nodes, the trend of path deviation direction, and the repulsion of route selection in the current task are combined and mapped as three behavioral parameters for each transportation task. This can be achieved by constructing a weight matrix or a normalized vector space for fusion, ultimately generating a path deviation label that reflects the customer's path selection deviation characteristics in the transportation task. The path deviation label is used to identify the overall behavioral difference level of the customer relative to the recommended path in this transportation task, and serves as the basis for subsequently constructing the path adoption deviation sequence, judging the stability of path behavior, and evaluating the conditions for freezing the profile label. This method, by introducing multi-task historical behavioral parameters, achieves systematic modeling of path selection strategies, improving the accuracy and robustness of freight forwarding customer profiles in deviation identification.

[0028] S2. When customers consistently use non-recommended paths over a long period, extract the path offset direction consistency rate, path recurrence frequency, and path dwell inertia span based on the path adoption offset sequence, and combine them to generate a path offset inertia parameter set to determine the stability of the customer's path behavior. In this embodiment, S2 specifically includes the following steps: S201. When customers continue to use non-recommended paths for a long time, based on the non-recommended path offset labels that appear consecutively in the path adoption offset sequence, count the path offset direction corresponding to each offset label, and calculate the proportion of the number of times the same offset direction appears in consecutive offset labels to the total number of consecutive offset labels, so as to extract the path offset direction consistency rate. Based on the number of times the path offset label corresponding to the same actual path appears in the path adoption offset sequence, count the number of path recurrences. Under the condition that customers consistently use non-recommended routes, all route offset tags are first sorted according to the completion time of the transportation task, and consecutive non-recommended route offset tag sequences are extracted. Each non-recommended route offset tag contains the actual route offset direction relative to the recommended route in the current task, such as an overall eastward offset, a northward offset, or a specific angular deviation. By statistically analyzing the frequency of each type of offset direction in these tags and calculating the proportion of the most frequent single direction in consecutive tags, the route offset direction consistency rate can be extracted, thereby determining whether the customer maintains consistent directional behavior across multiple tasks. To further identify the customer's dependence on specific routes, the actual routes corresponding to the route offset tags are filtered from the route adoption offset sequence, and the number of times the same route appears repeatedly in the sequence is recorded to calculate the route recurrence count. The route recurrence count and the route offset direction consistency rate jointly reflect the stability of the customer's preference for non-recommended routes. This method can effectively identify customer types that "seem to deviate from the recommended route but whose behavior is extremely stable," avoiding erroneous tag changes.

[0029] The consecutive non-recommended route offset labels in the route adoption offset sequence indicate that the customer has not used the recommended route in multiple transportation tasks, and each actual route deviates structurally from the recommended route. These offsets are continuously distributed over time, forming the basis data for stable offset behavior. The route offset direction corresponding to each offset label describes the directional deviation of the actual route relative to the recommended route on a two-dimensional map, such as the consistency of the offset angle and the uniformity of the directional trend. The proportion of the same offset direction appearing in consecutive offset labels to the total number of consecutive offset labels is a quantitative indicator of whether the customer consistently maintains the same offset logic. A higher proportion indicates a more stable offset direction, from which the route offset direction consistency rate is extracted. The number of times the route offset label corresponding to the same actual route appears in the route adoption offset sequence refers to the customer repeatedly choosing the exact same actual route in different transportation tasks. This statistic generates the route recurrence count, reflecting the customer's repeated preference and dependence on a particular transportation route, and is an important parameter for measuring behavioral inertia and stable route selection.

[0030] S202. Based on the continuous distribution of path offset labels corresponding to the same actual path in the path adoption offset sequence in the time dimension, determine the continuous duration of the customer in the state of keeping the same actual path unchanged, extract the path dwell inertia span with the continuous duration, and combine the path offset direction consistency rate, path recurrence times and path dwell inertia span after uniform dimension processing to generate a path offset inertia parameter set. When a customer consistently uses a non-recommended route, the first step is to sort each route offset label in the route adoption offset sequence by time and group them according to their corresponding actual routes, identifying label segments that repeatedly select the same actual route. For each segment of consecutive use of the same actual route, the number of transportation tasks traversed by that segment is calculated, which represents the continuous duration of the customer on that actual route. By statistically analyzing the duration of multiple groups of continuous routes, the customer's inertial behavior cycle without changing transportation routes can be extracted, and the route dwell inertial span can be calculated to characterize the customer's tendency to dwell on a specific transportation route. Since the route offset direction consistency rate, the number of route recurrences, and the route dwell inertial span have different unit dimensions, to avoid interference with subsequent evaluations, normalization or standard deviation standardization is required to process the three indicators separately, bringing them to the same numerical scale. After standardization, the three indicators are combined to form a route offset inertial parameter set, thus providing a multi-dimensional and comparable quantitative basis for subsequent stability assessments of customer route behavior. This process can effectively eliminate the risk of misjudgment caused by inconsistent indicator dimensions, and ensure that the stability assessment of path behavior is more credible.

