Product flow adjusting method and device and electronic equipment

Abstract

The invention provides a product process adjusting method, an adjusting device and electronic equipment. The method comprises the following steps: constructing a lost customer network and a passing customer network according to customer information of each stage in a product flow, and dividing the lost customer network into a plurality of lost customer network modules; respectively carrying out global characteristic analysis and local characteristic analysis on the client network, and respectively carrying out global characteristic analysis and local characteristic analysis on the lost client network and the plurality of lost client network modules; calculating a similarity coefficient between each lost client network module and the passing client network; under the condition that the lost client network module is similar to the passing client network, the loading speed and the response time of the target node of the product flow are at least adjusted, and under the condition that the lost client network module is not similar to the passing client network, the operation steps of the target node of the product flow are at least simplified. The method solves the problem that in the prior art, the optimization effect of the product process is poor.

Description

Technical Field

[0001] The present application relates to the field of product processes. Specifically, it relates to a method for adjusting a product process, a method for applying a product process, an apparatus for adjusting a product process, a computer-readable storage medium, and an electronic device. Background Art

[0002] With the development of the digital age, people's dependence on the Internet is getting stronger and stronger. The rise of Internet finance has brought huge impacts and changes to the traditional financial industry. Through the Internet, financial institutions can provide various financial products and services at lower costs and higher efficiencies, thus attracting a large number of customers. However, with the intensification of competition, relying solely on the cost performance of the product itself is no longer sufficient. Financial institutions need to consider how to retain and attract customers by optimizing the product process.

[0003] Traditional process optimization methods mainly analyze data such as the residence time of each page when customers use the product, the customer churn rate of each page, and the business conversion rate, and then optimize the entire product process. However, due to the similarities and differences among customer groups, the traditional unified optimization of the entire process cannot meet the personalized needs of different customer groups, resulting in poor optimization effects of the product process.

[0004] Therefore, there is an urgent need for a method to improve the optimization effect of the product process. Summary of the Invention

[0005] The main objective of the present application is to provide a method for adjusting a product process, a method for applying a product process, an apparatus for adjusting a product process, a computer-readable storage medium, and an electronic device, so as to at least solve the problem of poor optimization effects of the product process in the prior art.

[0006] To achieve the above object, according to one aspect of the present application, there is provided a method for adjusting a product process, including: obtaining customer information at each stage of the product process, where the customer information includes churn customer information and passing customer information. The churn customer information is the information of customers who have not completed the product process at the target node, and the passing customer information is the information of the customers who have completed the product process at the target node. The product process is at least composed of a user registration or login node, a data input node, a product selection node, and a transaction completion node, and the target node is any one of the nodes in the product process; constructing a churn customer network and a passing customer network according to the customer information, and using a community division algorithm to divide the churn customer network to obtain a plurality of churn customer network modules; performing a global characteristic analysis on the passing customer network to obtain a first analysis characteristic, and performing a local characteristic analysis on the passing customer network to obtain a second analysis characteristic; performing the global characteristic analysis on the churn customer network to obtain a third analysis characteristic, and performing the local characteristic analysis on a plurality of the churn customer network modules to obtain a plurality of fourth analysis characteristics; calculating a similarity coefficient between each of the churn customer network modules and the passing customer network at least according to the first analysis characteristic, the second analysis characteristic, the third analysis characteristic, and the plurality of fourth analysis characteristics; determining whether each of the churn customer network modules is similar to the passing customer network by determining whether the similarity coefficient is within a preset similarity range. In the case where the churn customer network module is similar to the passing customer network, at least adjust the loading speed and response time of the target node of the product process. In the case where the churn customer network module is not similar to the passing customer network, at least simplify the operation steps of the target node of the product process, thereby realizing the customization of the product process.

[0007] Optionally, constructing a churn customer network according to the customer information includes: mapping a plurality of the churn customer information to a plurality of first nodes in an initial churn customer network; using the Pearson correlation coefficient to calculate a first weight between two of the first nodes respectively, where the first weight is the similarity between the two first nodes; determining the first nodes to be connected and the first nodes to be deleted by comparing the magnitudes of a plurality of the first weights with a first preset value, and connecting the first nodes to be connected and deleting the first nodes to be deleted, thereby obtaining the churn customer network.

[0008] Optionally, by comparing the magnitudes of multiple said first weights with a first preset value, the first nodes to be connected and the first nodes to be deleted are determined, and the first nodes to be connected are connected and the first nodes to be deleted are deleted, thereby obtaining the churned customer network, including: a first comparison step, in the case where the first weight is greater than the first preset value, connecting the two first nodes corresponding to the first weight, taking the first weight as the weight of the first connection edge between the two first nodes, the first connection edge between the first nodes representing the similarity relationship between the two first nodes, and the weight of the first connection edge between the first nodes being the degree of similarity between the two first nodes; comparing multiple said first weights with the first preset value, and repeating the first comparison step at least once until it is determined whether to connect between any two first nodes, and deleting the first nodes that are not connected to any other first nodes, thereby obtaining the churned customer network.

[0009] Optionally, according to the customer information, a passed customer network is constructed, including: mapping multiple said passed customer information to multiple second nodes in an initial passed customer network; using the Pearson correlation coefficient to calculate the second weights between two said second nodes respectively, the second weight being the similarity between the two second nodes; by comparing the magnitudes of multiple said second weights with a second preset value, the second nodes to be connected and the second nodes to be deleted are determined, and the second nodes to be connected are connected and the second nodes to be deleted are deleted, thereby obtaining the passed customer network.

