A cross-border e-commerce logistics data intelligent management method
By analyzing the category representativeness of logistics data points and data distribution characteristics in the iterative self-organizing clustering algorithm, the location of the cluster center point is optimized, and the problem of the cluster center point is not optimal in the traditional method and the classification of logistics data is improved accuracy and efficiency.
Patent Information
- Application Number
- CN202411004541.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-25
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2044-07-25
AI Technical Summary
The traditional iterative self-organizing clustering algorithm does not consider the local data distribution when correcting the cluster center point, resulting in the cluster center point being not optimal, affecting the logistics data classification results.
During the iterative self-organizing clustering process, candidate offset data points are screened by analyzing the category representativeness of logistics data points, and the cluster center point is corrected using offset probability and data distribution characteristics, and the location of cluster center point is optimized to improve classification accuracy sex.
The number of iterations is reduced, the accuracy and efficiency of logistics data classification is improved, and the accurate classification of logistics data is ensured.
Smart Images

Figure CN118551247B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of logistics data management, and in particular to a method for intelligent management of cross-border e-commerce logistics data. Background Art
[0002] The logistics data generated during the operation of cross-border e-commerce records the order time and order address of the order. For the cross-border e-commerce industry, in order to ensure the consumer experience, the order should be completed as quickly and efficiently as possible. Therefore, during the logistics transportation process, it is necessary to classify the corresponding logistics data according to the order time and order address, and then make logistics arrangements based on the classification results to improve the transportation efficiency of the order.
[0003] Due to the complexity of logistics data, in order to accurately classify it, the existing technology can use an iterative self-organizing clustering algorithm, which continuously updates the cluster center points of the initial clusters obtained during the clustering process and merges and splits the clusters, which can obtain accurate and excellent classification results. However, in the traditional iterative self-organizing clustering algorithm, the update of the cluster center point is directly calculated by the centroid for correction, and the local data distribution within the cluster is not taken into account, resulting in the modified cluster center point not being the optimal result, which in turn affects the final logistics data classification result. Summary of the invention
[0004] In order to solve the technical problem that the iterative self-organizing clustering algorithm in the prior art has a poor correction effect on the cluster center point in the cluster, thereby affecting the classification of logistics data, the purpose of the present invention is to provide a cross-border e-commerce logistics data intelligent management method, and the technical solution adopted is as follows:
[0005] The present invention proposes a cross-border e-commerce logistics data intelligent management method, the method comprising:
[0006] Construct a two-dimensional sample space according to the order time and order address of the logistics data in the cross-border e-commerce platform, and obtain logistics data points in the two-dimensional sample space;
[0007] The logistics data points are clustered using an iterative self-organizing clustering algorithm to obtain logistics data categories; the method for correcting the cluster center point in the iterative self-organizing clustering algorithm is as follows: an initial cluster is obtained in each iteration process; in an initial cluster, the category representativeness of each logistics data point is obtained according to the position of each logistics data point and the distribution characteristics of local data points, and candidate offset data points are screened out according to the category representativeness; the offset probability of each candidate offset data point is obtained according to the category representativeness of each candidate offset data point and the distance between each candidate offset data point and other candidate offset data points; the offset logistics data point corresponding to the initial cluster is screened out according to the offset probability; the direction from the cluster center point of the initial cluster to the offset logistics data point is used as the offset direction, and the offset degree is obtained according to the data distribution characteristics on the offset direction and in the opposite direction; the cluster center point is corrected according to the offset degree and the offset direction to obtain a corrected cluster.
[0008] Furthermore, the method for acquiring the two-dimensional sample space includes:
[0009] The average time interval and average address distance between each logistics data and other logistics data are obtained; and the average time interval and the average address distance are used as two dimensions in the two-dimensional sample space.
[0010] Furthermore, the method for acquiring the local data point distribution feature includes:
[0011] Select any logistics data point as the target logistics data point; obtain the first k neighborhood of the target logistics data point according to the preset k value, and use the average distance between the target logistics data point and the first neighborhood data point in the first k neighborhood as the first distribution distance;
[0012] Obtaining a second k neighborhood of the cluster center point in the initial cluster according to a preset k value, and taking the average distance between the cluster center point and the second neighborhood data points in the second k neighborhood as the second distribution distance between the cluster center point and the target logistics data;
[0013] The ratio of the first distribution distance to the second distribution distance is used as the local data point distribution feature of the target logistics data point.