[0031] The continuous distribution of path offset labels over time refers to the sequence of offset labels ordered by the completion time of transportation tasks, where adjacent labels correspond to the same actual path and are not interrupted by other paths, forming a continuous adoption record. The continuous duration of a customer's use of the same actual path reflects their willingness to continuously use a non-recommended path; a higher value indicates more stable and fixed customer behavior. Path dwell inertia span is a quantitative expression of this continuous usage length, representing the customer's established habitual preference in path decisions. Standardized dimensional processing refers to standardizing or normalizing the path offset direction consistency rate, path recurrence frequency, and path dwell inertia span to address the inability to directly compare different data dimensions. The path offset inertia parameter set is a numerical set composed of the processed three-dimensional indicators. This set serves as a core feature input reflecting the stability of customer path preferences, supporting subsequent stability assessments and profile update control decisions.

[0032] S203. Based on the value distribution of each parameter in the path offset inertia parameter set in the path adoption offset sequence, construct a parameter determination interval to distinguish different stability levels. When the entire path offset inertia parameter set falls into the corresponding stability interval, determine the stability level of the customer's path behavior.

[0033] The core purpose of this approach is to address the technical challenge of effectively identifying the stability of customer path behavior in traditional path profiling updates. By constructing parameter judgment intervals based on the value distribution of each parameter in the path offset inertia parameter set within the path adoption offset sequence, the path selection patterns of customers during long-term transportation can be quantified into specific, comparable numerical segments. This allows for the differentiation between stable preferences and occasional temporary offsets. This approach introduces more precise classification criteria, enabling the system to move beyond simply relying on the consistency between paths and recommendation results when facing continuous path offsets. Instead, it establishes a mechanism for recognizing the customer's true path selection logic based on the continuity, repetition, and inertia of path behavior itself. This fundamentally improves the accuracy of customer profile label determination and prevents label drift and erroneous profile updates caused by short-term behavioral fluctuations.

[0034] In this embodiment, S203 specifically refers to: Based on the path adoption offset sequence, the historical value distribution ranges of path offset direction consistency rate, path recurrence frequency and path dwell inertia span in multiple customer transportation tasks are extracted respectively. Based on the variation range and central tendency of each parameter, the single parameter value threshold range used to represent stable behavioral characteristics is determined. To assess the stability of customer route behavior, it is necessary to extract complete records of route offset direction consistency rate, route recurrence frequency, and route dwell inertia span for multiple customers in historical transportation tasks based on route adoption offset sequences, and to construct historical value distribution intervals for each parameter. This operation can be performed by statistically analyzing the parameter value sets formed by different customers over multiple independent transportation cycles, calculating their minimum, maximum, average, and median values, and plotting their probability density distribution curves, thereby clarifying the value trend and fluctuation range of the parameter in actual business. After understanding the variation range and central tendency of each parameter in business practice, a single parameter value threshold range representing behavioral stability can be set based on methods such as standard deviation, high-density intervals, or 95% confidence intervals. For example, when the route offset direction consistency rate remains above a certain fixed value for a long period and its historical fluctuations are minimal, the parameter can be considered to exhibit stable characteristics. Similarly, for the route recurrence frequency and route dwell inertia span, stability can be determined only when their values ​​are significantly higher than the historical average level and within a high-density clustering interval. This process constructs quantifiable and comparable standard threshold ranges, providing clear reference for subsequent stability assessments, effectively avoiding subjectivity and ambiguity in the assessment process, and improving the scientific rigor and accuracy of profile label control.

[0035] After setting stable judgment segments for path offset direction consistency rate, path recurrence count and path dwell inertial span, the current value of each parameter in the path offset inertial parameter set is compared with the corresponding stable judgment segment to identify whether each parameter falls into the corresponding stable judgment segment, thereby constructing the stable hit status of the current path offset inertial parameter set. To accurately determine the stability of path behavior, corresponding stability judgment segments need to be set for the path offset direction consistency rate, path recurrence frequency, and path dwell inertia span. The stability judgment segment is a standard threshold range defined based on the historical value distribution interval to distinguish between stable and unstable behavior. For example, a high-stability region can be determined by setting the median and upper / lower quantile ranges of the parameters. After obtaining the current values ​​of the path offset inertia parameter set, the current value of each parameter is compared with its corresponding stability judgment segment one by one. If the value of a parameter falls into the corresponding stability judgment segment, the parameter is recorded as "hit"; otherwise, it is recorded as "missed". Statistical analysis of the comparison results of the three core parameters forms the stability hit status of the current path offset inertia parameter set. This stability hit status is usually quantified by the number of hit parameters, the hit ratio, or the hit combination type. For example, if all three parameters are hit, the customer's path behavior can be considered highly stable; if only one parameter is hit or none are hit, the behavior does not yet possess stability. This judgment process not only provides key criteria for subsequent label freezing, but also enhances the structure and interpretability of the judgment process, making the stability assessment more scientific and accurate.

