Offline grouping and online analysis method for communication scenario electronic business card

Abstract

The application provides a customer offline grouping and online analysis method for an electric name card in a communication scenario, comprising: extracting sample communication bill data and product application log data and performing feature processing to obtain standardized data; performing maximum and minimum processing on the standardized data to construct corresponding standard RANS vectors; selecting N standard RANS vectors and performing clustering by using a kmeans++ algorithm conforming to data distribution to obtain k clustering centroids; determining k customer grouping labels; and establishing a mapping relationship between the k clustering centroids and the k customer grouping labels. The offline clustering and grouping stage proposes a feature selection method with strong business interpretation and a label processing method, determines the k clustering centroids and the corresponding clusters, and the customer grouping labels, and after the mapping relationship is established, the customer type can be determined through the customer grouping labels, which is beneficial to subsequent feedback of corresponding product strategies based on the customer grouping labels, improves the call connection rate, and reduces the use cost of the electric name card.

Description

Technical Field

[0001] This invention relates to the field of intelligent calling, and more specifically, to a method for offline customer segmentation and online analysis of electronic business cards in communication scenarios. Background Technology

[0002] Caller ID cards are a unique communication product that displays the reason for an incoming call on a mobile phone. Mobile phone users selectively answer calls based on the caller ID displayed on the call interface. Enterprise users can choose to send caller ID cards to all users, but for scenarios with tens of millions of calls per day, the cost of caller ID cards is becoming increasingly significant, impacting their revenue. Therefore, a strategy of intelligently sending caller ID cards to mobile phone users can improve call connection rates (the probability of the called party answering the call) and reduce costs. Existing caller ID cards not only lack precise selection of feature engineering variables but also fail to consider the characteristics of changing customer behavior, making it impossible to provide accurate sending strategies for specific customers. They can only adopt full or random sending methods, increasing costs for enterprises, especially logistics companies with tens of millions of calls per day. Summary of the Invention

[0003] The present invention aims to overcome at least one of the defects of the prior art and provides a method for offline customer segmentation and online analysis of electronic business cards in communication scenarios, for solving the problem.

[0004] The technical solution adopted in this invention includes:

[0005] In a first aspect, the present invention provides a method for offline customer segmentation using electronic business cards in communication scenarios, comprising: extracting sample communication call detail records (CDRs) and product application log data from the current period to a historical period T; performing feature processing on the sample CDRs and application log data to obtain standardized data; performing maximum-minimum processing on the standardized data to construct corresponding standard RANS vectors; selecting N standard RANS vectors and using the kmeans++ algorithm conforming to the data distribution to perform clustering to obtain k cluster centroids; determining k customer segmentation labels based on the k cluster centroids; and establishing a mapping relationship between the k cluster centroids and the k customer segmentation labels.

[0006] The present invention provides a method for offline customer segmentation using business cards in communication scenarios. Based on a large amount of sample communication call detail records and product application log data, it obtains k cluster centroids and their corresponding clusters through a clustering algorithm. Each cluster can correspond to a customer segmentation label. After establishing the mapping relationship, the customer type can be quickly determined through the customer segmentation label, which is conducive to subsequent feedback of corresponding product strategies (such as the specific sending method of business cards) based on the customer segmentation label, thereby improving the call connection rate and reducing the usage cost of business cards.

[0007] Furthermore, the standardized data is subjected to minimization processing to construct the corresponding standard RANS vector. Specifically, the standardized data includes four variables: r, n, a, and s; according to... For each standardized data point, a minimization process is performed; where r', n', a', and s' are the four variables of the constructed standard RANS vector, r, n, a, and s are the four variables in the standardized data, minR, minN, minA, and minS are the minimum values ​​of the four variables in the standardized data, and maxR, maxN, maxA, and maxS are the maximum values ​​of the four variables in the standardized data.

[0008] The standardized data includes four variables: r, n, a, and s, representing the response rate, last call time, average call duration within the statistical period, and product reach percentage, respectively. After performing a minimization process on each standardized data point, a standard RANS vector can be constructed, enabling more accurate cluster analysis.