[0014] Furthermore, the method for obtaining the category representativeness includes:
[0015] The distance between the target logistics data point and the cluster center point, the first distribution distance and the local data point distribution characteristics are accumulated to obtain the category deviation degree, and the category deviation degree is negatively correlated and normalized to obtain the category representativeness degree of the target logistics data point.
[0016] Furthermore, the method for obtaining the candidate offset data points includes:
[0017] The category representativeness of each logistics data point in the initial cluster is obtained, and the logistics data points in the initial cluster are divided into two data point categories according to the size of the category representativeness, and the logistics data points in the data point category with the largest category degree are used as the candidate offset data points.
[0018] Furthermore, the method for obtaining the offset probability includes:
[0019] In an initial cluster, any candidate offset data point is selected as a target candidate offset data point, and the average distance between the target candidate offset data point and other candidate offset data points is taken as a third distribution distance; the offset probability is obtained according to the category representativeness of the target candidate offset data point and the third distribution distance, and the offset probability is negatively correlated with the third distribution distance, and positively correlated with the category representativeness of the target candidate offset data point.
[0020] Furthermore, the method for obtaining the offset logistics data point includes:
[0021] In an initial cluster, a candidate offset data point with the largest offset probability is selected as the offset logistics data point.
[0022] Furthermore, the method for acquiring the data distribution characteristics includes:
[0023] Select any one of the offset direction and the direction opposite to the offset direction as the direction to be analyzed, take the cluster center as the starting point, and select a preset number of logistics data points along the direction to be analyzed as the first data point to be analyzed; construct a third k-neighborhood of the first data point to be analyzed according to the selected number, and use the logistics data points in the third k-neighborhood as the second data point to be analyzed corresponding to the first data point to be analyzed;
[0024] The average distance between the first data point to be analyzed and the corresponding second data point to be analyzed is used as the fourth distribution distance, and the cumulative sum of the fourth distribution distances of all the first data points to be analyzed is used as the data distribution feature.
[0025] Furthermore, the method for obtaining the offset degree includes:
[0026] In an initial clustering, characteristic differences of the data distribution characteristics in the offset direction and in the opposite direction are obtained, and the ratio of the number of candidate offset data points in the offset direction to the total number of candidate offset data points is used as the offset weight; the product of the characteristic difference and the offset weight is normalized to obtain an offset coefficient; the offset coefficient is multiplied by a preset adjustment parameter to obtain the offset degree.
[0027] Furthermore, the logistics data points in the initial cluster are divided into two data point categories according to the degree of category representativeness, including:
[0028] The logistics data points in the initial cluster are divided into two data point categories using the K-means clustering algorithm according to the representativeness of the categories.
[0029] The present invention has the following beneficial effects:
[0030] The embodiment of the present invention first constructs a two-dimensional sample space based on the order time and order address of the logistics data, and then obtains the logistics data points. In the process of iterative self-organizing clustering of the logistics data points, the category representativeness of each logistics data point in the initial cluster is first analyzed. The category representativeness can effectively characterize the information representativeness of each logistics data point in the initial cluster, and then provide a reference for the correction of the cluster center point. The offset logistics data points can be screened out according to the category representativeness. The direction in which the cluster center point points to the offset logistics data point is the offset direction. The data distribution in the offset direction is further used to obtain the offset degree. The cluster center point is corrected by the offset degree and the offset direction to obtain an effective correction result, which reduces the number of iterations of iterative self-organizing clustering while increasing the accuracy of the clustering results, thereby obtaining accurately classified logistics data categories. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0032] Figure 1 A flow chart of a cross-border e-commerce logistics data intelligent management method provided by one embodiment of the present invention;
[0033] Figure 2 A flow chart of a method for correcting cluster center points in an iterative self-organizing clustering process provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0034] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation method, structure, features and effects of a cross-border e-commerce logistics data intelligent management method proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.
[0035] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
[0036] The specific scheme of the cross-border e-commerce logistics data intelligent management method provided by the present invention is described in detail below with reference to the accompanying drawings.
[0037] See also Figure 1 , which shows a flow chart of a cross-border e-commerce logistics data intelligent management method provided by an embodiment of the present invention, the method comprising:
[0038] Step S101: construct a two-dimensional sample space according to the order time and order address of the logistics data in the cross-border e-commerce platform, and obtain logistics data points in the two-dimensional sample space.