[0036] Based on the stability hit statistics of the path offset inertial parameter set, the number of parameters falling into the stability judgment segment is counted. When the current value of all parameters falls into their corresponding stability judgment segment at the same time, it is determined that the stability of the customer's path behavior has reached a stable state, which is used to support the status control operation of the profile tag.

[0037] To ultimately determine whether customer path behavior has reached a stable state, it is necessary to count the number of parameters falling into their respective stable judgment segments based on the stable hit status of the path offset inertia parameter set. In the specific implementation, firstly, the current values ​​of three parameters—path offset direction consistency rate, path recurrence frequency, and path dwell inertia span—are checked one by one to see if they fall into their respective preset stable judgment segments. Each time a parameter meets the condition, it is considered a hit. After counting the hit results of all parameters, if the current values ​​of all three parameters hit their corresponding stable judgment segments, it can be determined that the customer path behavior exhibits high consistency, high frequency of recurrence, and continuous inertia at this stage, thus determining that the customer path behavior has reached a stable state. This confirmation of a stable state is a formal data-level assessment of the customer's behavior of consistently refusing to adopt recommended routes and insisting on using their preferred routes during transportation tasks. This judgment result will serve as the core basis for the profile tag control logic, providing stable support for whether to execute tag freezing operations subsequently, avoiding erroneous tag updates due to short-term behavioral fluctuations, and effectively improving the long-term effectiveness and accuracy of the profile.

[0038] S3. Map the path offset inertia parameter set to the frozen confidence interval, determine whether to generate a label freeze signal, which is used to indicate that the customer's path behavior has entered a stable preference state and trigger the corresponding profile label update stop operation. In this embodiment, S3 specifically refers to: After normalizing the path offset direction consistency rate, path recurrence count, and path dwell inertia span in the path offset inertia parameter set, a joint evaluation vector is constructed. Based on the stability distribution of the joint evaluation vector in historical task data, the path offset inertia parameter set is mapped to a frozen confidence interval to reflect the confidence level achieved by the offset inertia characteristics of customer path behavior. To accurately identify whether a customer has entered a stable path preference state, it is necessary to standardize the different dimensions of the path offset inertia parameter set. A min-max normalization method can be used to map the values ​​of path offset direction consistency rate, path recurrence frequency, and path dwell inertia span to the same fixed interval, such as 0 to 1, to eliminate the influence of inconsistent numerical dimensions. The three normalized parameters are then used to construct a three-dimensional joint evaluation vector, representing the customer's path offset inertia characteristics in the current time period. Subsequently, a multi-dimensional probability distribution model of the joint evaluation vector is constructed from historical transportation task data, such as using density estimation methods or cluster boundary identification, to delineate stable regions appearing in the joint evaluation vector set. Finally, based on the position of the current customer's evaluation vector within this stable distribution, a frozen confidence value is mapped, reflecting the confidence level of the current offset inertia. This frozen confidence value will be used to determine whether the customer's current path behavior has sufficient stability, thereby supporting the decision on whether to generate a tag freezing signal.

[0039] Normalization unifies multiple indicators to a single metric, enabling effective comparison of path deviation direction consistency rate, path recurrence frequency, and path dwell inertia span. The joint evaluation vector is an ordered set of values ​​containing the three normalized parameters, used to characterize the comprehensive features of customer path deviation behavior in three-dimensional space. The stability distribution of the joint evaluation vector in historical task data refers to the clustered region of joint vector values ​​formed by most customers before and after entering a stable path preference state in a large number of transportation tasks. The frozen confidence interval is a continuous numerical segment defined based on this stability distribution, used to measure the closeness between the current customer deviation inertia characteristics and historical stable characteristics; higher confidence indicates that the customer behavior is closer to a perceived stable state. This mapping process is the key foundation for freeze determination, possessing the ability to accurately judge the state of customer behavior.

[0040] Set the freezing critical range in the freezing confidence interval, and compare the mapping result of the path offset inertial parameter set with the freezing critical range. When the entire path offset inertial parameter set falls into the freezing critical range, generate a label freezing signal to indicate that the customer's path behavior has entered a stable preference state. To accurately identify the stability of customer path behavior, a frozen critical range can be set within the frozen confidence interval to identify behavior states that have reached high stability. The frozen confidence interval is typically a continuous segment defined between 0 and 1, with the range closest to 1 designated as the frozen critical range, for example, 0.85 to 1. After the path offset inertia parameter set is mapped to a joint confidence value, it is compared with the frozen critical range to determine if the value falls within a highly reliable behavior offset inertia state. If the mapping result falls within the frozen critical range, it indicates a high consistency rate in the current path offset direction, frequent path recurrence, and a long path dwell inertia span, comprehensively reflecting that the customer path behavior has stabilized. Therefore, a tag freezing signal can be generated. This signal serves as a trigger condition for marking changes in customer behavior states, pausing the dynamic updating of customer profile tags in subsequent processes to prevent erroneous offset changes in profile tags and ensure their accuracy and long-term stability.

