Intelligent express batch quick taking system based on user behavior analysis

Abstract

The invention discloses an intelligent express batch quick taking system based on user behavior analysis, and relates to the technical field of express. The system comprises an information acquisition module used for acquiring identity identification information of a pickup user; the storage information acquisition module is used for calling target storage area information corresponding to the identity identification information; and the pick-up module is used for pushing the target storage area position where all the to-be-picked-up expresses are located and the express list in the area to the user according to the target storage area information, and the user goes to the target storage area to pick up all the goods. According to the method, the express storage position can be accurately positioned, the pick-up process is simplified, blind search of the user is avoided, and the batch pick-up efficiency of the user is improved.

Description

Technical Field

[0001] This invention belongs to the field of express delivery technology, specifically relating to an intelligent express delivery batch quick pickup system based on user behavior analysis. Background Technology

[0002] With the rapid development of e-commerce, the express delivery industry has experienced explosive growth. Last-mile delivery, as a crucial link in the entire express delivery chain, directly impacts user experience and industry operating costs through its efficiency and service quality. Currently, last-mile delivery has become the most challenging and difficult scenario in the express delivery chain. Traditional manual delivery models face problems such as large volume of business, low efficiency, and high operating costs. Furthermore, the management of traditional store stations requires significant manpower and effort, making it difficult to meet the demands of large-scale delivery. To alleviate the pressure on last-mile delivery, smart parcel lockers, with their 24-hour self-service storage and retrieval functions, have become an important platform. By the end of 2024, their nationwide stock exceeded 1.2 million sets, with an urban penetration rate of 85%, handling over 120 million parcels daily, effectively improving delivery efficiency.

[0003] However, neither smart parcel lockers nor traditional parcel stations analyze and apply user pickup habits and parcel volume. The allocation of shelves or compartments lacks specificity, requiring users to search for each parcel individually, which can easily lead to misplacing or missing parcels, resulting in low efficiency for users picking up parcels in bulk. Summary of the Invention

[0004] The purpose of this invention is to solve the problems of traditional express locker configuration that does not combine user behavior with resource allocation, which leads to easy errors and omissions in package retrieval and low efficiency in batch retrieval. Therefore, this invention proposes an intelligent express batch retrieval system based on user behavior analysis.

[0005] This invention proposes a system, wherein the intelligent express bulk pickup system based on user behavior analysis includes: The information acquisition module is used to acquire the identity information of the user picking up the package; The storage information acquisition module is used to retrieve the target storage area information corresponding to the identity information; The package pickup module is used to push the location of all the packages to be picked up to the user and the package list within the target storage area based on the target storage area information, and the user can go to the target storage area to pick up all the packages.

[0006] Optionally, the information acquisition module includes: The data acquisition module is used to collect original identity data by scanning the graphic pickup code; The identifier extraction module is used to parse the original identity data and extract the structured user identifier field; The verification module is used to match and verify the user identifier field with the locally cached list of valid users; The identity information generation module is used to output the successfully verified identity information after the matching verification is passed; the identity information is the last five digits of the user's mobile phone number.

[0007] Optionally, the storage information acquisition module includes: The extraction module is used to extract the user's unique identification code from the identity information; The query module is used to query the target storage area identifier bound to the user based on the user's unique identification code; The location acquisition module is used to acquire the physical location parameters corresponding to the target storage area identifier; the physical location parameters include at least the shelf number and the shelf plate number. The information generation module is used to obtain target storage area information based on the physical location parameters and the waybill number.

[0008] Optionally, the process of determining the target storage area includes: Acquire multi-dimensional user data; the multi-dimensional user data includes historical package pickup behavior data, package feature data, and real-time system status data; the historical package pickup behavior data includes average package volume, pickup frequency, and pickup time. Users are categorized into hierarchical tags based on the historical package pickup behavior data; Based on the hierarchical tags and the historical pickup behavior data, predict the user's pickup needs in the future preset period; The target storage area is allocated to the user based on the hierarchical label; Based on the prediction results and real-time system status, candidate target storage areas are allocated to users; When the time a package stays in the target storage area exceeds a preset time threshold, the package will be automatically moved to a candidate target storage area, and the candidate target storage area will be used as the updated target storage area.