[0010] Optionally, by comparing the magnitudes of multiple said second weights with a second preset value, the second nodes to be connected and the second nodes to be deleted are determined, and the second nodes to be connected are connected and the second nodes to be deleted are deleted, thereby obtaining the passed customer network, including: a second comparison step, in the case where the second weight is greater than the second preset value, connecting the two second nodes corresponding to the second weight, taking the second weight as the weight of the second connection edge between the two second nodes, the second connection edge between the second nodes representing the similarity relationship between the two second nodes, and the weight of the second connection edge between the second nodes being the degree of similarity between the two second nodes; comparing multiple said second weights with the second preset value, and repeating the second comparison step at least once until it is determined whether to connect between any two second nodes, and deleting at least the second nodes that have no connections, thereby obtaining the passed customer network.

[0011] Optionally, perform global feature analysis and local feature analysis on the customer network to obtain a first analysis feature and a second analysis feature, including: using a non-linear transformation method to convert the second weight between the two second nodes into a first distance between the two second nodes; according to Calculate the first global efficiency GE1, where L 1ij is the first distance between the second node i and the second node j, and N1 is the number of second nodes in the customer network; according to Calculate the first average shortest path length ASPL1, so as to obtain the first analysis feature [GE1, ASPL1]; according to Calculate the first local efficiency LE1; according to Calculate the first clustering coefficient C of the second node 1i , where E 1i is the number of edges connected by the second nodes adjacent to the second node i, and k 1i is the total number of adjacent nodes of the second node i; according to Calculate the first node degree D of the customer network 1i , where a 1ij is obtained according to whether there is a second connection edge between the second node i and the second node j. When there is a second connection edge between the second node i and the second node j, a 1ij is 1, and when there is no second connection edge between the second node i and the second node j, a 1ij is 0. The second analysis feature is [LE1, C 11i …C 1N1i , D 1i .

[0012] Optionally, perform the global feature analysis on the churned customer network and perform the local feature analysis on multiple churned customer network modules to obtain a third analysis feature and multiple fourth analysis features, including: using a non-linear transformation method to convert the first weight between the two first nodes into a second distance between the two first nodes; according to Calculate the second global efficiency GE2, where L 2mn is the second distance between the first node m and the first node n, and N2 is the number of first nodes in the churned customer network; according to Calculate the second average shortest path length ASPL2, and the third analysis feature is [GE2, ASPL2]; according to Calculate the second local efficiency LE2; according to Calculate the second clustering coefficient C of the first node in each churned customer network module2m , where E 2m is the number of edges connected by the first node adjacent to the first node m, and k 2m is the total number of adjacent nodes of the first node m; According to calculate the second node degree D of each of the churned customer network modules 2m , where a 2mn is obtained based on whether there is the first connection edge between the first node m and the first node n. When there is the first connection edge between the first node m and the first node n, a 2mn is 1. When there is no first connection edge between the first node m and the first node n, a 2mn is 0. The multiple fourth analysis features are [LE2, C 21m , D 21m …[LE2, C 2Mm , D 2Mm , where M is the number of the churned customer network modules.

[0013] Optionally, calculate the similarity coefficient between each of the churned customer network modules and the passed customer network at least according to the first analysis feature, the second analysis feature, the third analysis feature, and the multiple fourth analysis features, including: performing vector fusion on the first analysis feature and the second analysis feature to obtain a first feature vector; normalizing the first feature vector to obtain a second feature vector; performing vector fusion on the third analysis feature and the multiple fourth analysis features respectively to obtain multiple third feature vectors; normalizing the multiple third feature vectors respectively to obtain multiple fourth feature vectors; using the cosine similarity method to calculate the similarity between the multiple fourth feature vectors and the second feature vector respectively to obtain the similarity coefficients between the multiple churned customer network modules and the passed customer network.

[0014] Optionally, determine whether each of the churned customer network modules is similar to the passed customer network by determining whether the similarity coefficient is within a preset similarity range, including: when the similarity coefficient is within the preset similarity range, determine that the churned customer network module is similar to the passed customer network; when the similarity coefficient is not within the preset similarity range, determine that the churned customer network module is not similar to the passed customer network.

[0015] Optionally, when the churned customer network module is similar to the passed customer network, at least adjust the loading speed and response time of the target node of the product process, including: when the churned customer network module is similar to the passed customer network, at least use asynchronous loading technology so that when the user browses the target node of the product process, the content of other nodes of the product process is not loaded, thereby adjusting the loading speed and the response time of the target node of the product process.

[0016] Optionally, when the churned customer network module is not similar to the passed customer network, at least simplify the operation steps of the target node of the product process, including: when the churned customer network module is not similar to the passed customer network, at least simplify the form tax design of the target node and reduce the required fields of the form, thereby simplifying the operation steps of the target node of the product process.

[0017] According to another aspect of the present application, there is provided a method for applying a product process, including: adjusting a product process by using any one of the above-mentioned product process adjustment methods to obtain a target product process; pushing the target product process to a target customer, where the target customer at least includes the churned customers of the product process.

[0018] According to still another aspect of the present application, there is provided a product process adjustment device, including:

[0019] An acquisition unit that acquires customer information at each stage of a product process, where the customer information includes churned customer information and passed customer information, the churned customer information is the information of customers who have not completed the product process of the target node, the passed customer information is the information of the customers who have completed the product process of the target node, the product process is at least composed of a user registration or login node, a data input node, a product selection node, and a transaction completion node, and the target node is any node in the product process;

[0020] A construction unit that constructs a churned customer network and a passed customer network according to the customer information, and uses a community division algorithm to divide the churned customer network to obtain a plurality of churned customer network modules;

[0021] A first analysis unit that performs a global characteristic analysis on the passed customer network to obtain a first analysis characteristic, and performs a local characteristic analysis on the passed customer network to obtain a second analysis characteristic;

[0022] A second analysis unit performs the global feature analysis on the lost customer network to obtain a third analysis feature, and performs the local feature analysis on a plurality of the lost customer network modules to obtain a plurality of fourth analysis features;

[0023] A calculation unit calculates a similarity coefficient between each of the lost customer network modules and the passed customer network based on at least the first analysis feature, the second analysis feature, the third analysis feature, and the plurality of fourth analysis features;

[0024] A determination unit determines whether each of the lost customer network modules is similar to the passed customer network by determining whether the similarity coefficient is within a preset similarity range. When a lost customer network module is similar to the passed customer network, at least the loading speed and response time of the target node of the product process are adjusted. When a lost customer network module is not similar to the passed customer network, at least the operation steps of the target node of the product process are simplified, thereby realizing the customization of the product process.