[0041] The frozen confidence interval is a quantitative segment used to measure the strength of path offset inertia. Formed through historical data analysis, it assesses the stability trend of customer path behavior. The frozen critical range is a further numerical sub-interval defined within the frozen confidence interval. It identifies behavioral data with significantly stable characteristics and is generally located at the high end of the confidence interval, used to rigorously determine whether the freezing conditions are met. The mapping result of the path offset inertia parameter set, i.e., the current customer's comprehensive behavioral score in the path offset dimension, is a key input for the freezing determination. The tag freezing signal is a judgment signal used to control changes in the customer profile tag status. The significance of generating this signal is to clearly indicate that the customer has entered a stable path behavior stage, thereby triggering a pause in tag updates in the profile system, thus avoiding misleading intelligent recommendations caused by continuous tag changes. This signal is not generated based on a single indicator but rather on a highly reliable judgment result based on the stability of a multi-parameter combination.

[0042] After generating the tag freeze signal, the tag freeze signal is associated with the path adoption class tag in the customer profile, and the signal is used as the control trigger to execute the update stop operation of the path adoption class tag, so that the path adoption class tag no longer changes with the subsequent path adoption offset sequence.

[0043] To achieve precise control over path adoption tags in customer profiles, a direct correlation needs to be established between the tag freeze signal and the path adoption tags in the customer profile after the freeze signal is generated. This can be achieved by setting a mapping control table to bind each customer identifier to the freeze status of its path adoption tag. When the freeze signal is marked as active, the system uses the freeze signal as a control criterion during the customer profile tag update process, intercepting tag changes originally triggered based on the latest path adoption offset sequence. Specifically, the tag update logic first checks whether the corresponding customer is in a frozen state. If it is, the path offset sequence analysis process used to recalculate tag values ​​is not invoked; instead, the tag values ​​remain unchanged from before the freeze. This mechanism ensures that for customers whose path behavior stability has been confirmed, their profile tags are no longer affected by occasional short-term path adjustments or abnormal transportation tasks, thereby improving the long-term representativeness of the tags and the quality of the profile.

[0044] Route adoption tags in customer profiles record customer responses to recommended routes during historical transportation tasks. These tags typically include fields such as route acceptance level, route deviation frequency, or route adoption ratio. Essentially, these tags reflect customer behavioral feedback to route optimization suggestions and are key variables used to predict customer preferences in intelligent scheduling and service recommendation. The route adoption tag update halt operation refers to the system proactively suspending the regular update process for a customer's route adoption tags when the customer's route behavior is identified as highly stable. This operation is not a permanent clearing but rather a dynamic control guided by a freeze signal, allowing tags to resume updates when customer behavior changes again. This design avoids frequent changes in profile tags due to short-term fluctuations, thus ensuring the stability and accuracy of the data relied upon by the freight forwarding service system when making intelligent decisions such as route recommendation and capacity matching.

[0045] S4. After generating the tag freeze signal, establish behavior monitoring annotations, expand the record of the path adoption offset sequence generated by subsequent transportation tasks, continuously track the path behavior trend and calculate the degree of offset between the path and historical parameters. In this embodiment, S4 specifically refers to: After generating the tag freeze signal, establish a behavior monitoring annotation, write the path adoption offset sequence and path offset inertia parameter set at the time the tag freeze signal is generated into the behavior monitoring annotation, and bind the customer identifier for continuous comparison of subsequent path behavior trend changes. After generating the tag freeze signal, a behavior monitoring annotation needs to be established to enable continuous observation of customer path behavior. This behavior monitoring annotation can be achieved by constructing a data record entry bound to a customer identifier. This record entry contains the path adoption offset sequence and path offset inertia parameter set corresponding to the time the tag freeze signal was generated. This data can be stored in a customer behavior monitoring table in the database, with a key-value pair mapping each customer identifier to its behavior state at the time of freeze. The path adoption offset sequence records the time-ordered sequence of path offset tags generated by the customer in each transportation task before the freeze. The path offset inertia parameter set is generated by combining the path offset direction consistency rate, path recurrence frequency, and path dwell inertia span. Packing these key behavior parameters into the monitoring annotation at the time of freeze allows for future identification of whether the customer has experienced behavior drift or path strategy adjustments. For example, if a customer's transportation route changes multiple times after the freeze, the parameters recorded at the time of freeze can be compared to assess whether their path behavior remains stable.

[0046] Behavioral monitoring annotations are structured data records used for long-term tracking of dynamic changes in customer path behavior, possessing the function of binding time location and behavioral features. Path adoption offset sequences are time-series data used to characterize the path adoption status of each customer's transportation task. Their key component is the path offset label, representing the degree of deviation between the recommended path and the actual path in each task. The path offset inertia parameter set is a quantitative set of behavioral stability generated through comprehensive analysis of indicators such as offset direction consistency, path recurrence frequency, and path maintenance span. Customer identifiers are unique identification codes used to identify a specific customer, which can be customer ID, tax ID, or registration information. This identifier ensures the unique attribution and rapid indexing capability of behavioral monitoring annotations in the database. Binding path behavior features to customer identifiers helps to achieve continuous tracking, offset monitoring, and dynamic management of individual customer path behavior.