[0039] In order to accurately classify logistics data, it is necessary to obtain all logistics data to be processed based on the cross-border e-commerce platform. The logistics data contains some basic data of the order, among which the data that affects logistics transportation mainly include order time and order address. Therefore, a two-dimensional sample space is constructed based on order time and order address, and the logistics data is mapped into the two-dimensional sample space to obtain all logistics data points.
[0040] Preferably, in one embodiment of the present invention, considering that the data point features in the two-dimensional sample space constructed directly using the order time and order address of the logistics data are not significant, which will affect the subsequent clustering process, the method for obtaining the two-dimensional sample space specifically includes:
[0041] Get the average time interval and average address distance between each logistics data and other logistics data. The average time interval and average address distance are used as two dimensions in the two-dimensional sample space.
[0042] The larger the average time interval, the earlier or later the order time of the corresponding logistics data is compared with other logistics data. The logistics data point of the logistics data in the two-dimensional sample space is more isolated than other logistics data points. In the clustering process, it will not be classified into the same category as other logistics data points, avoiding the wrong operation of the logistics data in the subsequent transportation management of the order. For example, when all orders are shipped uniformly, the order corresponding to the logistics data is placed later, while other orders are placed earlier. In the case of unified shipment, there is a risk of cancellation of the order or a long stocking cycle for the order. Therefore, it is necessary to distinguish the order from other orders that were placed earlier and process it separately to enhance transportation efficiency.
[0043] Similar to the average time interval, the larger the average address distance, the more the order data corresponding to the logistics data requires more logistics resources than other order data or needs to bear the risk of non-delivery, and needs to be processed separately.
[0044] Therefore, the two-dimensional sample space constructed by the average time interval and the average address distance can amplify the characteristic differences between logistics data, facilitate subsequent clustering operations, and make the final clustering effect better.
[0045] In one embodiment of the present invention, the average time interval is used as the horizontal axis in the two-dimensional sample space, and the average address distance is used as the vertical axis of the two-dimensional sample data.
[0046] It should be noted that the data in the embodiments of the present invention are normalized before and after calculation, which can avoid the dimensional influence between the data and facilitate the processing and calculation of the data.
[0047] Step S102: Cluster the logistics data points using an iterative self-organizing clustering algorithm to obtain logistics data categories.
[0048] The logistics data categories can be obtained by clustering the logistics data sheets using an iterative self-organizing clustering algorithm in a two-dimensional sample space. When managing logistics on a cross-border e-commerce platform, logistics resources can be uniformly allocated in a targeted manner based on the logistics data categories to improve logistics efficiency.
[0049] The iterative self-organizing clustering algorithm does not need to predefine the number of clusters during the clustering process. It continuously analyzes the initial clusters of each iteration, forms new clusters using merging and splitting operations, and continues to analyze to obtain the final clustering results. The iterative self-organizing clustering algorithm continuously optimizes the clustering results so that each cluster center can better represent the sample set to which it belongs. Therefore, the cluster center point correction method for forming clusters in the algorithm is an important factor affecting the final clustering results. The traditional algorithm directly uses the method of calculating the centroid to correct the cluster center point without considering the data distribution in the initial cluster. Therefore, the embodiment of the present invention optimizes the cluster center point correction method in the iterative self-organizing clustering algorithm. Please refer to Figure 2 , which shows a flow chart of a method for correcting cluster center points in an iterative self-organizing clustering process provided in one embodiment of the present invention, specifically comprising:
[0050] Step S201: obtaining an initial cluster in each iteration process; in an initial cluster, obtaining the category representativeness of each logistics data point according to the position of each logistics data point and the distribution characteristics of the local data points, and screening out candidate offset data points according to the category representativeness.
[0051] In the initial cluster formed in each iteration process, there is an original cluster center point, which needs to be corrected to a position that can truly represent the initial cluster. For a cluster, the location where the data points are most concentrated can represent the information of the cluster. Therefore, the category representativeness of each logistics data point is obtained according to the position of each logistics data point and the distribution characteristics of the local data points. The greater the category representativeness, the more the cluster center point needs to be corrected towards the corresponding logistics data point, so the candidate offset data points can be screened out according to the category representativeness.