[0009] Optionally, the hierarchical classification of users based on the historical pickup behavior data to obtain hierarchical tags includes: Users whose pickup frequency exceeds a preset first threshold and whose average package volume exceeds a preset second threshold are defined as high-frequency users, and users whose pickup frequency is lower than a third threshold are defined as low-frequency users. Extract users who are not defined as high-frequency hierarchical tags and low-frequency hierarchical tags to form a candidate user set; Construct a user feature vector set based on the historical parcel pickup behavior data of the candidate user set; Cluster analysis is performed on the user feature vector set to obtain the mid-frequency hierarchical user set; Each of the high-frequency, low-frequency, and mid-frequency user groups is assigned a corresponding hierarchical label to achieve hierarchical classification of all users.

[0010] Optionally, clustering analysis of the user feature vector set to obtain the mid-frequency hierarchical user set includes: An optimized feature set is obtained by performing multi-dimensional correlation analysis on the user feature vectors, followed by feature filtering, integration, and dimensionality reduction. Calculate the local density of each user feature point in the optimized feature set, and take the user points with local density greater than a preset density threshold as high-density points to obtain a high-density point set; For each high-density point in the high-density point set, calculate the minimum distance between it and all other high-density points, and identify the high-density points whose minimum distance is greater than a preset distance threshold as density core points to obtain a density core point set. For the pickup time dimension in the user feature vector, calculate the pickup time complementarity between different users, and group the users according to the pickup time complementarity to obtain a time complementarity candidate user set; An initial hierarchical set is constructed based on the density core point set and the time complementary candidate users; For each remaining feature point in the optimized feature set except for the density core point, calculate its feature similarity with the density core point in the initial level set, and assign the remaining feature points to the initial level with the highest feature similarity to obtain the initial level user set. Calculate the feature Euclidean distance between any two levels in the initial hierarchical user set, merge the two levels whose Euclidean distance is less than a preset merging threshold into a new level and assign a level identifier to obtain the mid-frequency hierarchical user set.

[0011] Optionally, the optimized feature set obtained by performing multi-dimensional correlation analysis on the user feature vectors, including feature filtering and dimensionality reduction integration, includes: The correlation coefficients for the pickup frequency dimension and the pickup time dimension, and the correlation coefficients for the average package volume dimension and the pickup time dimension are calculated based on the user feature vector to obtain the first correlation coefficient and the second correlation coefficient. If the first correlation coefficient is greater than the preset correlation threshold, the feature value of the pickup frequency dimension and the feature value of the pickup time dimension are weighted to obtain the first coupling feature; If the second correlation coefficient is greater than the preset correlation threshold, the feature value of the average volume dimension of the package and the feature value of the pickup time dimension are weighted to obtain the second coupling feature. Based on the comparison results of the first correlation coefficient and the second correlation coefficient with the preset correlation threshold, the optimized feature set is obtained by statistically analyzing all coupled features and the original feature dimensions that are determined to remain independent.

[0012] The beneficial effects of this invention are as follows: This invention proposes an intelligent bulk parcel retrieval system based on user behavior analysis. By acquiring the identity information of the user picking up the parcels, retrieving the corresponding target storage area information, and pushing the specific storage location of the parcels to be picked up and the parcel list within the area to the user, the system guides the user to retrieve all the parcels. This method can accurately locate the parcel storage location, simplify the retrieval process, avoid users searching blindly, and improve the efficiency of bulk parcel retrieval. Attached Figure Description

[0013] The present invention will now be further described with reference to the accompanying drawings.

[0014] Figure 1 A framework diagram of an intelligent express delivery batch pickup system based on user behavior analysis provided in an embodiment of the present invention; Figure 2 A flowchart illustrating the process of determining the target storage area provided in an embodiment of the present invention. Detailed Implementation

[0015] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0016] Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] This invention provides an intelligent bulk express delivery system based on user behavior analysis. See also... Figure 1 , Figure 1 This is a framework diagram of an intelligent express delivery batch pickup system based on user behavior analysis, provided as an embodiment of the present invention. It includes: The information acquisition module is used to acquire the identity information of the user picking up the package; The storage information acquisition module is used to retrieve the target storage area information corresponding to the identity information; The pickup module is used to push the location of all the packages to be picked up to the user and the package list within the target storage area based on the target storage area information. The user can then go to the target storage area to pick up all the packages.