[0025] According to another aspect of the present application, there is provided a computer-readable storage medium, where the computer-readable storage medium includes a stored program, and when the program runs, it controls the device where the computer-readable storage medium is located to execute any one of the above product process adjustment methods.

[0026] According to another aspect of the present application, there is provided an electronic device, including: one or more processors, a memory, and one or more programs, where the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include those for executing any one of the above product process adjustment methods.

[0027] Applying the technical solution of the present application, first, obtain customer information at each stage of the product process; secondly, construct a churned customer network and a passed customer network based on the churned customer information and passed customer information in the above customer information, and use a community partitioning algorithm to divide the above churned customer network into multiple churned customer network modules; then, perform global feature analysis and local feature analysis on the passed customer network respectively, perform global feature analysis on the churned customer network, and perform local feature analysis on multiple churned customer network modules. According to the above analysis feature results, calculate the similarity coefficient between each churned customer network module and the passed customer network; finally, determine whether each churned customer network module is similar to the passed customer network according to the similarity coefficient. In the case where the churned customer network module is similar to the passed customer network, at least adjust the loading speed and response time of the target node of the product process. In the case where the churned customer network module is not similar to the passed customer network, at least simplify the operation steps of the target node of the product process. Compared with the prior art, the solution of the present application classifies customers into passed customers and churned customers, constructs a churned customer network and a passed customer network, uses a community partitioning algorithm to divide the churned customer network into different churned customer network modules according to features, and determines whether the churned customer network module is similar to the passed customer network according to the global feature analysis and local feature analysis of the customer network, the global feature analysis of the churned customer network, and the local feature analysis of the churned customer network module. In the case of similarity, at least adjust the loading speed and response time of the target node of the product process. In the case of dissimilarity, at least simplify the operation steps of the target node of the product process, so as to realize the customization of the product process according to the different characteristics of churned customers, that is, the method formulates corresponding processes for different churned customers, meets the personalized needs of different customer groups, and thus solves the problem of poor optimization effect of the product process in the prior art. Description of the Drawings

[0028] The specification drawings forming a part of the present application are used to provide a further understanding of the present application. The schematic embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation of the present application. In the drawings:

[0029] Figure 1 Shows a hardware structure block diagram of a mobile terminal for performing an adjustment method of a product process provided in an embodiment of the present application;

[0030] Figure 2 Shows a flowchart of an adjustment method of a product process provided in an embodiment of the present application;

[0031] Figure 3 Shows a step diagram of an adjustment method of a product process provided in an embodiment of the present application;

[0032] Figure 4 The structural block diagram of an adjustment device for a product process provided according to an embodiment of the present application is shown;

[0033] Figure 5 The schematic flow chart of an application method for a product process provided according to an embodiment of the present application is shown. Detailed implementation manners

[0034] It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments may be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

[0035] In order to enable those skilled in the art to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0036] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so as to implement the embodiments of the present application described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or are inherent to these processes, methods, products or devices.

[0037] As introduced in the background art, the problem of poor optimization effect of the product process in the prior art. To solve the above problems, the embodiments of the present application provide an adjustment method for a product process, an application method for a product process, an adjustment device for a product process, a computer-readable storage medium, and an electronic device.

[0038] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0039] The method embodiments provided in the embodiments of the present application can be executed on a mobile terminal, a computer terminal, or a similar computing device. Taking running on a mobile terminal as an example, Figure 1 is the hardware structural block diagram of a mobile terminal of an adjustment method for a product process according to an embodiment of the present invention. As Figure 1As shown, the mobile terminal may include one or more ( Figure 1 only one is shown in the figure) processors 102 (the processor 102 may include, but is not limited to, processing devices such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data. Among them, the above-mentioned mobile terminal may further include a transmission device 106 for communication functions and an input / output device 108. Those of ordinary skill in the art can understand that Figure 1 the structure shown is only schematic and does not limit the structure of the above-mentioned mobile terminal. For example, the mobile terminal may further include more or fewer components than Figure 1 shown in the figure, or have a different configuration from Figure 1 shown in the figure.

[0040] The memory 104 can be used to store computer programs. For example, software programs and modules of application software, such as the computer program corresponding to the display method of device information in the embodiments of the present invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, implements the above-mentioned method. The memory 104 may include a high-speed random access memory and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memories, or other non-volatile solid-state memories. In some instances, the memory 104 may further include a memory remotely set relative to the processor 102, and these remote memories can be connected to the mobile terminal through a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof. The transmission device 106 is used to receive or send data via a network. Specific examples of the above-mentioned network may include a wireless network provided by a communication provider of the mobile terminal. In one instance, the transmission device 106 includes a network adapter (abbreviated as NIC), which can be connected to other network devices through a base station and thus can communicate with the Internet. In one instance, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0041] In this embodiment, a method for adjusting the product process running on a mobile terminal, a computer terminal, or a similar computing device is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. And although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order from that here.