[0009] Furthermore, N standard RANS vectors are selected, and clustering is performed using the kmeans++ algorithm that conforms to the data distribution. Specifically, when the number of selected standard RANS vectors is N, the number of cluster trees is k, and the maximum number of iterations is m, the following steps are performed for clustering: The set of input standard RANS vectors is X = {x1, x2, x3, ... x...} n The output clustering result is C = {C1, C2, C3, ... C}. n Step a: Randomly select k standard RANS vectors as k temporary cluster centers μ. j Step b: For each x in the input set i ,according to Calculate x i With temporary cluster centers μ j Distance D(x) i Step c: Select x corresponding to the larger D(x) value as a cluster centroid; Step d: Repeat steps b and c until k cluster centroids μ are selected. j Step e: Use the determined k cluster centroids as initial centroids μ j ; after determining k initial centroids μ j Then, for iteration numbers m = 1, 2, 3, ... according to dist(x) i μ j )=1-cos(x i μ j ) Calculate each standard RANS vector x i To each initial centroid μ j cosine distance dist(x)i μ j ), determine each x i The shortest cosine distance dist(x) i μ j The corresponding initial centroid μ j , will x i The initial centroid μ is marked as determined. j The corresponding cluster; output the clustering results.

[0010] Based on the distance between standard RANS vectors, k cluster centroids can be obtained after multiple calculations. The standard RANS vectors are then classified into the clusters corresponding to each centroid, thus completing the cluster analysis.

[0011] Furthermore, k customer segmentation labels are determined based on the k cluster centroids, specifically including: determining the mean r of the four variables corresponding to the selected N standard RANS vectors. avg n avg a avg s avg The mean r of the four variables avg n avg a avg s avg The values ​​of the four variables corresponding to the k cluster centroids are compared, and the k customer group labels of the k cluster centroids are determined based on the comparison results.

[0012] The mean of the four variables corresponding to the standard RANS vector is used as the threshold for judging customer group labels. The values ​​of the four variables corresponding to the k cluster centroids are compared with the mean to determine the k customer group labels.

[0013] Furthermore, the value of k is 7; based on the comparison results, k customer segmentation labels for the k cluster centroids are determined, specifically including: if R k >r avg N k >n avg A k >a avg S k >s avg The customer segment labels corresponding to the cluster centroids that meet the conditions are identified as high-response customer groups in the long term; if R k >r avg N k >n avg A k >a avg S k avg The customer segmentation labels corresponding to the cluster centroids that meet the conditions are identified as high-value customers in the long term; if R k <r avg ​N k <n avg A k avg S k >s avg The customer segment labels corresponding to the cluster centroids that meet the conditions are identified as customers with low response rates in the near future; if R k <r avg N k >n avg A k avg S k >s avg The customer segment labels corresponding to the cluster centroids that meet the conditions are identified as low-response customers in the long term; if R k >r avg N k <n avg A k >a avg S k >s avg The customer segment labels corresponding to the cluster centroids that meet the conditions are identified as high-response customers in the near future; if R k <r avg N k >n avg A k avg S k avg The customer segment labels corresponding to the cluster centroids that meet the conditions are identified as potential customers; if R k >r avg N k <n avg A k >a avg S k avg The customer segmentation labels corresponding to the cluster centroids that meet the conditions are identified as high-value customers in the near future; among them, R k N k A k S k The data consists of four variables representing the centroids of the cluster.

[0014] Once customer segmentation tags are identified, these tags can be directly used to determine specific product strategies for responding to customers, increasing the probability of called users answering calls while maintaining low business costs.

[0015] ​​​​​Secondly, this invention provides an online customer segmentation analysis method for electronic business cards in communication scenarios, comprising: determining whether a target customer is a historical customer or a new customer; when the target customer is determined to be a historical customer, obtaining customer segmentation labels already assigned to historical customers, and providing feedback on the product strategy corresponding to the assigned customer segmentation labels; when the target customer is determined to be a new customer, obtaining the standard RANS vector corresponding to the new customer, determining the distance between the standard RANS vector of the new customer and each of the k cluster centroids determined by the aforementioned customer segmentation method for electronic business cards in communication scenarios, and determining the cluster centroid with the smallest distance; determining the standard RANS vector of the new customer. The distance between the vector and other vectors within the cluster corresponding to the initial clustering result is used to determine whether offline clustering needs to be restarted. If so, the above steps of clustering using the kmeans++ algorithm that conforms to the data distribution are re-executed based on the standard RANS vector of the new customer, new cluster centroids are republished, and the customer segment label corresponding to the cluster centroid with the smallest distance to the standard RANS vector of the new customer is determined. The product strategy corresponding to the determined customer segment label is then applied. If not, the customer segment label corresponding to the cluster centroid with the smallest distance is determined, and the product strategy corresponding to the determined customer segment label is applied.