[0018] This invention provides an intelligent bulk parcel retrieval system based on user behavior analysis. By collecting the user's identification information and retrieving associated target storage area data, the system pushes the specific storage location and parcel details of the user's parcels to the user, guiding them to the corresponding location to collect all their parcels. This method achieves precise matching of parcel storage locations, simplifies the retrieval process, avoids disordered searching by users, and improves the efficiency of bulk parcel retrieval.

[0019] In one implementation, the system, based on the determined user-specific target storage area information, first retrieves all currently stored parcels associated with that user within that area, generating a detailed location list and parcel list including specific shelf numbers, shelf positions, and the last few digits of the tracking number. Subsequently, the integrated location guidance information and parcel list are pushed to the corresponding user all at once via SMS, application notifications, or in-station displays. After receiving the notification, the user can directly go to the designated fixed area, find and pick up all their parcels at once within that area, without having to travel between multiple scattered shelves, thus achieving an efficient and accurate batch self-pickup service.

[0020] In one embodiment, the information acquisition module includes: The data acquisition module is used to collect original identity data by scanning the graphic pickup code; The identifier extraction module is used to parse the raw identity data and extract the structured user identifier field; The verification module is used to match and verify the user identifier field against the locally cached list of valid users; The identity information generation module is used to output the successfully verified identity information after the matching verification is passed; the identity information is the last five digits of the user's mobile phone number.

[0021] In one implementation, the process achieves a secure and efficient conversion from physical media to a trusted business identifier. First, scanning ensures accurate information entry. Then, structured fields are parsed and extracted to unify the data format. Next, the data is verified against the local cache to quickly verify user legitimacy and block invalid requests before operation. Finally, the last five digits of the phone number are output as a business identifier that balances privacy protection and identification efficiency. The advantage of this layered processing is that the responsibilities of each module are clear, errors are not easily propagated, local verification improves response speed and system reliability, and using a portion of the phone number balances user privacy and business needs. Overall, it provides accurate, secure, and efficient identity credentials for subsequent core businesses such as package location and storage allocation.

[0022] In one embodiment, the stored information acquisition module includes: The extraction module is used to extract the user's unique identification code from the identity information; The query module is used to query the identifier of the target storage area bound to the user based on the user's unique identification code; The location acquisition module is used to acquire the physical location parameters corresponding to the target storage area identifier; the physical location parameters include at least the shelf number and the shelf plate number; The information generation module is used to obtain target storage area information based on physical location parameters and waybill number.

[0023] In one implementation, the system first uses an extraction module to parse the user's provided identity identifier and extract the user's unique identification code used internally by the system. Then, the query module uses this identification code to quickly search the pre-established and maintained user-area binding relationship database to find the shelf location identifier assigned to the user, such as a logical code like A-area-3-2-layer. Subsequently, the location acquisition module converts this logical identifier into directly operable physical location parameters, specifying the corresponding specific shelf number and shelf number. Finally, the information generation module combines the physical location with the current express delivery tracking number to generate a complete storage location instruction containing the shelf number, shelf number, and the last digit of the tracking number, such as A3-2-8895. This instruction can directly guide staff to accurately store the package in the user's designated area or be used to send the user a pickup notification containing this detailed location information.

[0024] In one embodiment, see Figure 2 , Figure 2 A flowchart of the target storage region determination process provided in this embodiment of the invention; the target storage region determination process includes: S101, acquire multi-dimensional user data; multi-dimensional user data includes historical pickup behavior data, package characteristic data, and real-time system status data; historical pickup behavior data includes average package volume, pickup frequency, and pickup time. S102, Based on historical package pickup behavior data, users are classified into hierarchical categories to obtain hierarchical labels; S103, based on hierarchical tags and historical pickup behavior data, predict the user's pickup needs within a future preset period; S104, allocate target storage areas to users based on hierarchical labels; S105, allocates candidate target storage areas to users based on prediction results and real-time system status; S106, when the time a package stays in the target storage area exceeds a preset time threshold, the package is automatically migrated to a candidate target storage area, and the candidate target storage area is used as the updated target storage area.