[0042] Figure 2 is a flowchart of the method for adjusting the product process according to the embodiments of the present application. As Figure 2As shown, the method includes the following steps:

[0043] Step S201: Obtain customer information at each stage of the product process. Among them, the above customer information includes churned customer information and passed customer information. The above churned customer information is the information of customers who have not completed the product process of the target node. The above passed customer information is the information of the above customers who have completed the product process of the above target node. The above product process is at least composed of a user registration or login node, a data input node, a product selection node, and a transaction completion node. The above target node is any one of the nodes in the above product process.

[0044] Specifically, the customer information mainly includes: (1) Personal information indicators: including age, gender, marital status, education level, occupation, etc. These indicators can reflect the basic personal situation and stability of customers. (2) Financial status indicators: including income, expenditure, savings, liabilities, etc. These indicators can reflect the financial status and debt repayment ability of customers. (3) Credit record indicators: including credit rating, credit card usage, overdue situation, etc. These indicators can reflect the credit record and credit risk of customers. (4) Behavioral characteristic indicators: including consumption habits, online shopping situation, social media usage situation, etc. These indicators can reflect the behavioral characteristics and lifestyle of customers. (5) Personal preference indicators: including investment preference, consumption preference, lifestyle, etc. These indicators can reflect the personal preferences and needs of customers. (6) Loan history indicators: the number of loan applications in the previous year, the total loan amount in the previous year, overdue situation, etc. These indicators can reflect the historical loan situation of customers.

[0045] Step S202: According to the above customer information, construct a churned customer network and a passed customer network, and use a community detection algorithm to partition the above churned customer network to obtain multiple churned customer network modules.

[0046] Specifically, the community detection algorithm is an algorithm used to partition a network or graph into several subgraphs or communities. The purpose of the community detection algorithm is to discover subgraphs with high-density connections and close associations in the network, that is, communities. These communities usually have certain functional or structural meanings in the network, such as friend circles in social networks, groups for information dissemination, etc.

[0047] Specifically, common community detection algorithms include, for example, the Louvain algorithm, label propagation algorithm, genetic algorithm, etc. Among them, the Louvain algorithm: measures the quality of the community structure based on modularity, and maximizes the modularity by iteratively optimizing the community membership of nodes in the network; the label propagation algorithm: each node is initially assigned a unique label, and then the labels are propagated through the relationships between nodes until convergence, forming communities; the genetic algorithm: searches for the optimal community detection scheme by simulating the process of biological evolution. Through the iterative optimization process of the genetic algorithm, a better community detection scheme can be found, enabling the nodes in the network to be reasonably divided into different communities, thereby better understanding the network structure and the relationships between nodes.

[0048] Specifically, this method is described by taking the particle swarm optimization algorithm in community detection algorithms as an example. The basic idea of the particle swarm optimization algorithm is as follows: A group of particles is randomly initialized in a continuous space. The positions of the particles can theoretically take any values in the continuous search space. Moreover, each position can be regarded as a feasible solution to the target problem, and the quality of the solution is evaluated by fitness. The particles fly in the continuous search space to search for better solutions. Each movement is an iteration. The particles reach new positions by adding the increment of velocity to the old positions. After each iteration, the particles determine whether to update their own best solutions Pbest according to the fitness of the new positions. The particle swarm determines whether to update the global best solution Gbest of the particle swarm according to the optimal fitness of the current positions of all particles. Now, assume that the number of particles in the initialized particle swarm is P, and the dimensionality of the solution space is q. For a certain particle i (1 ≤ i ≤ q), the current velocity of the particle is v i =(v i1 ,v i2 ,v i3 ,…v iq ), the current best solution of the particle is pbest i4 =(pbest i1 ,pbest i2 ,pbest i3 ,…pbest iq ), the current best solution of the entire particle swarm is gbest i =(gbest i2 ,gbest i3 ,…gbest iq ), and the particle swarm updates the position x i =(x i1 ,x i2 ,x i3 ,…x iq ) of the particles according to these three values. Excessive dimensions of the particles will cause severe fluctuations of the particles, so they need to be limited within a range [-v max, v max . The velocity and position update formulas of the particles are as follows:

[0049]

[0050] where c1 and c2 are acceleration coefficients, w is the inertia coefficient, r1 and r2 are random numbers between 0 and 1, and v max is the maximum velocity that the particle can reach.

[0051] Furthermore, the steps of the particle swarm optimization algorithm are as follows: Step 1: Read the initialization parameters of the particle swarm, and initialize the position and velocity of each particle in the particle swarm, the individual optimal solution, the global optimal solution, etc.; Step 2: Calculate the fitness of each particle and update its own optimal solution Pbest; Step 3: Update the global optimal solution Gbest; Step 4: Update the velocity according to ; Step 5: Update the position according to the position update formula; Step 6: Determine whether the iteration is completed. If the iteration is completed, go to Step 7. If the iteration is not completed, return to Step 2 to recalculate the current particle fitness and its own optimal solution Gbest; Step 7: The module division is completed, and the result is output.

[0052] Step S203: Perform the above-mentioned global characteristic analysis on the customer network to obtain the first analysis characteristic, and perform the above-mentioned local characteristic analysis on the customer network to obtain the second analysis characteristic.

[0053] Step S204: Perform the above-mentioned global characteristic analysis on the churned customer network to obtain the third analysis characteristic, and perform the above-mentioned local characteristic analysis on multiple modules of the churned customer network to obtain multiple fourth analysis characteristics.

[0054] Step S205: Calculate the similarity coefficient between each of the above-mentioned churned customer network modules and the above-mentioned customer network at least according to the above-mentioned first analysis characteristic, the above-mentioned second analysis characteristic, the above-mentioned third analysis characteristic, and the multiple above-mentioned fourth analysis characteristics.

[0055] Specifically, compare the analysis characteristic results of the churned customer network module with the analysis characteristic results of the customer network, and study the similarities and differences therein.