[0047] After each subsequent transportation task is completed, the newly generated path offset labels are added to the path adoption offset sequence in the order of the transportation task completion time, forming an extended record that includes path behavior before and after freezing, and the extended path adoption offset sequence is written into the behavior monitoring annotation; After each subsequent transportation task is completed, a path mapping is performed between the recommended route and the actual route for that task. A path offset label is then generated based on node overlap, offset direction trends, and segment rejection. This offset label is then appended to the end of the existing path adoption offset sequence according to the completion time of the transportation tasks, thus gradually building an extended record containing path behavior before and after the freeze. This operation extends the original behavior sequence via a timeline, ensuring continuous connection between old and new data without requiring a reconstruction of the overall data structure. The updated path adoption offset sequence is rewritten into the behavior monitoring annotation and continues to be bound and stored with the customer identifier, enabling long-term tracking of changes in customer path behavior after the label freeze. Continuous recording allows for quantitative analysis of behavior trends, identifying whether customers have changed their path strategies, and providing data support for adjusting subsequent label freeze states. For example, if a customer uses a completely new route for five consecutive tasks after the freeze, the system can determine whether a new path preference trend exists based on the extended record.

[0048] "Extended records containing path behavior before and after freezing" refers to the process of sequentially concatenating newly generated transportation task path offset tags with existing path adoption offset tags from before the freeze, forming a complete timeline of path behavior after the tag freeze signal is generated. Path offset tags are abstract representations of the differences between the recommended and actual paths in each transportation task, while the path adoption offset sequence is an ordered set of these tags arranged chronologically. When adding new path offset tags to an existing sequence, the chronological order and tag type of the original tags must be preserved to ensure logical coherence throughout the sequence. Behavior monitoring annotations, as record objects bound to customer identifiers, carry updates to the extended path adoption offset sequence, reflecting whether the customer's behavior under the frozen tag state maintains stable preferences or exhibits a path strategy drift trend. This extended record mechanism not only supports dynamic tracking but also provides a fundamental data source for subsequent offset calculations and tag status control.

[0049] Based on the path adoption offset sequence after extended recording, the path offset direction consistency rate, path recurrence count, and path dwell inertia span are re-extracted to calculate the current path offset inertia parameter set. This set is then compared at the parameter level with the historical path offset inertia parameter set recorded in the behavior monitoring annotations to obtain the degree of offset between the current and historical data, which is used to continuously track path behavior trends.

[0050] After expanding the path adoption offset sequence, three key parameters need to be re-extracted based on the updated complete sequence: path offset direction consistency rate, path recurrence frequency, and path dwell inertia span. The extraction method can be based on the same calculation logic as before freezing, namely, statistically analyzing the consistency of the direction distribution of non-recommended path labels in the sequence, the recurrence frequency of the same path in the sequence, and the time span. By unifying the dimensions of the three parameters, a current path offset inertia parameter set is constructed. Subsequently, the historical path offset inertia parameter set recorded at the time of freezing the label is retrieved from the behavior monitoring annotations. The current parameter set is compared with the historical parameter set item by item, and the numerical deviation of each parameter is calculated to obtain the degree of offset between the current and historical values. This process can identify the magnitude of change by setting thresholds. For example, if the path recurrence frequency decreases significantly and the path offset direction changes, it indicates that the customer's path behavior strategy may have been adjusted, thus providing a basis for subsequent label freezing status judgment.

[0051] The current path offset inertia parameter set is a centralized abstraction of the path behavior characteristics of the expanded path adoption offset sequence within the current time period, reflecting the customer's recent inertial behavior in path selection. The historical path offset inertia parameter set is a set of parameters of the same type extracted based on the sequence at the time of frozen tag signal generation, representing the customer's original path preference characteristics. Parameter-level comparison compares each corresponding parameter in the current parameter set with that in the historical parameter set one by one, identifying whether the behavior remains consistent by calculating the numerical differences between the two. For example, when the consistency rate of path offset direction drops from 90% to 50%, it indicates that the customer may abandon their original preferred path. The degree of offset between the current and historical values ​​is an overall expression of the differences across the three parameter dimensions, used to quantitatively reflect whether path behavior deviates from its original stable state. The core significance of this operation lies in dynamically judging changes in customer behavior strategies through parameter evolution trends, achieving precise control of path tags in the customer profile.

[0052] S5. Based on the updated path offset inertia parameter set in the behavior monitoring annotation, determine whether the freeze release condition is met. If the freeze release condition is met, release the label freeze state and resume the profile label update process to achieve dynamic control of customer path profile labels.