[0025] In one implementation, the preset time threshold is determined by technicians, for example, 48 hours.

[0026] In one implementation, the system allocates target storage areas to users based on their hierarchical tags. First, the system maintains a storage area configuration table, which pre-defines fixed storage area specifications for each hierarchical tag. For example, high-frequency, large-item user tags are allocated continuous storage compartments with larger areas located on the lower shelves, while low-frequency user tags are allocated scattered compartments with smaller areas located on the upper shelves. For each mid-frequency level obtained through clustering, a shared area consisting of two adjacent shelves is allocated. When the system allocates an area to a user, it directly queries the configuration table to find the pre-defined area specifications corresponding to the user's hierarchical tag and instantiates the specific shelf and shelf positions from the available areas of the current site, thereby completing a precise and differentiated mapping from abstract tags to physical storage locations.

[0027] In one implementation, the system first collects multi-dimensional user data, including each user's historical pickup behavior, package characteristics, and real-time information such as the current shelf availability in the warehouse. Next, the system categorizes users based on this behavioral data, for example, labeling users who frequently pick up large packages as "high-frequency large-item users" and those who frequently pick up small packages as "low-frequency standard users," assigning each user a corresponding tier label. Then, the system combines the tier labels with historical data and uses an LSTM model to predict a user's pickup needs for the next week; for example, predicting that a high-frequency large-item user will receive two packages within three days. Based on this classification, the system allocates a fixed basic storage area to the user, such as the third shelf in shelf area A, and reserves a candidate area in the shared area, such as the second shelf in shelf area B, based on the prediction results and current shelf space. Finally, the system continuously monitors the package status. If a user's package in their fixed area remains unpicked for more than 48 hours, the system automatically moves the package to the previously reserved candidate area and notifies the user of the new pickup location, thus optimizing and efficiently utilizing storage space.

[0028] In one embodiment, classifying users into hierarchical tags based on historical package pickup behavior data includes: Users whose pickup frequency exceeds a preset first threshold and whose average package volume exceeds a preset second threshold are defined as high-frequency users, while users whose pickup frequency is lower than a third threshold are defined as low-frequency users. Extract users who are not defined as high-frequency hierarchical tags and low-frequency hierarchical tags to form a candidate user set; Construct a user feature vector set based on the historical pickup behavior data of the candidate user set; Cluster analysis of user feature vector sets yields a mid-frequency hierarchical user set; Assign a hierarchical label to each of the high-frequency, low-frequency, and mid-frequency user groups to achieve hierarchical classification of all users.

[0029] In one implementation, the preset first threshold, preset second threshold, and preset third threshold are set by technicians. The preset first threshold can be 10 times per month, the preset second threshold can be 0.05 cubic meters, and the preset third threshold can be 2 times per month.

[0030] In one implementation, the process of constructing a user feature vector set involves the system first extracting three core dimensions from the historical parcel pickup behavior data of the candidate user set: average parcel volume, pickup frequency, and pickup time. Next, the system quantifies the data for each dimension, for example, converting the average parcel volume to a value in liters, quantifying the pickup frequency to the average number of pickups per month, and parsing the pickup time into specific hourly values ​​or classifying it by time period. Then, the system combines these three processed values ​​for each user into a three-dimensional vector in a unified order. For example, a user's feature vector could be [15, 8, 19], representing an average parcel volume of approximately 15 liters, an average of 8 pickups per month, and the most frequent pickup time being 7 PM, respectively. Finally, all such three-dimensional vectors from all candidate users constitute a structured user feature vector set. This dataset concisely and centrally characterizes the differences in users' core pickup behaviors, providing standardized input for subsequent clustering analysis.