[0056] Step S206: Determine whether each of the above-mentioned churn customer network modules is similar to the above-mentioned passing customer network by judging whether the above similarity coefficient is within a preset similarity range. In the case where the churn customer network module is similar to the passing customer network, at least adjust the loading speed and response time of the above target node of the above product process. In the case where the churn customer network module is not similar to the passing customer network, at least simplify the operation steps of the above target node of the above product process, so as to realize the customization of the above product process.

[0057] Specifically, according to the characteristics of the above customer groups, match a similar passing customer network for the churn customer network module. If the similarity is high, at least adjust the loading speed and response time of the target node of the product process. If the similarity is low, at least simplify the operation steps of the target node of the product process to realize the customization of the product process.

[0058] In the above embodiment, first, obtain the customer information at each stage of the product process; secondly, construct a churn customer network and a passing customer network according to the churn customer information and passing customer information in the above customer information, and use the community division algorithm to divide the above churn customer network into multiple churn customer network modules; then, perform global feature analysis and local feature analysis on the passing customer network respectively, perform global feature analysis on the churn customer network, and perform local feature analysis on multiple churn customer network modules. According to the above analysis feature results, calculate the similarity coefficient between each churn customer network module and the passing customer network; finally, determine whether each churn customer network module is similar to the passing customer network according to the similarity coefficient. In the case where the churn customer network module is similar to the passing customer network, at least adjust the loading speed and response time of the target node of the product process. In the case where the churn customer network module is not similar to the passing customer network, at least simplify the operation steps of the target node of the product process. Compared with the prior art, the solution of the present application classifies customers into passing customers and churn customers, constructs a churn customer network and a passing customer network, uses the community division algorithm to divide the churn customer network into different churn customer network modules according to characteristics, and judges whether the churn customer network module is similar to the passing customer network according to the global feature analysis and local feature analysis of the customer network, the global feature analysis of the churn customer network, and the local feature analysis of the churn customer network module. In the case of similarity, at least adjust the loading speed and response time of the target node of the product process. In the case of dissimilarity, at least simplify the operation steps of the target node of the product process, so as to realize the customization of the product process according to the different characteristics of the churn customers, that is, the method formulates corresponding processes for different churn customers, meets the personalized needs of different customer groups, and thus solves the problem of poor optimization effect of the Internet product process in the prior art.

[0059] In an embodiment of the present application, according to the above customer information, a churned customer network is constructed, including: mapping a plurality of the above churned customer information to a plurality of first nodes in an initial churned customer network; using the Pearson correlation coefficient to calculate the first weight between any two of the above first nodes respectively, where the first weight is the similarity between the two first nodes; by comparing the magnitude relationship between a plurality of the above first weights and a first preset value, determining the above first nodes to be connected and the above first nodes to be deleted, and connecting the above first nodes to be connected and deleting the above first nodes to be deleted, thereby obtaining the above churned customer network. In this method, by mapping customer information to nodes in the network and using the Pearson correlation coefficient to calculate the weights between nodes, the correlation and structure among the churned customer groups can be more effectively revealed, and by comparing the magnitudes of the weights and the preset value, it can be determined which nodes need to be connected and which nodes need to be deleted, so as to more accurately construct the churned customer network, and further more accurately adjust the product process, and further improve the optimization effect of the product process.

[0060] Specifically, any two customers or any two churned customers are defined as two variables X and Y, and the Pearson correlation coefficient between X and Y is defined as the quotient of the covariance and standard deviation between the two variables:

[0061]

[0062] Estimating the covariance and standard deviation of the sample, the Pearson correlation coefficient r can be obtained:

[0063]

[0064] Further simplifying r in the above formula, and representing it by the mean of the standard scores of the sample points (X i , Y i ), the following new expression is obtained:

[0065]

[0066] Among them, σ X and are the mean, standard deviation and standard score of X i respectively.

[0067] In order to more accurately determine the first nodes to be connected and the first nodes to be deleted, so as to more accurately construct the churned customer network, in an embodiment of the present application, by comparing the magnitudes of multiple above-mentioned first weights and a first preset value, the above-mentioned first nodes to be connected and the above-mentioned first nodes to be deleted are determined, and the above-mentioned first nodes to be connected are connected and the above-mentioned first nodes to be deleted are deleted, so as to obtain the above-mentioned churned customer network, including: a first comparison step, when the above-mentioned first weight is greater than the above-mentioned first preset value, connecting the two above-mentioned first nodes corresponding to the above-mentioned first weight, taking the above-mentioned first weight as the weight of the first connection edge between the two above-mentioned first nodes, the above-mentioned first connection edge between the above-mentioned first nodes represents the similarity relationship between the two above-mentioned first nodes, and the above-mentioned weight of the above-mentioned first connection edge between the above-mentioned first nodes is the degree of similarity between the two above-mentioned first nodes; comparing multiple above-mentioned first weights with the above-mentioned first preset value, and repeating the above-mentioned first comparison step at least once until it is determined whether to connect between any two above-mentioned first nodes, and deleting the above-mentioned first nodes that are not connected to any other above-mentioned first nodes, so as to obtain the above-mentioned churned customer network.

[0068] Specifically, the selection of the preset value usually needs to be determined according to specific business requirements, data characteristics, and subsequent analysis goals. For example, if the goal is to identify highly similar customer groups, the preset value can be set relatively high; if the goal is to construct a comprehensive network structure to discover potential association patterns, the preset value can be set relatively low.