[0016] This invention also provides an online analysis method for customer segmentation of business cards in communication scenarios. Based on customer segmentation labels determined by offline segmentation methods, specific product strategies for customer feedback are determined. Specifically, it is necessary to first distinguish between historical customers and new customers. If they are historical customers, the previously determined product strategies can be directly fed back. If they are new customers, their data needs to be classified into the standard RANS vectors that have been constructed and clustered. However, there is a possibility that the standard RANS vectors do not belong to any of the pre-determined clusters. Therefore, it may be necessary to restart the offline clustering process, reissue new cluster centroids, and determine new customer segmentation labels to determine a more accurate and effective way to process business cards.

[0017] Furthermore, when the target customer is determined to be a new customer, before obtaining the standard RANS vector corresponding to the new customer, the process includes: determining whether the observation period time window of the new customer meets the time window threshold; if not, the new customer is set as the target user to be observed, and the process is repeated until the new customer is a user in the observation period; if so, the standard RANS vector within the observation period time window is determined as the standard RANS vector corresponding to the new customer, and subsequent steps are executed.

[0018] Furthermore, the distance between the standard RANS vector of the new customer and other vectors within the cluster corresponding to the initial clustering result is determined. Based on the determined distance, it is determined whether offline clustering needs to be restarted. Specifically, this includes: based on... Determine whether offline clustering needs to be restarted; where s(i) is the evaluation metric, and Dcii D is the average distance from the standard RANS vector of a new customer to other vectors within the corresponding cluster. cij s(i) represents the average distance from the standard RANS vector of a new customer to vectors within other clusters. If the value of s(i) is close to 1, then it is determined that offline clustering does not need to be restarted. If the value of s(i) is close to -1 or 0, then it is determined whether the number of new customers whose s(i) values ​​are close to -1 or 0 has reached the restart threshold. If yes, it is determined that offline clustering needs to be restarted; otherwise, it is determined that offline clustering does not need to be restarted.

[0019] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-mentioned offline customer segmentation method for electronic business cards in communication scenarios, and / or the above-mentioned online customer segmentation analysis method for electronic business cards in communication scenarios.

[0020] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the above-mentioned offline customer segmentation method for electronic business cards in communication scenarios, and / or the above-mentioned online customer segmentation analysis method for electronic business cards in communication scenarios.

[0021] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0022] This invention provides a method for offline customer segmentation and online analysis of business cards in communication scenarios. First, standardized data is obtained based on sample communication call detail records and product application log data within a certain period. Then, a clustering algorithm is used to cluster the standardized data, resulting in k cluster centroids and their corresponding clusters. Each cluster is divided into k customer segmentation labels based on the variable data of the cluster centroid and the mean. A mapping relationship is established between the cluster centroids and the customer segmentation labels. The offline clustering stage of this invention proposes a feature selection method and label processing method with strong business interpretability, selecting initial centroids based on the principle of maximizing probability. In the online analysis method, new users can be quickly identified by customer segmentation labels based on the mapping relationship, thereby facilitating rapid feedback of corresponding product strategies (such as the specific method of sending business cards), improving call connection rates, and reducing the cost of using business cards. When a new user does not belong to any pre-determined cluster, the offline clustering step is re-triggered under certain conditions to obtain new cluster centroids. New customer segmentation labels are defined for the new cluster centroids, and corresponding product strategies are determined. The system dynamically assesses the rationality of classification and determines whether clustering needs to be restarted, accurately applying different business strategies to different customer groups. This overcomes the difficulty of sending electronic business cards without a target and reduces the company's operating costs. Attached Figure Description

[0023] Figure 1This is a schematic diagram of the overall steps in Embodiment 1 of the present invention.

[0024] Figure 2 This is a flowchart illustrating steps S110 to S160 of Embodiment 1 of the present invention.

[0025] Figure 3 This is a schematic diagram of the overall steps in Embodiment 2 of the present invention.

[0026] Figure 4 This is a flowchart illustrating steps S210 to S270 of Embodiment 2 of the present invention.