[0052] Preferably, in one embodiment of the present invention, a method for acquiring distribution characteristics of local data points includes:
[0053] Select any logistics data point as the target logistics data point; obtain the first k neighborhood of the target logistics data point according to the preset k value, that is, select the k other logistics data points closest to the target logistics data point in the initial cluster to form the first k neighborhood, and the other logistics data points in the first k neighborhood are used as the first neighborhood data points of the target logistics data point. The average distance between the target logistics data point and the first neighborhood data point in the first k neighborhood is used as the first distribution distance. The smaller the first distribution distance, the more concentrated the neighborhood data points around the target logistics data are.
[0054] Similarly, the second k neighborhood of the cluster center point in the initial cluster is obtained according to the preset k value, and the average distance between the cluster center point and the second neighborhood data points in the second k neighborhood is used as the second distribution distance between the cluster center point and the target logistics data.
[0055] The first distribution distance and the second distribution distance can respectively characterize the degree of aggregation of the target logistics data points and the cluster center points in the local range. Therefore, the ratio of the first distribution distance to the second distribution distance is used as the local data point distribution feature of the target logistics data points. The smaller the first distribution distance and the larger the second distribution distance, the more concentrated the data points of the target logistics data points are in the local range relative to the cluster center points, and the smaller the local data point distribution feature. The local data point distribution feature can be used to analyze the relative relationship between the local aggregation degree of the target logistics data points and the cluster center points based on the idea of comparison. The embodiment of the present invention uses the local data point distribution feature to more effectively obtain the category representativeness of each logistics data point.
[0056] It should be noted that by changing the selection of the target logistics data point, the local data point distribution characteristics of each logistics data point can be obtained. The local data point distribution characteristics are expressed by the formula:
[0057] ;in, Represents the local data point distribution characteristics of the target logistics data point s, is the preset k value, represents the jth second neighborhood data point of the cluster center point s, represents the jth second neighboring data point of the target logistics data point s, In one embodiment of the present invention, k is set to 5.
[0058] It should be noted that, in other embodiments of the present invention, the first distribution distance can also be directly used as the local data point distribution feature; the number of data points within a preset neighborhood range can also be analyzed to characterize the local density of the target logistics data points to obtain the local data point distribution feature, which is not limited here.
[0059] Preferably, in one embodiment of the present invention, the method for obtaining the category representativeness comprises:
[0060] The distance between the target logistics data point and the cluster center, the first distribution distance and the distribution characteristics of the local data points are accumulated to obtain the category deviation degree. The larger the distance between the target logistics data point and the cluster center, the more the target logistics data point deviates from the original cluster center. In order to ensure the stability of the initial cluster, the position of the target logistics data point is not suitable as the corrected position of the cluster center; the larger the first distribution distance, the smaller the local density of the target logistics data point is, and it cannot effectively represent the corresponding initial cluster; the larger the distribution characteristics of the local data points, the smaller the local density of the target logistics data point is relative to the original cluster center, and the target logistics data point cannot effectively represent the corresponding initial cluster. Therefore, the greater the category deviation degree obtained by the accumulation and sum of the three characteristics, the smaller the category representation degree of the target logistics data point is. Therefore, the category deviation degree is negatively correlated and normalized to obtain the category representation degree of the target logistics data point. In one embodiment of the present invention, the category representation degree of the target logistics data point s is expressed by the formula:
[0061] ;in, Indicates the category representativeness of the target logistics data point s, is an exponential function with a natural constant as base, is the Euclidean distance between the target logistics data point s and the cluster center point, is the first distribution distance, Represents the local data point distribution characteristics of the target logistics data point s. That is, in the category representative formula, an exponential function with a natural constant as the base is used to perform negative correlation mapping and normalization. It should be noted that the negative correlation mapping and normalization methods involved in the embodiments of the present invention can be replaced by conventional mathematical means, which will not be elaborated or limited here.
[0062] By changing the selection of the target logistics data point, the category representativeness of each logistics data point can be obtained. The greater the category representativeness, the more likely the logistics data point is to be the offset reference data of the cluster center. In one embodiment of the present invention, the method for obtaining the candidate offset data point includes:
[0063] The category representativeness of each logistics data point in the initial cluster is obtained, and the logistics data points in the initial cluster are divided into two data point categories according to the size of the category representativeness, and the logistics data points in the data point category with the largest category degree are taken as candidate offset data points.