[0069] Specifically, when the first weight (i.e., the similarity between two first nodes) is greater than the first preset value, it indicates that there is a significant correlation or similar feature between these two churned customers. Therefore, a first connection edge is established in the network and the first weight is assigned as the weight value of this first connection edge. The magnitude of the weight value represents the degree of similarity or correlation between the two nodes. Conversely, if the first weight is less than or equal to the first preset value, the first connection edge is not established, indicating that the correlation between these two churned customers is weak or not significant enough to form an effective connection in the network. After comparing all node pairs, there may be some isolated first nodes in the network, that is, nodes that are not connected to any other nodes. These isolated points may not contain sufficient information, or their churn reasons are significantly different from other customer groups.

[0070] In another embodiment, according to the above customer information, a passing customer network is constructed, including: mapping a plurality of the above passing customer information to a plurality of second nodes in the initial passing customer network; using the Pearson correlation coefficient to calculate the second weight between two of the above second nodes respectively, where the second weight is the similarity between the two second nodes; by comparing the magnitude relationship between a plurality of the above second weights and a second preset value, determining the second nodes to be connected and the second nodes to be deleted among the above second nodes, and connecting the second nodes to be connected and deleting the second nodes to be deleted, thereby obtaining the above passing customer network. In this method, by mapping customer information to nodes in the network and using the Pearson correlation coefficient to calculate the weights between nodes, the relevance and structure among passing customer groups can be more effectively revealed. And by comparing the magnitudes of the weights and the preset value, it can be determined which nodes need to be connected and which nodes need to be deleted, thereby more accurately constructing the passing customer network and further more effectively improving the optimization effect of the product process.

[0071] In order to more accurately determine the second nodes to be connected and the second nodes to be deleted, and thus more accurately construct the passing customer network, in an embodiment of the present application, by comparing the magnitude relationship between a plurality of the above second weights and a second preset value, determining the second nodes to be connected and the second nodes to be deleted among the above second nodes, and connecting the second nodes to be connected and deleting the second nodes to be deleted, thereby obtaining the above passing customer network, including: a second comparison step, in the case where the above second weight is greater than the above second preset value, connecting the two above second nodes corresponding to the above second weight, taking the above second weight as the weight of the second connection edge between the two above second nodes, where the above second connection edge between the above second nodes represents the relationship of similarity between the two above second nodes, and the above weight of the above second connection edge between the above second nodes is the degree of similarity between the two above second nodes; comparing a plurality of the above second weights with the above second preset value, and repeating the above second comparison step at least once until it is determined whether any two of the above second nodes are connected, and at least deleting the above second nodes without any connection, thereby obtaining the above passing customer network.

[0072] In yet another embodiment, a global characteristic analysis and a local characteristic analysis are performed on the above passing customer network to obtain a first analysis characteristic and a second analysis characteristic, including: using a non-linear transformation method to transform the above second weight between two of the above second nodes into a first distance between the two above second nodes; according to calculating the first global efficiency GE1, where L 1ij is the above first distance between the above second node i and the above second node j, and N1 is the number of the above second nodes in the above passing customer network; according to Calculate the first average shortest path length ASPL1, so as to obtain the above first analysis feature [GE1, ASPL1]; According to Calculate the first local efficiency LE1; According to Calculate the first clustering coefficient C of the above second node 1i , where, E 1i is the number of edges connected by the above second nodes adjacent to the above second node i, k 1i is the total number of adjacent nodes of the above second node i; According to Calculate the first node degree D of the above through the customer network 1i , in the formula, a 1ij is obtained according to whether there is the above second connection edge between the above second node i and the above second node j. When there is the above second connection edge between the above second node i and the above second node j, a 1ij is 1. When there is no the above second connection edge between the above second node i and the above second node j, a 1ij is 0. The above second analysis feature is [LE1, C 11i …C 1N1i , D 1i . In this method, the first global efficiency, the first average shortest path length, the first local efficiency, the first clustering coefficient and the first node degree are combined to form the first analysis feature and the second analysis feature, so that the characteristics of the customer group in the customer network can be evaluated more accurately.

[0073] Specifically, the global efficiency describes the network from the perspective of the overall information flow, which can better reflect the processing ability and transmission ability of the network global information, and can also reflect the integration degree of the network; The average shortest path length represents the ability of information transfer between network nodes, which can better reflect the functional integration level between regions. The lower this value is, the higher the functional integration level between regions is; The local efficiency is usually used to measure the separation degree of the network, which can better reflect the processing ability and transmission ability of local information; The clustering coefficient is usually used to measure the aggregation degree of nodes, and is positively correlated with the aggregation degree; The node degree is the sum of the out-degree and in-degree of nodes in the network, which is related to the complexity of the network and is mostly used in the research of network modules.

[0074] Specifically, combining the first global efficiency, the first average shortest path length, the first local efficiency, the first clustering coefficient and the first node degree to form the first analysis feature and the second analysis feature can more comprehensively evaluate the characteristics of the customer group. This combined feature analysis method provides an in-depth understanding of the behavior and association patterns of the customer group in the customer network from both the global and local levels, and provides a solid data basis for subsequent customer group comparison analysis and process customization.