[0031] In one embodiment, clustering analysis of the user feature vector set to obtain the mid-frequency hierarchical user set includes: An optimized feature set is obtained by performing multi-dimensional correlation analysis on user feature vectors, followed by feature filtering, integration, and dimensionality reduction. Calculate the local density of each user feature point in the optimized feature set, and use user points whose local density is greater than a preset density threshold as high-density points to obtain a high-density point set. For each high-density point in the high-density point set, calculate the minimum distance between it and all high-density points, and identify high-density points whose minimum distance is greater than a preset distance threshold as density core points to obtain the density core point set. For the pickup time dimension in the user feature vector, calculate the pickup time complementarity between different users, and group users according to the pickup time complementarity to obtain a time complementarity candidate user set; An initial hierarchical set is constructed based on the density core point set and the time complementary candidate users; For each remaining feature point in the optimized feature set except for the density core point, calculate its feature similarity with the density core point in the initial level set, and assign the remaining feature points to the initial level with the highest feature similarity to obtain the initial level user set. Calculate the feature Euclidean distance between any two levels in the initial hierarchical user set, merge the two levels whose Euclidean distance is less than the preset merging threshold into a new level and assign a level identifier to obtain the mid-frequency hierarchical user set.

[0032] In one implementation, the preset density threshold and the preset distance threshold are set by a technician. The preset density threshold can be 2.5 and the preset distance threshold can be 1.8.

[0033] In one implementation, the local density calculation process first delineates the influence neighborhood for each user feature point in the optimized feature set. The radius of this neighborhood is not a fixed value but is adaptively determined based on the average distance between the point and its K nearest neighbors and the overall sparsity of the feature space. Next, the system calculates the feature similarity between the point and other points within its influence neighborhood and converts the similarity into a density contribution value of the point to other points in the neighborhood. The contribution value decays exponentially with distance. Then, for each feature point, not only are the density contributions received from all points in the neighborhood accumulated, but a first-order propagation mechanism is also introduced, whereby the density contributions of each point in the neighborhood are partially transmitted to its own neighboring points, thus forming a small-scale density diffusion network. Finally, the direct contribution and indirect propagation contribution received by each point are weighted and fused to obtain a comprehensive local density value that reflects its degree of aggregation in the local feature space. This method effectively overcomes the limitations of traditional density calculation, which is sensitive to distance and ignores the feature distribution structure. The minimum distance is calculated using Euclidean distance.

[0034] In one implementation, the pickup time dimension data is first extracted from the user feature vector, which is usually represented in time series form. Then, the complementarity of pickup times is quantified by calculating the overlap or conflict index between the time distributions of any two users. For example, if user A picks up their packages in the morning and user B picks up their packages in the evening, the overlap area of ​​their time distributions is small and the complementarity is high. Subsequently, based on the calculated complementarity matrix, the system uses a hierarchical clustering method to prioritize users with complementarity higher than a preset threshold, i.e., whose time peaks are significantly staggered, into the same group. Finally, multiple candidate user sets with complementary time coverage are formed, where the pickup times of internal users are staggered and the overall time coverage is balanced.

[0035] In one implementation, the initial hierarchy set is constructed by taking each density core point as the hierarchy center and selecting the user whose time behavior is most significantly different from that of the candidate user with complementary time to bind together to form an initial hierarchy. The set of all the hierarchy sets is the initial hierarchy set.

[0036] In one implementation, user characteristics are first screened and optimized based on dimensions such as pickup frequency, package volume, and pickup time to extract key features that accurately reflect user behavior. Next, high-density core users with representative pickup behavior are identified, and candidate users are selected based on the complementarity of pickup times. An initial user hierarchy is built around these core users. The remaining users are then assigned to the initial hierarchy with the most similar characteristics. Groups with similar characteristics are merged by calculating the similarity between hierarchies, ultimately forming a structured user group that satisfies both user characteristic similarity and allows for staggered pickup times. The resulting user hierarchy directly corresponds to pickup areas in the system, enabling more efficient batch pickups and avoiding congestion and waste of shelf resources through staggered pickup times. For example, users at the same hierarchy share adjacent shelf units.

[0037] In one embodiment, the optimized feature set obtained by performing multi-dimensional correlation analysis on user feature vectors and then integrating and reducing the dimensionality includes: The first correlation coefficient and the second correlation coefficient are obtained by calculating the correlation between the pickup frequency dimension and the pickup time dimension and the average package volume dimension and the pickup time dimension based on the user feature vector. If the first correlation coefficient is greater than the preset correlation threshold, the feature value of the pickup frequency dimension and the feature value of the pickup time dimension are weighted to obtain the first coupling feature; If the second correlation coefficient is greater than the preset correlation threshold, the feature value of the average package volume dimension and the feature value of the pickup time dimension are weighted to obtain the second coupling feature. Based on the comparison results of the first correlation coefficient and the second correlation coefficient with the preset correlation threshold, the optimized feature set is obtained by statistically analyzing all coupled features and the original feature dimensions that are judged to remain independent.