[0027] Figure 5 This is a flowchart illustrating steps T210 to T230 of Embodiment 2 of the present invention. Detailed Implementation

[0028] The accompanying drawings are for illustrative purposes only and should not be construed as limiting the invention. To better illustrate the following embodiments, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions; it is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0029] Example 1

[0030] like Figure 1 , 2 As shown in the figure, this embodiment provides a method for offline customer segmentation using electronic business cards in a communication scenario, including:

[0031] S110. Extract sample communication call detail records and product application log data from the current period to the historical period T.

[0032] Specifically, this step extracts a certain amount (e.g., tens of millions) of sample communication call detail records from the call topic domain of the data warehouse from the current period to the previous statistical period T, extracts product application log data from the business system product logs, and sets the historical sample data for the statistical period T.

[0033] S120. Perform feature processing on the sample communication call detail records and application log data to obtain standardized data;

[0034] Specifically, this step performs ETL feature processing on the sample communication call detail records and application log data to obtain processable standardized data.

[0035] S130. Perform maximum-minimum processing on the standardized data to construct the corresponding standard RANS vector;

[0036] Standardized data includes four variables: r, n, a, and s.

[0037] r represents the response rate;

[0038] n is near_time, the time of the last call;

[0039] 'a' represents avg_duration, the average call duration within the statistical period.

[0040] s represents send_rate, the percentage of product reach.

[0041] We construct standard RANS vectors by performing a maximum-minimum operation on the four variables.

[0042] according to Perform maximization and minimization on each standardized data point.

[0043] Where r', n', a', and s' are the four variables of the constructed standard RANS vector, r, n, a, and s are the four variables of the standardized data, minR, minN, minA, and minS are the minimum values ​​of the four variables of the standardized data, and maxR, maxN, maxA, and maxS are the maximum values ​​of the four variables of the standardized data.

[0044] S140. Select N standard RANS vectors and use the kmeans++ algorithm that conforms to the data distribution to perform clustering, and obtain k cluster centroids;

[0045] Specifically, in this step, when the number of selected standard RANS vectors is N, the number of cluster trees is k, and the maximum number of iterations is m, the following steps are performed for clustering:

[0046] The set of input standard RANS vectors is X = {x1, x2, x3, ... x}. n The output clustering result is C = {C1, C2, C3, ... C}. n}

[0047] Step a: Randomly select k standard RANS vectors as k temporary cluster centers μ j ;

[0048] Step b: For each x in the input set i ,according to Calculate x i With temporary cluster centers μ j Distance D(x) i );

[0049] Step c: Select x corresponding to the larger D(x) value as a cluster centroid;

[0050] A larger D(x) value indicates that the center has a larger radiation radius and can be used as the centroid of the cluster.

[0051] Step d: Repeat steps b and c until k cluster centroids μ are selected. j ;

[0052] Step e: Use the determined k cluster centroids as initial centroids μ j ;

[0053] After determining k initial centroids μ j Then, for iteration numbers m = 1, 2, 3, ...

[0054] According to dist(x) i μ j )=1-cos(x i )μ j ) Calculate each standard RANS vector x i To each initial centroid μ j cosine distance dist(x) i μ j ), determine each x i The shortest cosine distance dist(x) i μ j The corresponding initial centroid μ j , will x i The initial centroid μ is marked as determined. j The corresponding cluster;

[0055] Output the clustering results.

[0056] S150. Determine k customer segmentation labels based on k cluster centroids;

[0057] Specifically, this step first determines the mean r of the four variables corresponding to the selected N standard RANS vectors. avg n avg a avg s avg .

[0058] The mean r of the four variables avg n avg a avg s avg The values ​​of the four variables corresponding to the k cluster centroids are compared, and the k customer group labels of the k cluster centroids are determined based on the comparison results.

[0059] In this specific implementation, the value of k is 7.

[0060] If R k >r avg N k>n avg A k >a avg S k >s avg The customer segmentation labels corresponding to the cluster centroids that meet the criteria are identified as high-response customer groups in the long term. These customer groups are characterized by frequent product use in the long term, higher response rates, and longer average call durations.

[0061] If R k >r avg N k >n avg A k >a avg S k avg The customer segmentation labels corresponding to the cluster centroids that meet the criteria are identified as high-value customers in the long term. This customer group is characterized by a high response rate and long average call duration in the long term, but a low product application rate. This type of customer group has a high connection rate even without product application notifications.