[0053] In this embodiment, S5 specifically refers to: Based on the updated path offset inertial parameter set in the behavior monitoring annotation, the current path offset direction consistency rate, path recurrence count, and path dwell inertial span are compared with the corresponding parameters in the historical path offset inertial parameter set recorded when the tag freeze signal is generated. This results in three sets of parameter offset amplitude data. The offset amplitude of each parameter is then compared with the corresponding release judgment threshold. When the offset amplitude of all three sets of parameters exceeds their corresponding release judgment threshold, it is determined that the freeze release condition is met. After generating the tag freeze signal, the current path offset inertia parameter set can be dynamically updated by continuously recording the path adoption offset sequence of the customer's subsequent transportation tasks. To determine whether the customer has stopped maintaining the previously stable path selection preference, the current parameter set needs to be compared item by item with the historical parameter set recorded when the tag freeze signal was generated. During the comparison, the difference between the current path offset direction consistency rate and the historical value, the difference between the current path recurrence count and the historical value, and the difference between the current path dwell inertia span and the historical value need to be calculated, resulting in three sets of offset amplitude data. Each set of offset amplitude is then compared with the corresponding preset release judgment threshold. If the offset amplitude of all three parameters is greater than their respective release judgment thresholds, it indicates that the customer's path behavior has changed significantly and no longer maintains the stable state before freezing. At this time, it can be determined that the freeze release condition is met. This process is usually implemented through a path parameter monitoring model. The model continuously receives new task data and triggers the comparison logic to achieve rapid identification of changes in customer behavior trends. For example, if a customer’s continuously used new routes show a significant shift in direction and the types of routes used repeatedly decrease significantly, while the frequency of switching routes increases significantly, then the magnitude of these three deviations will exceed the release judgment threshold, and the system will determine that the customer’s behavior is no longer stable.

[0054] Path offset direction consistency rate indicates the stability of path selection in terms of directional trend; path recurrence count measures the intensity of a customer's repeated use of a particular actual path; and path dwell inertia span characterizes the length of time a customer continuously maintains a single path. Release judgment thresholds are numerical limits set for these three types of parameters in the system, reflecting the minimum degree of change required for customer behavior to transition from a "stable" to an "unstable" state. Release judgment thresholds are typically obtained through training on a large amount of historical customer transportation data, for example, using discrete point clustering or skewed distribution analysis to determine the lower limit of fluctuation for each type of parameter in an unstable state. When the current parameter set is compared with the historical parameter set at the time of freezing, if all three offset amplitudes exceed their respective release judgment thresholds, it indicates that the customer no longer continues their original stable path preference, thus meeting the tag freezing and release conditions. This mechanism is designed to prevent customer profile tags from remaining frozen for extended periods, ensuring that the tag change logic can adapt to changes in actual customer behavior and achieve dynamic control of customer path profile tags.

[0055] If the conditions for freezing and releasing are met, the freeze signal of the tag associated with the customer identifier will be set to invalid, the tag will be unfrozen, and the current state switch time and state change result will be recorded in the behavior monitoring annotation. After identifying that a customer's path behavior has met the conditions for freezing and releasing, the label freezing state needs to be promptly lifted through a state switching operation. Specifically, the label freezing signal bound to the customer identifier in the system is first marked as invalid, preventing it from being used as the control basis for the profile label update process. Next, the status flag of the profile label management module is updated, switching the customer path profile label from a frozen state to an updatable state, and allowing dynamic correction of the label based on path behavior offset data. After completing the state change, a timestamp of the current switch and a result identifier for lifting the freezing state need to be added to the behavior monitoring annotation for subsequent analysis and backtracking of behavior evolution trajectories. For example, if a customer has recently frequently changed their path selection, no longer maintaining the original path repetition and offset trend, and is deemed to have lost path behavior stability through threshold judgment, the system immediately marks the freezing signal as invalid, resuming dynamic updates of its label, ensuring that the profile system can closely adjust to changes in the customer's actual behavior.

[0056] The tag freeze signal is a control flag generated after a customer's path behavior enters a stable preference state, used to temporarily freeze the update operation of path profile tags. The customer identifier is an identity index used to uniquely bind information such as the tag freeze signal, path offset data, and monitoring records. After continuous behavior changes and the release judgment condition is met, the tag freeze signal must be set to an invalid state to prevent the profile tag from continuing to be bound by outdated data. Behavior monitoring annotations serve to record key state change moments, including the time of state transition, the background of parameter changes, and the old and new state tags, which helps the system perform future operations such as trend prediction, tag optimization, and behavior anomaly investigation. Through this mechanism, high-frequency dynamic adjustment of customer profile tags can be achieved, ensuring that customer feature recognition and path prediction remain highly synchronized with actual behavior.

[0057] After the label freeze is lifted, the expanded path adoption offset sequence is reintegrated into the profile label update process, and path adoption class labels are calculated based on the path adoption offset sequence to restore the dynamic control of customer path profile labels.

[0058] After the label freeze is lifted, the expanded route adoption offset sequence needs to be reintegrated into the profile label update process to achieve continuous dynamic adjustment of customer route profile labels. Specifically, the system retrieves the complete route adoption offset sequence, containing both pre-freeze and post-freeze route behaviors, from behavior monitoring annotations as the new base dataset for label calculation. The system traverses this sequence chronologically and, combined with the route offset labels generated for each transportation task, reassesses the changing characteristics of route adoption behavior, including the stability of route selection, the persistence of deviation trends, and the evolution trajectory of behavioral inertia. Subsequently, based on the latest behavioral trajectory, the system recalculates the customer's current route adoption category labels, such as preferred route type, detour level, and route change sensitivity. For example, if a customer continuously selects new routes after the freeze is lifted and exhibits new recurring offset characteristics, the system will update their route preference feature labels accordingly, ensuring that the label status accurately reflects the customer's current transportation decision-making behavior.