[0075] In order to more accurately evaluate the characteristics of customer groups in the churned customer network module, so as to more accurately adjust the product process and then more effectively improve the optimization effect of the product process. In an embodiment of the present application, the above-mentioned global characteristics analysis is performed on the above-mentioned churned customer network, and the above-mentioned local characteristics analysis is performed on multiple above-mentioned churned customer network modules to obtain a third analysis characteristic and multiple fourth analysis characteristics, including: using a non-linear transformation method to convert the above-mentioned first weight between two above-mentioned first nodes into a second distance between the two above-mentioned first nodes; according to Calculate the second global efficiency GE2, where L 2mn is the above-mentioned second distance between the above-mentioned first node m and the above-mentioned first node n, and N2 is the number of the above-mentioned first nodes in the above-mentioned churned customer network; according to Calculate the second average shortest path length ASPL2, and the above-mentioned third analysis characteristic is [GE2, ASPL2]; according to Calculate the second local efficiency LE2; according to Calculate the second clustering coefficient C of the above-mentioned first nodes in each of the above-mentioned churned customer network modules 2m , where E 2m is the number of edges connected by the above-mentioned first nodes adjacent to the above-mentioned first node m, and k 2m is the total number of adjacent nodes of the above-mentioned first node m; according to Calculate the second node degree D of each of the above-mentioned churned customer network modules 2m , where a 2mn is obtained according to whether there is the above-mentioned first connection edge between the above-mentioned first node m and the above-mentioned first node n. When there is the above-mentioned first connection edge between the above-mentioned first node m and the above-mentioned first node n, a 2mn is 1, and when there is no the above-mentioned first connection edge between the above-mentioned first node m and the above-mentioned first node n, a 2mn is 0. The multiple above-mentioned fourth analysis characteristics are [LE2, C 21m , D 21m …[LE2, C 2Mm , D 2Mm , where M is the number of the above-mentioned churned customer network modules.

[0076] In yet another embodiment, at least based on the above-mentioned first analysis feature, the above-mentioned second analysis feature, the above-mentioned third analysis feature, and multiple above-mentioned fourth analysis features, calculate the similarity coefficient between each of the above-mentioned churned customer network modules and the above-mentioned passed customer network, including: performing vector fusion on the above-mentioned first analysis feature and the above-mentioned second analysis feature to obtain a first feature vector; normalizing the above-mentioned first feature vector to obtain a second feature vector; performing vector fusion on the above-mentioned third analysis feature and multiple above-mentioned fourth analysis features respectively to obtain multiple third feature vectors; normalizing multiple above-mentioned third feature vectors respectively to obtain multiple fourth feature vectors; using the KNN algorithm to calculate the similarity between multiple above-mentioned fourth feature vectors and the above-mentioned second feature vector respectively, to obtain the above-mentioned similarity coefficient between each of the above-mentioned churned customer network modules and the above-mentioned passed customer network. In this method, by performing feature analysis on the churned customer network module and the passed customer network, and performing fusion and normalization operations on different feature vectors, finally using the KNN algorithm to calculate the similarity coefficient, it is possible to more accurately understand the similarity degree between the churned customer network module and the passed customer network, and then more accurately adjust the product process, further improving the optimization effect of the product process.

[0077] Specifically, the steps of calculating the similarity coefficient using KNN: Select a suitable distance metric method to quantify the distance between two vectors. Common ones include Euclidean distance, Manhattan distance, Minkowski distance, or cosine similarity, etc.; calculate the similarity coefficient. Based on the KNN algorithm, define a similarity coefficient to evaluate the similarity of two vectors. In this method, the distance metric is transformed into a similarity metric. For example, the distance value is transformed into its reciprocal or a certain exponential function is used to weaken the influence of the distance.

[0078] In order to more accurately determine whether the churned customer network module is similar to the passed customer network, so as to more accurately adjust the product process, and further improve the optimization effect of the product process, in an embodiment of the present application, by determining whether the above-mentioned similarity coefficient is within a preset similarity range, determine whether each of the above-mentioned churned customer network modules is similar to the above-mentioned passed customer network, including: when the above-mentioned similarity coefficient is within the above-mentioned preset similarity range, determine that the above-mentioned churned customer network module is similar to the above-mentioned passed customer network; when the above-mentioned similarity coefficient is not within the above-mentioned preset similarity range, determine that the above-mentioned churned customer network module is not similar to the above-mentioned passed customer.

[0079] In order to more accurately adjust the product process when the lost customer network module is similar to the passing customer network, so as to further improve the optimization effect of the product process. In an embodiment of the present application, when the lost customer network module is similar to the passing customer network, at least adjust the loading speed and response time of the target node of the above product process, including: when the lost customer network module is similar to the passing customer network, at least use asynchronous loading technology so that when the user browses the target node of the above product process, the content of other nodes of the above product process is not loaded, thereby adjusting the above loading speed and the above response time of the target node of the above product process.

[0080] Specifically, in web development, asynchronous loading technology allows a page to load only the specified parts during loading, while other parts are postponed to be loaded under specific conditions. For example, using JavaScript's AJAX (Asynchronous JavaScript and XML) or related frameworks (such as React, Vue, etc.), asynchronous loading of specific nodes can be achieved. When a customer accesses a target node, only the data and elements related to that node are loaded, while the elements of other nodes remain in a lazy loading state and are loaded only when they are actually needed. This can significantly reduce the initial loading time and improve the response speed.

[0081] Specifically, the adjustment process of asynchronous loading technology is as follows: resource grouping, according to the analysis results of the lost customer network and the passing customer network similarity module, divide the nodes in the product process into multiple groups, and each group contains nodes with similar characteristics; loading strategy customization, formulate different asynchronous loading strategies for each group of nodes. For the target node, implement depth-first asynchronous loading, that is, give priority to loading the target node to ensure fast response. For non-target nodes, adopt deferred loading or on-demand loading to reduce unnecessary resource loading; optimize the front-end code, by optimizing the front-end HTML, CSS, and JavaScript codes, ensure that the elements and resources of the target node can be quickly recognized and loaded. This may include using preloading, prefetching, and resource lazy loading and other technologies; back-end optimization, in cooperation with front-end asynchronous loading, the back-end needs to provide support, including on-demand data provision (that is, only return data related to the target node when requested), data compression, and caching strategies to further improve the loading speed and response time; testing and adjustment, after implementing asynchronous loading, strict testing is required, including user behavior testing, performance testing, and compatibility testing, to ensure that the optimized process not only improves the loading speed and response time but also does not affect the user experience and the integrity of the process. According to the test results, it may be necessary to adjust the parameters and strategies of asynchronous loading to achieve the best effect.