[0062] If R k <r avg N k <n avg A k avg S k >s avg The customer segment labels corresponding to the cluster centroids that meet the criteria are identified as low-response customers in the near term. This customer group is characterized by a low recent response rate and short average call duration, but they use the product extensively, indicating that this type of customer group has a low response rate to product applications.

[0063] If R k <r avg N k >n avg A k avg S k >s avg The customer segment labels corresponding to the cluster centroids that meet the criteria are identified as low-response customers in the long term. This customer group is characterized by a low response rate and short average call duration in the long term, but they use the product extensively, indicating that this type of customer group has a low response rate to product applications.

[0064] If R k >r avg N k <n avg A k >a avg S k >s avg ​​​The customer segment labels corresponding to the cluster centroids that meet the criteria are identified as high-response customers in the near future. This customer group is characterized by a high recent response rate, high average call duration, and multiple product applications.

[0065] If R k <r avg N k >n avg A k avg S k avg The customer segment labels corresponding to the cluster centroids that meet the criteria are identified as potential customers. This customer group is characterized by a low response rate in the long term, low average call duration, and limited product usage. These customers are far from the date of the call and have not used the product before.

[0066] If R k >r avg N k <n avg A k >a avg S k avg The customer segmentation labels corresponding to the cluster centroids that meet the criteria are identified as high-value customers in the near future. This customer group is characterized by a high recent response rate, long average call duration, and limited product usage.

[0067] Among them, R k N k A k S k The data consists of four variables representing the centroids of the cluster.

[0068] S160. Establish the mapping relationship between k cluster centroids and k customer group labels.

[0069] Example 2

[0070] like Figure 3 , 4 As shown, this embodiment provides a method for online customer segmentation analysis of electronic business cards in communication scenarios, including:

[0071] S210. Determine whether the target customer is a historical customer; if the target customer is determined to be a historical customer, proceed to step S220; if the target customer is determined not to be a historical customer, proceed to step S230.

[0072] S220. Obtain customer segmentation tags that have been assigned to historical customers, provide feedback on the product strategies corresponding to the assigned customer segmentation tags, and end the execution of the method.

[0073] ​​​Specifically, this step retrieves the cluster tags from the cluster tag mapping table stored in high-performance middleware such as Redis, and returns the product strategy based on the cluster tags.

[0074] S230. Determine the target customer as a new customer, obtain the standard RANS vector corresponding to the new customer, determine the distance between the standard RANS vector of the new customer and each of the k cluster centroids determined by the method provided in Example 1, and determine the cluster centroid with the smallest distance.

[0075] Specifically, before performing this step, such as Figure 3 , 5 As shown, it also includes the following execution steps:

[0076] T210. Determine whether the observation period window for the new customer meets the time window threshold; if not, proceed to step T220; if yes, proceed to step T230.

[0077] T220. Set the new customer as a target user to be observed, and re-evaluate whether the new customer is a user in the observation period until the new customer is a user in the observation period.

[0078] T230. Determine the standard RANS vector within the observation period's time window as the standard RANS vector corresponding to the new customer, and execute step S230.

[0079] S240. Determine the distance between the standard RANS vector of the new customer and the other vectors within the cluster corresponding to the initial clustering result;

[0080] S250. Determine whether offline clustering needs to be restarted based on the determined distance. If yes, proceed to step S260; otherwise, proceed to step S270.

[0081] Specifically, in this step, according to Determine whether offline clustering needs to be restarted. Offline clustering refers to re-executing step S140 in Example 1.

[0082] Where s(i) is the evaluation index, D cii D is the average distance from the standard RANS vector of a new customer to other vectors within the corresponding cluster. cij The average distance from the standard RANS vector of a new customer to vectors within other clusters.

[0083] D cii The smaller the value, the more likely it is that vector x has been added. i It is close to the centroid of the cluster.

[0084] D cij The larger the value, the more likely it is that vector x has been added. i The centroid of the cluster is far away from other clusters.

[0085] If the value of s(i) is close to 1, then the vector is classified reasonably, and it is determined that there is no need to restart offline clustering;

[0086] If the value of s(i) is close to -1, the vector is classified into other classes, and it is determined that offline clustering needs to be restarted.