[0059] The path adoption offset sequence is a data sequence composed of multiple path offset labels arranged in chronological order, recording the deviation of customers from recommended path adoption in different transportation tasks. The extended path adoption offset sequence refers to the addition of path behavior data after the freeze, based on the original pre-freeze data, used to construct the evolution trajectory of customer behavior throughout its entire lifecycle. Path adoption tags are core tags in the profiling system reflecting customer path usage preferences and changing characteristics, typically including dimensions such as path repetition rate, change frequency, and avoidance intensity. Reintegrating the extended path adoption offset sequence into the tag update process ensures that these tags respond promptly to changes in customer behavior trends, avoiding information lag or bias in customer profiles after the freeze is lifted, thereby achieving continuous dynamic control of customer profile tags and high-fidelity mapping of behavioral trends.

[0060] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions according to the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means (e.g., infrared, wireless, microwave, etc.). A computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0061] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0062] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0063] In the several embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0064] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0065] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0066] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included 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 method for generating a customer portrait of a freight forwarder based on transportation path optimization, characterized by, Specifically, the following steps are included: S1. Map the recommended route and the actual route for each transportation task of the freight forwarding customer to a complete route. Generate a route offset label based on the overlap of route nodes, the trend of route offset direction, and the rejection of route selection. Form a route adoption offset sequence according to the order of task completion time to determine whether the customer has been using non-recommended routes for a long time. S2. When customers consistently use non-recommended paths over a long period, extract the path offset direction consistency rate, path recurrence frequency, and path dwell inertia span based on the path adoption offset sequence, and combine them to generate a path offset inertia parameter set to determine the stability of the customer's path behavior. S3. Map the path offset inertia parameter set to the frozen confidence interval, determine whether to generate a label freeze signal, which is used to indicate that the customer's path behavior has entered a stable preference state and trigger the corresponding profile label update stop operation. S4. After generating the tag freeze signal, establish behavior monitoring annotations, expand the record of the path adoption offset sequence generated by subsequent transportation tasks, continuously track the path behavior trend and calculate the degree of offset between the path and historical parameters. S5. Based on the updated path offset inertia parameter set in the behavior monitoring annotation, determine whether the freeze release condition is met. If the freeze release condition is met, release the label freeze state and resume the profile label update process to achieve dynamic control of customer path profile labels. 2.The method of claim 1, wherein, S1 specifically includes the following steps: S101. Compare the path node sequence corresponding to the recommended path in each transportation task of the freight forwarding customer with the path node sequence corresponding to the actual path node, establish a mapping relationship between the recommended path node and the actual path node based on the spatial location correspondence of the path node, and form a complete path mapping covering the entire recommended path and the actual path based on the mapping relationship. S102. Calculate the path node overlap between the recommended path and the actual path based on the complete path mapping, extract the overall path offset direction trend of the actual path relative to the recommended path, and determine the road segment selection rejection degree based on the actual path's avoidance of specific road segments in multiple transportation tasks. Combine the path node overlap degree, path offset direction trend and road segment selection rejection degree to generate the path offset label corresponding to each transportation task. S103. Arrange the path offset labels according to the order of completion time of the transportation task to form a path adoption offset sequence. Divide the path adoption offset sequence into segments based on whether adjacent path offset labels are consecutive non-recommended path types. When the number of non-recommended path offset labels in a consecutive segment exceeds a preset threshold and the trend of path offset direction does not reverse within the consecutive segment, it is determined that the customer has been using non-recommended paths for a long time. 3.The freight forwarder customer profiling method based on transportation path optimization of claim 2, wherein, S102 specifically refers to: Based on the complete path mapping, the path node sequence corresponding to the recommended path and the path node sequence corresponding to the actual path are extracted. By statistically analyzing the ratio of the number of spatially matched nodes in the two sets of path node sequences to the total number of path nodes, the path node overlap between the recommended path and the actual path is calculated. Based on the spatial arrangement order of each path node in the complete path mapping, the overall offset direction change trajectory of the actual path relative to the recommended path is calculated, and by performing consistency analysis on the offset direction of continuous path nodes, the overall path offset direction trend reflecting the actual path relative to the recommended path is extracted. Based on the complete path mapping results of multiple transportation tasks, the set of road segments that were repeatedly detoured or not used in the actual path is counted. The frequency of occurrence of each road segment being avoided in multiple transportation tasks is calculated to determine the road segment selection rejection degree. The path node overlap degree, path offset direction trend and road segment selection rejection degree are combined and mapped to generate the path offset label corresponding to each transportation task. 4.The method of claim 1, wherein, S2 specifically includes the following steps: S201. When customers continue to use