[0064] In one embodiment of the present invention, the logistics data points in the initial cluster are divided into two data point categories according to the degree of category representativeness, including:
[0065] The logistics data points in the initial cluster are divided into two data point categories using the K-means clustering algorithm according to the degree of category representativeness. That is, the K value in the K-means clustering algorithm is set to 2. It should be noted that the K-means clustering algorithm is a technical means well known to those skilled in the art, and will not be described in detail here. In other embodiments of the present invention, the Otsu threshold algorithm can also be used to achieve the classification of logistics data points according to the degree of category representativeness, which will not be described in detail here.
[0066] Step S202: Obtain the offset probability of each candidate offset data point according to the category representativeness of each candidate offset data point and the distance between each candidate offset data point and other candidate offset data points; and select the offset logistics data points corresponding to the initial cluster according to the offset probability.
[0067] The candidate offset data points obtained in step S201 are a set, which includes multiple candidate offset data points. When searching for the final offset logistics data point among the multiple candidate offset data points, the embodiment of the present invention not only takes into account the degree of category representativeness, but also further analyzes the distance between each candidate offset data point and other candidate offset data points, that is, takes into account the relative distribution of each candidate offset data point and other candidate offset data points. If a candidate offset data point is more centrally located than other candidate offset data points, it means that the candidate offset data point can better represent the overall information of the candidate offset data point set, and thus the offset probability is obtained. According to the offset probability, the offset logistics data points corresponding to the initial cluster can be screened out.
[0068] Preferably, in one embodiment of the present invention, a method for obtaining the offset probability includes:
[0069] In an initial cluster, any candidate offset data point is selected as the target candidate offset data point, and the average distance between the target candidate offset data point and other candidate offset data points is taken as the third distribution distance. The offset probability is obtained according to the category representativeness of the target candidate offset data point and the third distribution distance. The smaller the third distribution distance is, the more the target candidate offset data point is in the center of the entire candidate offset data points, and the more it can represent the candidate offset data, the larger the corresponding offset probability should be. Therefore, the offset probability is negatively correlated with the third distribution distance. The greater the category representativeness is, the more the target candidate offset data point can represent the corresponding initial cluster, that is, the category representativeness of the target candidate offset data point is positively correlated with the offset probability. In one embodiment of the present invention, the offset probability is expressed by the formula:
[0070] ;in is the displacement probability of the target candidate displacement data point h in the lth initial cluster, is the category representativeness of the target candidate offset data point h in the lth initial cluster, is the average category representativeness of all candidate offset data points in the lth initial cluster, is the third distribution distance.
[0071] In the deviation probability formula, the negative correlation between the third distribution distance and the deviation probability is constructed in the form of an inverse; As the weight of the category representativeness, the larger the item is, the greater the category representativeness of the target candidate offset data point is relative to the overall category representativeness, and the information expressed by the category representativeness should be stronger, that is, the item is used to amplify the characteristics of the category representativeness; further, the product is used to combine the items to obtain the offset probability. The larger the offset probability is, the more the corresponding target candidate offset data point is an offset logistics data point.
[0072] In one embodiment of the present invention, in an initial cluster, a candidate offset data point with the largest offset probability is selected as the offset logistics data point.
[0073] Step S203: The direction from the cluster center point of the initial cluster to the offset logistics data point is used as the offset direction, and the offset degree is obtained according to the data distribution characteristics in the offset direction and in the opposite direction; the cluster center point is corrected according to the offset degree and the offset direction to obtain a corrected cluster.
[0074] After obtaining the offset logistics data point, the direction in which the cluster center points to the offset logistics data point is the offset direction, and the cluster center needs to be offset along the offset direction. However, in the actual offset process, the cluster center cannot be directly offset to the offset logistics data point position, because the data distribution characteristics at different positions in the cluster are different, and the degree of cluster center offset should also be different. Directly offsetting to the offset logistics data point position may result in the final cluster center not being in the optimal position. Therefore, the embodiment of the present invention analyzes the data distribution characteristics in the offset direction and in the opposite direction to obtain the offset degree. The cluster center is along the offset direction, and the offset degree is used as the offset step to obtain the corrected cluster center point and obtain the corrected cluster.