[0087] If the value of s(i) is close to 0, the vector classification belongs to the boundary between clusters, and it is determined that offline clustering needs to be restarted;

[0088] S260. Based on the standard RANS vector of the new customer, re-execute step S140 of Example 1, republish the new cluster centroid, determine the customer segment label corresponding to the cluster centroid with the smallest distance to the standard RANS vector of the new customer, and formulate the product strategy corresponding to the determined customer segment label.

[0089] S270. Determine the customer segmentation label corresponding to the cluster centroid with the smallest distance, and formulate a product strategy based on the determined customer segmentation label.

[0090] Determine if the number of new customers whose value of s(i) is close to -1 or 0 has reached the restart threshold. If yes, determine that offline clustering needs to be restarted; otherwise, determine that offline clustering does not need to be restarted.

[0091] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the technical solution of the present invention, and are not intended to limit the specific implementation of the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the claims of the present invention should be included within the protection scope of the claims of the present invention.

Claims

1. A method for offline customer segmentation using electronic business cards in a communication scenario, characterized in that, include: Extract sample communication call detail records and product application log data from the current period to the historical period T; The sample communication call detail records and application log data are subjected to feature processing to obtain standardized data; The standardized data is processed by minimization to construct the corresponding standard RANS vectors; Specifically, standardized data includes four variables: r, n, a, and s. r represents the response rate; n is near_time, the time of the last call; 'a' represents avg_duration, the average call duration within the statistical period. s represents send_rate, the percentage of products reached; Max-minimization is performed on the four variables to construct a standard RANS vector; N standard RANS vectors are selected and clustered using the kmeans++ algorithm that conforms to the data distribution to obtain k cluster centroids; N standard RANS vectors are selected, and clustering is performed using the kmeans++ algorithm that conforms to the data distribution. Specifically, this includes: When the number of selected standard RANS vectors is N, the number of cluster trees is k, and the maximum number of iterations is m, the following steps are performed for clustering: The set of input standard RANS vectors is X ={ x 1 ,x 2 ,x 3 ,…x k The output clustering result is C={C1,C2,C3,…C}. k }; Step a: Randomly select k standard RANS vectors as k temporary cluster centers. j ; Step b: For each element in the input set ,according to calculate With temporary cluster center j Distance D ( ); Step c: Select D ( Larger values ​​correspond to x It can serve as a cluster centroid; Step d: Repeat steps b and c until k cluster centroids are selected. j ; Step e: Use the determined k cluster centroids as initial centroids. j ; Determine k initial centroids j Then, for iteration numbers m=1, 2, 3, ... according to Calculate each standard RANS vector x i To each initial centroid j cosine distance dist( , j ), determine each x i The shortest cosine distance dist( , j The corresponding initial centroid j ,Will x i Marked as the determined initial centroid j The corresponding cluster; Output the clustering results; Determine k customer segmentation labels based on k cluster centroids; Establish a mapping relationship between k cluster centroids and k customer group labels.

2. The customer offline grouping method for electronic business cards in communication scenarios according to claim 1, characterized in that, The standardized data is processed by minimization to construct the corresponding standard RANS vectors, specifically including: Standardized data includes four variables: r, n, a, and s. according to , , , Perform maximization and minimization on each standardized data point; Where r', n', a', and s' are the four variables of the constructed standard RANS vector, r, n, a, and s are the four variables of the standardized data, minR, minN, minA, and minS are the minimum values ​​of the four variables of the standardized data, and maxR, maxN, maxA, and maxS are the maximum values ​​of the four variables of the standardized data.

3. The customer offline grouping method for electronic business cards in communication scenarios according to claim 1, characterized in that, Based on the k cluster centroids, k customer segmentation labels are determined, specifically including: Determine the mean r of the four variables corresponding to the selected N standard RANS vectors. avg n avg a avg s avg ; The mean r of the four variables avg n avg a avg s avg The values ​​of the four variables corresponding to the k cluster centroids are compared, and the k customer group labels of the k cluster centroids are determined based on the comparison results.