[0038] In one implementation, the preset relevant threshold is determined by technical personnel, for example, 0.6. In one implementation, the correlation coefficient is calculated as follows: First, the pickup frequency and pickup time dimensions in the user feature vector are rearranged to map discrete pickup time points and corresponding frequency values ​​into a synchronous sequence according to a time window. Then, an adaptive sliding window algorithm is used to calculate the cross-information entropy and local variance ratio of the two sequences under optimal alignment, and this composite index is normalized to the first correlation coefficient. Similarly, for the average package volume and pickup time dimensions, the system introduces a volume-time distribution matrix, estimates the kernel density of the volume data according to the time dimension, and performs a convolution operation with the pickup time distribution to extract the coupling spectral features of the two distributions. Then, the second correlation coefficient is obtained by calculating the spectral energy concentration. This method overcomes the dependence of traditional correlation coefficients on linear relationships through sequence alignment and distribution coupling analysis, and can more sensitively capture complex nonlinear correlation patterns between behavioral features.

[0039] In one implementation, the construction process of the optimized feature set is as follows: First, the first correlation coefficient and the second correlation coefficient are respectively input into a two-layer decision model. The first layer determines whether feature fusion is triggered based on a preset correlation threshold. If triggered, the corresponding coupled feature is generated and its source dimension is marked as fused. The second layer performs state tracking on all original feature dimensions, and unmarked dimensions are automatically classified as independent features. Subsequently, a temporary feature pool is constructed, and all generated coupled features and independent features are encoded by type and stored in the pool. Finally, through a round of redundancy verification, if an independent feature is highly semantically overlapping with any coupled feature, the independent feature is removed from the pool. The remaining features in the pool constitute the optimized feature set with a simplified structure and non-redundant information.

[0040] In one implementation, the process first calculates the correlation between user behavior features, such as analyzing whether the frequency of package pickup is strongly correlated with pickup within a fixed time period, or whether the size of the package is intrinsically related to a specific pickup time, thereby obtaining first and second correlation coefficients. Then, the system judges based on preset thresholds. If the correlation is strong enough, the two strongly correlated features are merged into a new coupled feature through a weighted method. For example, high frequency and evening pickup are merged into a high frequency evening activity index, or large volume and weekend pickup are merged into a large item weekend preference index. If the correlation is weak, the original independent features are retained. Finally, the system summarizes all the generated coupled features and the retained independent features to form a more concise and information-focused optimized feature set.

[0041] The foregoing has described one embodiment of the present invention in detail, but this content is merely a preferred embodiment and should not be considered as limiting the scope of the present invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the scope of the claims of this invention.

Claims

1. A smart express bulk pickup system based on user behavior analysis, characterized in that, The system includes: The information acquisition module is used to acquire the identity information of the user picking up the package; The storage information acquisition module is used to retrieve the target storage area information corresponding to the identity information; The package pickup module is used to push the location of all the packages to be picked up to the user and the package list within the target storage area based on the target storage area information, and the user can go to the target storage area to pick up all the packages.

2. The intelligent express bulk pickup system based on user behavior analysis according to claim 1, characterized in that, The information acquisition module includes: The data acquisition module is used to collect original identity data by scanning the graphic pickup code; The identifier extraction module is used to parse the original identity data and extract the structured user identifier field; The verification module is used to match and verify the user identifier field with the locally cached list of valid users; The identity information generation module is used to output the successfully verified identity information after the matching verification is passed; the identity information is the last five digits of the user's mobile phone number.

3. The intelligent express delivery batch pickup system based on user behavior analysis according to claim 1, characterized in that, The stored information acquisition module includes: The extraction module is used to extract the user's unique identification code from the identity information; The query module is used to query the target storage area identifier bound to the user based on the user's unique identification code; The location acquisition module is used to acquire the physical location parameters corresponding to the target storage area identifier; the physical location parameters include at least the shelf number and the shelf plate number. The information generation module is used to obtain target storage area information based on the physical location parameters and the waybill number.