[0075] Preferably, in one embodiment of the present invention, the method for acquiring data distribution characteristics includes:
[0076] Select one direction from the offset direction and the opposite direction of the offset direction as the direction to be analyzed, take the cluster center as the starting point, and select a preset number of logistics data points along the direction to be analyzed as the first data point to be analyzed. That is, the first data point to be analyzed is a neighborhood data point of the cluster center point in the direction to be analyzed.
[0077] A third k-neighborhood of the first data point to be analyzed is constructed according to the selected quantity, and the logistics data points in the third k-neighborhood serve as second data points to be analyzed corresponding to the first data point to be analyzed.
[0078] The average distance between the first data point to be analyzed and the corresponding second data point to be analyzed is taken as the fourth distribution distance, and the cumulative sum of the fourth distribution distances of all the first data points to be analyzed is taken as the data distribution feature. That is, the smaller the fourth distribution distance is, the greater the local density of the first data point to be analyzed is, and the greater the data distribution feature is, that is, the data distribution feature represents the local distribution density of the local area in the corresponding direction.
[0079] Preferably, in one embodiment of the present invention, a method for obtaining the degree of deviation includes:
[0080] In an initial cluster, the characteristic difference of the data distribution characteristics in the offset direction and the reverse direction is obtained. The greater the characteristic difference, the worse the uniformity of the data point distribution, the worse the position of the current cluster center point, and the greater the degree of correction required.
[0081] The ratio of the number of candidate offset data points in the offset direction to the total number of candidate offset data points is used as the offset weight. The larger the offset weight, the more concentrated the candidate offset data points are in the offset direction, and the more the cluster center needs to be corrected. Therefore, the product of the feature difference and the offset weight is normalized to obtain the offset coefficient, and the offset coefficient is multiplied by the preset adjustment parameter to obtain the offset degree. In one embodiment of the present invention, the offset degree is expressed by the formula:
[0082] ;in Indicates the degree of deviation of the cluster center point in the lth initial cluster, is the normalization function, is the data distribution characteristic in the offset direction, is the data distribution characteristic in the direction opposite to the offset direction, is the number of candidate offset data points in the lth initial cluster in the offset direction, is the total number of candidate offset data points in the lth initial cluster, To adjust the parameters.
[0083] In one embodiment of the present invention, the adjustment parameter is set to 0.2, and the normalization function adopts range normalization.
[0084] At this point, the cluster center point correction of the initial cluster is completed in the iterative process, and the final cluster can be obtained until the iterative self-organizing clustering algorithm ends. The logistics data in each final cluster belongs to the same logistics data category. The order time and order address of the orders involved in a logistics data category are similar, so the cross-border e-commerce platform can process the orders in each logistics data category in a targeted manner, and the logistics data belonging to the same category can be uniformly supervised and managed, reducing the chaos of logistics transportation and improving logistics efficiency.
[0085] In summary, the embodiment of the present invention constructs a two-dimensional sample space based on the order time and order address of the logistics data, and then obtains the logistics data points. In the process of iterative self-organizing clustering of the logistics data points, the category representativeness of each logistics data point in the initial cluster is first analyzed, and the offset logistics data points can be screened out according to the category representativeness. The direction in which the cluster center points to the offset logistics data point is the offset direction. The offset degree is obtained by using the data distribution in the offset direction, and the cluster center is corrected by the offset degree and the offset direction to obtain the logistics data category. The embodiment of the present invention obtains the logistics data category with good classification effect by optimizing the iterative self-organizing clustering algorithm in the logistics data classification process, thereby realizing the effective management of cross-border e-commerce logistics data.
[0086] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0087] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.