non-recommended paths for a long time, based on the non-recommended path offset labels that appear consecutively in the path adoption offset sequence, count the path offset direction corresponding to each offset label, and calculate the proportion of the number of times the same offset direction appears in consecutive offset labels to the total number of consecutive offset labels, so as to extract the path offset direction consistency rate. Based on the number of times the path offset label corresponding to the same actual path appears in the path adoption offset sequence, count the number of path recurrences. S202. Based on the continuous distribution of path offset labels corresponding to the same actual path in the path adoption offset sequence in the time dimension, determine the continuous duration of the customer in the state of keeping the same actual path unchanged, extract the path dwell inertia span with the continuous duration, and combine the path offset direction consistency rate, path recurrence times and path dwell inertia span after uniform dimension processing to generate a path offset inertia parameter set. S203. Based on the value distribution of each parameter in the path offset inertia parameter set in the path adoption offset sequence, construct a parameter determination interval to distinguish different stability levels. When the entire path offset inertia parameter set falls into the corresponding stability interval, determine the stability level of the customer's path behavior. 5.The freight forwarder customer profiling method based on transportation path optimization of claim 4, wherein, S203 specifically refers to: Based on the path adoption offset sequence, the historical value distribution ranges of path offset direction consistency rate, path recurrence frequency and path dwell inertia span in multiple customer transportation tasks are extracted respectively. Based on the variation range and central tendency of each parameter, the single parameter value threshold range used to represent stable behavioral characteristics is determined. After setting stable judgment segments for path offset direction consistency rate, path recurrence count and path dwell inertial span, the current value of each parameter in the path offset inertial parameter set is compared with the corresponding stable judgment segment to identify whether each parameter falls into the corresponding stable judgment segment, thereby constructing the stable hit status of the current path offset inertial parameter set. Based on the stability hit statistics of the path offset inertial parameter set, the number of parameters falling into the stability judgment segment is counted. When the current value of all parameters falls into their corresponding stability judgment segment at the same time, it is determined that the stability of the customer's path behavior has reached a stable state, which is used to support the status control operation of the profile tag. 6.The method of claim 1, wherein, S3 specifically refers to: After normalizing the path offset direction consistency rate, path recurrence count, and path dwell inertia span in the path offset inertia parameter set, a joint evaluation vector is constructed. Based on the stability distribution of the joint evaluation vector in historical task data, the path offset inertia parameter set is mapped to a frozen confidence interval to reflect the confidence level achieved by the offset inertia characteristics of customer path behavior. Set the freezing critical range in the freezing confidence interval, and compare the mapping result of the path offset inertial parameter set with the freezing critical range. When the entire path offset inertial parameter set falls into the freezing critical range, generate a label freezing signal to indicate that the customer's path behavior has entered a stable preference state. After generating the tag freeze signal, the tag freeze signal is associated with the path adoption class tag in the customer profile, and the signal is used as the control trigger to execute the update stop operation of the path adoption class tag, so that the path adoption class tag no longer changes with the subsequent path adoption offset sequence. 7.The freight forwarder customer profiling method based on transportation path optimization of claim 1, wherein, S4 specifically refers to: After generating the tag freeze signal, establish a behavior monitoring annotation, write the path adoption offset sequence and path offset inertia parameter set at the time the tag freeze signal is generated into the behavior monitoring annotation, and bind the customer identifier for continuous comparison of subsequent path behavior trend changes. After each subsequent transportation task is completed, the newly generated path offset labels are added to the path adoption offset sequence in the order of the transportation task completion time, forming an extended record that includes path behavior before and after freezing, and the extended path adoption offset sequence is written into the behavior monitoring annotation; Based on the path adoption offset sequence after extended recording, the path offset direction consistency rate, path recurrence count, and path dwell inertia span are re-extracted to calculate the current path offset inertia parameter set. This set is then compared at the parameter level with the historical path offset inertia parameter set recorded in the behavior monitoring annotations to obtain the degree of offset between the current and historical data, which is used to continuously track path behavior trends. 8.The method of claim 1, wherein, S5 specifically refers to: Based on the updated path offset inertial parameter set in the behavior monitoring annotation, the current path offset direction consistency rate, path recurrence count, and path dwell inertial span are compared with the corresponding parameters in the historical path offset inertial parameter set recorded when the tag freeze signal is generated. This results in three sets of parameter offset amplitude data. The offset amplitude of each parameter is then compared with the corresponding release judgment threshold. When the offset amplitude of all three sets of parameters exceeds their corresponding release judgment threshold, it is determined that the freeze release condition is met. If the conditions for freezing and releasing are met, the freeze signal of the tag associated with the customer identifier will be set to invalid, the tag will be unfrozen, and the current state switch time and state change result will be recorded in the behavior monitoring annotation. After the label freeze is lifted, the expanded path adoption offset sequence is reintegrated into the profile label update process, and path adoption class labels are calculated based on the path adoption offset sequence to restore the dynamic control of customer path profile labels.

Images