4. The intelligent express delivery batch pickup system based on user behavior analysis according to claim 1, characterized in that, The process of determining the target storage area includes: Acquire multi-dimensional user data; the multi-dimensional user data includes historical package pickup behavior data, package feature data, and real-time system status data; the historical package pickup behavior data includes average package volume, pickup frequency, and pickup time. Users are categorized into hierarchical tags based on the historical package pickup behavior data; Based on the hierarchical tags and the historical pickup behavior data, predict the user's pickup needs in the future preset period; The target storage area is allocated to the user based on the hierarchical label; Based on the prediction results and real-time system status, candidate target storage areas are allocated to users; When the time a package stays in the target storage area exceeds a preset time threshold, the package will be automatically moved to a candidate target storage area, and the candidate target storage area will be used as the updated target storage area.

5. The intelligent express delivery batch pickup system based on user behavior analysis according to claim 4, characterized in that, Based on the historical package pickup behavior data, users are categorized into hierarchical tags, including: Users whose pickup frequency exceeds a preset first threshold and whose average package volume exceeds a preset second threshold are defined as high-frequency users, and users whose pickup frequency is lower than a third threshold are defined as low-frequency users. Extract users who are not defined as high-frequency hierarchical tags and low-frequency hierarchical tags to form a candidate user set; Construct a user feature vector set based on the historical parcel pickup behavior data of the candidate user set; Cluster analysis is performed on the user feature vector set to obtain the mid-frequency hierarchical user set; Each of the high-frequency, low-frequency, and mid-frequency user groups is assigned a corresponding hierarchical label to achieve hierarchical classification of all users.

6. The intelligent express delivery batch pickup system based on user behavior analysis according to claim 5, characterized in that, Cluster analysis of the user feature vector set yields the mid-frequency hierarchical user set, which includes: An optimized feature set is obtained by performing multi-dimensional correlation analysis on the user feature vectors, followed by feature filtering, integration, and dimensionality reduction. Calculate the local density of each user feature point in the optimized feature set, and take the user points with local density greater than a preset density threshold as high-density points to obtain a high-density point set; For each high-density point in the high-density point set, calculate the minimum distance between it and all other high-density points, and identify the high-density points whose minimum distance is greater than a preset distance threshold as density core points to obtain a density core point set. For the pickup time dimension in the user feature vector, calculate the pickup time complementarity between different users, and group the users according to the pickup time complementarity to obtain a time complementarity candidate user set; An initial hierarchical set is constructed based on the density core point set and the time complementary candidate users; For each remaining feature point in the optimized feature set except for the density core point, calculate its feature similarity with the density core point in the initial level set, and assign the remaining feature points to the initial level with the highest feature similarity to obtain the initial level user set. Calculate the feature Euclidean distance between any two levels in the initial hierarchical user set, merge the two levels whose Euclidean distance is less than a preset merging threshold into a new level and assign a level identifier to obtain the mid-frequency hierarchical user set.

7. The intelligent express delivery batch pickup system based on user behavior analysis according to claim 6, characterized in that, The optimized feature set obtained by performing multi-dimensional correlation analysis on the user feature vectors, including feature filtering, integration, and dimensionality reduction, includes: The correlation coefficients for the pickup frequency dimension and the pickup time dimension, and the correlation coefficients for the average package volume dimension and the pickup time dimension are calculated based on the user feature vector to obtain the first correlation coefficient and the second correlation coefficient. If the first correlation coefficient is greater than the preset correlation threshold, the feature value of the pickup frequency dimension and the feature value of the pickup time dimension are weighted to obtain the first coupling feature; If the second correlation coefficient is greater than the preset correlation threshold, the feature value of the average volume dimension of the package and the feature value of the pickup time dimension are weighted to obtain the second coupling feature. Based on the comparison results of the first correlation coefficient and the second correlation coefficient with the preset correlation threshold, the optimized feature set is obtained by statistically analyzing all coupled features and the original feature dimensions that are determined to remain independent.

Images