Claims
1. A cross-border e-commerce logistics data intelligent management method, characterized in that: The method comprises: Construct a two-dimensional sample space according to the order time and order address of the logistics data in the cross-border e-commerce platform, and obtain logistics data points in the two-dimensional sample space; The logistics data points are clustered using an iterative self-organizing clustering algorithm to obtain logistics data categories; the method for correcting the cluster center point in the iterative self-organizing clustering algorithm is as follows: an initial cluster is obtained in each iteration process; in an initial cluster, the category representativeness of each logistics data point is obtained according to the position of each logistics data point and the distribution characteristics of local data points, and candidate offset data points are screened out according to the category representativeness; the offset probability of each candidate offset data point is obtained according to the category representativeness of each candidate offset data point and the distance between each candidate offset data point and other candidate offset data points; the offset logistics data point corresponding to the initial cluster is screened out according to the offset probability; the direction from the cluster center point of the initial cluster to the offset logistics data point is taken as the offset direction, and the offset degree is obtained according to the data distribution characteristics in the offset direction and in the opposite direction; the cluster center point is corrected according to the offset degree and the offset direction to obtain a corrected cluster; The method for acquiring the local data point distribution characteristics includes: Select any logistics data point as the target logistics data point; obtain the first k neighborhood of the target logistics data point according to the preset k value, and use the average distance between the target logistics data point and the first neighborhood data point in the first k neighborhood as the first distribution distance; Obtaining a second k neighborhood of the cluster center point in the initial cluster according to a preset k value, and taking the average distance between the cluster center point and the second neighborhood data points in the second k neighborhood as the second distribution distance between the cluster center point and the target logistics data; Using the ratio of the first distribution distance to the second distribution distance as the local data point distribution feature of the target logistics data point; The method for obtaining the category representative degree includes: The distance between the target logistics data point and the cluster center point, the first distribution distance and the local data point distribution characteristics are accumulated to obtain the category deviation degree, and the category deviation degree is negatively correlated and normalized to obtain the category representativeness degree of the target logistics data point.
2. According to claim 1, a cross-border e-commerce logistics data intelligent management method is characterized in that: The method for acquiring the two-dimensional sample space includes: The average time interval and average address distance between each logistics data and other logistics data are obtained; and the average time interval and the average address distance are used as two dimensions in the two-dimensional sample space.
3. According to claim 1, a cross-border e-commerce logistics data intelligent management method is characterized in that: The method for obtaining the candidate offset data points includes: The category representativeness of each logistics data point in the initial cluster is obtained, and the logistics data points in the initial cluster are divided into two data point categories according to the size of the category representativeness, and the logistics data points in the data point category with the largest category degree are used as the candidate offset data points.
4. According to claim 1, a cross-border e-commerce logistics data intelligent management method is characterized in that: The method for obtaining the offset probability includes: In an initial cluster, any candidate offset data point is selected as a target candidate offset data point, and the average distance between the target candidate offset data point and other candidate offset data points is taken as a third distribution distance; the offset probability is obtained according to the category representativeness of the target candidate offset data point and the third distribution distance, and the offset probability is negatively correlated with the third distribution distance, and positively correlated with the category representativeness of the target candidate offset data point.
5. According to claim 1, a cross-border e-commerce logistics data intelligent management method is characterized in that: The method for obtaining the offset logistics data point includes: In an initial cluster, a candidate offset data point with the largest offset probability is selected as the offset logistics data point.
6. A cross-border e-commerce logistics data intelligent management method according to claim 1, characterized in that: The method for acquiring the data distribution characteristics includes: Select any one of the offset direction and the direction opposite to the offset direction as the direction to be analyzed, take the cluster center as the starting point, and select a preset number of logistics data points along the direction to be analyzed as the first data point to be analyzed; construct a third k-neighborhood of the first data point to be analyzed according to the selected number, and use the logistics data points in the third k-neighborhood as the second data point to be analyzed corresponding to the first data point to be analyzed; The average distance between the first data point to be analyzed and the corresponding second data point to be analyzed is used as the fourth distribution distance, and the cumulative sum of the fourth distribution distances of all the first data points to be analyzed is used as the data distribution feature.
7. The method for intelligent management of cross-border e-commerce logistics data according to claim 1, characterized in that: The method for obtaining the offset degree includes: In an initial clustering, characteristic differences of the data distribution characteristics in the offset direction and in the opposite direction are obtained, and the ratio of the number of candidate offset data points in the offset direction to the total number of candidate offset data points is used as the offset weight; the product of the characteristic difference and the offset weight is normalized to obtain an offset coefficient; the offset coefficient is multiplied by a preset adjustment parameter to obtain the offset degree.
8. The method for intelligent management of cross-border e-commerce logistics data according to claim 3 is characterized in that: The method of dividing the logistics data points in the initial cluster into two data point categories according to the degree of category representativeness includes: The logistics data points in the initial cluster are divided into two data point categories using the K-means clustering algorithm according to the representativeness of the categories.
Citation Information
Cited By
Cross-border e-commerce visual report generation method and system based on drag-and-drop component
CN122018767A
A Method and System for Generating Visual Reports for Cross-Border E-commerce Based on Drag-and-Drop Components
CN122018767B