4. The customer offline grouping method for electronic business cards in communication scenarios according to claim 3, characterized in that, The value of k is 7; Based on the comparison results, k customer segmentation labels are determined for the k cluster centroids, specifically including: If R k >r avg N k >n avg A k >a avg S k >s avg The customer segmentation labels corresponding to the cluster centroids that meet the conditions are identified as high-response customer groups in the long term; If R k >r avg N k >n avg A k >a avg S k avg The customer group labels corresponding to the cluster centroids that meet the conditions are identified as high-value customers in the long term.​ If R k <r avg N k <n avg A k avg S k >s avg The customer group labels corresponding to the cluster centroids that meet the conditions are identified as customers with low response in the near future.​ If R k <r avg N k >n avg A k avg S k >s avg The customer group labels corresponding to the cluster centroids that meet the conditions are identified as low-response customers in the long term.​ If R k >r avg N k <n avg A k >a avg S k >s avg The customer segmentation labels corresponding to the cluster centroids that meet the conditions are identified as high-response customers in the near future. If R k <r avg N k >n avg A k avg S k avg The customer group labels corresponding to the cluster centroids that meet the conditions are identified as potential customers.​​ If R k >r avg N k <n avg A k >a avg S k avg The customer segmentation labels corresponding to the cluster centroids that meet the conditions are identified as high-value customers in the near future.​ Among them, R k N k A k S k The data consists of four variables representing the centroids of the cluster.

5. A method for online customer segmentation analysis of electronic business cards in communication scenarios, characterized in that, include: Determine whether the target customers are existing customers or new customers; When the target customer is identified as a historical customer, obtain the customer segmentation tags that have been assigned to the historical customer and provide feedback on the product strategy corresponding to the assigned customer segmentation tags. When the target customer is determined to be a new customer, the standard RANS vector corresponding to the new customer is obtained, the distance between the standard RANS vector of the new customer and the k cluster centroids determined by the customer grouping method of the electronic display business card in the communication scenario as described in any one of claims 1 to 4 is determined, and the cluster centroid with the smallest distance is determined. Determine the distance between the standard RANS vector of the new customer and other vectors within the cluster corresponding to the initial clustering result. Based on the determined distance, determine whether offline clustering needs to be restarted. If so, re-execute the steps of clustering using the kmeans++ algorithm conforming to the data distribution as described in any one of claims 1 to 4 based on the standard RANS vector of the new customer, republish new cluster centroids, determine the customer segment label corresponding to the cluster centroid with the smallest distance to the standard RANS vector of the new customer, and formulate a product strategy according to the determined customer segment label. If not, determine the customer segment label corresponding to the cluster centroid with the smallest distance and formulate a product strategy according to the determined customer segment label.

6. The online customer segmentation analysis method for electronic business cards in communication scenarios according to claim 5, characterized in that, When the target customer is identified as a new customer, before obtaining the standard RANS vector corresponding to the new customer, the process also includes: determining whether the observation period of the new customer meets the time window threshold. If not, the new customer will be set as a target user to be observed, and the new customer will be re-evaluated as a user to be observed until the new customer is a user to be observed. If so, the standard RANS vector within the observation period time window is determined as the standard RANS vector corresponding to the new customer, and subsequent steps are executed.

7. The online customer segmentation analysis method for electronic business cards in communication scenarios according to claim 5, characterized in that, Determine the distance between the standard RANS vector of the new customer and other vectors within the cluster corresponding to the initial clustering result. Based on the determined distance, determine whether offline clustering needs to be restarted. Specifically, this includes: according to Determine whether offline clustering needs to be restarted; in, s ( i D is the evaluation indicator. cii D is the average distance from the standard RANS vector of a new customer to other vectors within the corresponding cluster. cij The average distance from the standard RANS vector of a new customer to vectors within other clusters; like s ( i If the value of ) is close to 1, then it is determined that offline clustering does not need to be restarted; like s ( i If the value of ) is close to -1 or 0, then determine s ( i If the number of new customers whose value is close to -1 or 0 has reached the restart threshold, then offline clustering needs to be restarted; otherwise, offline clustering does not need to be restarted.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the offline customer segmentation method for electronic business cards in communication scenarios as described in any one of claims 1 to 4, and / or the online customer segmentation analysis method for electronic business cards in communication scenarios as described in any one of claims 5 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the offline customer segmentation method for electronic business cards in communication scenarios as described in any one of claims 1 to 4, and / or the online customer segmentation analysis method for electronic business cards in communication scenarios as described in any one of claims 5 to 7.

Images