[0082] In order to more accurately adjust the product process when the lost customer network module is not similar to the passing customer network, so as to further improve the optimization effect of the product process, in an embodiment of the present application, when the lost customer network module is not similar to the passing customer network, at least simplify the operation steps of the target node of the product process, including: when the lost customer network module is not similar to the passing customer network, at least simplify the form tax design of the target node, and reduce the required fields in the form, so as to simplify the operation steps of the target node of the product process.

[0083] Specifically, the form design optimization mainly targets the target node (for example, the product application form), redesigns the form structure to make it more intuitive and easy to fill out. This may include reducing the required fields in the form, splitting a long form into multiple short steps, using the pre-fill function, or providing a more user-friendly interface prompt.

[0084] Specifically, the required field evaluation is to re-evaluate the necessity of the required fields in the form. Consider whether some items can be made optional, or whether they can be pre-filled in other ways (such as user account information), so as to reduce the burden on users when filling out the form.

[0085] In order to enable those skilled in the art to more clearly understand the technical solution of the present application, the implementation process of the product process adjustment method of the present application will be described in detail below in conjunction with specific embodiments.

[0086] This embodiment relates to a specific product process adjustment method, as Figure 3 shown, including the following steps:

[0087] Step 1: Obtain customer information: Obtain customer information at each stage of the product process, where the customer information includes lost customer information and passing customer information;

[0088] Step 2: Construct a lost customer network: According to the lost customer information, use the Pearson correlation coefficient to calculate the first weight between two first nodes in the initial lost customer network, and connect the corresponding two first nodes when the first weight is greater than the first preset value. Repeat to determine the size relationship between the first weight and the first preset value until it is confirmed whether any two first nodes are connected, and delete the first nodes without any connections, so as to construct a lost customer network;

[0089] Step 3: Construct the passing customer network: According to the passing customer information, use the Pearson correlation coefficient to calculate the second weight between two second nodes in the initial passing customer network, and connect the corresponding two second nodes when the second weight is greater than the second preset value. Repeat the determination of the size relationship between the second weight and the second preset value until it is confirmed whether any two second nodes are connected, and delete the second nodes without any connections, thereby constructing the passing customer network;

[0090] Step 4: Divide the churned customer network into multiple churned customer network modules: Use the community detection algorithm to divide the churned customer network to obtain multiple churned customer network modules;

[0091] Step 5: Conduct global characteristic analysis and local characteristic analysis on the passing customer network: Use the non-linear transformation method to transform the second weight between two second nodes into the first distance between two second nodes, calculate the first global efficiency and the first average shortest path length to obtain the first analysis characteristics, and calculate the first local efficiency, the first clustering coefficient and the first node degree to obtain the second analysis characteristics;

[0092] Step 6: Conduct global characteristic analysis on the churned customer network and conduct local characteristic analysis on multiple churned customer network modules: Use the non-linear transformation method to transform the first weight between two first nodes into the second distance between two first nodes, calculate the second global efficiency and the second average shortest path length to obtain the third analysis characteristics, and calculate the second local efficiency, the second clustering coefficient and the second node degree of multiple churned customer networks to obtain the fourth analysis characteristics;

[0093] Step 7: Calculate the similarity coefficient between each churned customer network module and the passing customer network: Perform a vector fusion operation on the first analysis characteristics and the second analysis characteristics to obtain the first characteristic vector, and perform a normalization operation on the first characteristic vector to obtain the second characteristic vector. Perform a vector fusion operation on the third analysis characteristics and the fourth analysis characteristics to obtain the third characteristic vector, and perform a normalization operation on the third characteristic vector to obtain the fourth characteristic vector. Use the KNN algorithm to calculate the similarity between multiple fourth characteristic vectors and the second characteristic vector respectively, thereby obtaining the similarity coefficient between multiple churned customer network modules and the passing customer network;

[0094] Step 8: Determine whether each churned customer network module is similar to the passing customer network: When the similarity coefficient is within the preset similarity range, determine that the churned customer network module is similar to the passing customer network; when the similarity coefficient is not within the preset similarity range, determine that the churned customer network module is not similar to the passing customer network;

[0095] Step 9: Adjust the product process: When the lost customer network module is similar to the passed customer network, at least use asynchronous loading technology so that when the user browses the target node of the product process, the content of other nodes of the product process is not loaded, thereby adjusting the loading speed and response time of the target node of the product process. When the lost customer network module is not similar to the passed customer network, at least simplify the form tax design of the target node and reduce the required items in the form, thereby simplifying the operation steps of the target node of the product process.

[0096] It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. And although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0097] The embodiment of the present application also provides an adjustment device for the product process. It should be noted that the adjustment device for the product process in the embodiment of the present application can be used to execute the adjustment method for the product process provided by the embodiment of the present application. The device is used to implement the above embodiments and preferred implementation manners, and those that have been described will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that can achieve a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware is also possible and contemplated.

[0098] The following introduces the adjustment device for the product process provided by the embodiment of the present application.

[0099] Figure 4 is a schematic diagram of the adjustment device for the product process according to the embodiment of the present application. As Figure 4 shown, the device includes:

[0100] An acquisition unit 10, which acquires customer information at each stage of the product process. Among them, the above customer information includes lost customer information and passed customer information. The above lost customer information is the information of customers who have not completed the above product process of the target node, and the above passed customer information is the above information of the above customers who have completed the above product process of the above target node. The above product process is at least composed of a user registration or login node, a data input node, a product selection node, and a transaction completion node, and the above target node is any node in the above product process.

[0101] Specifically, customer information mainly includes: (1) Personal information indicators: including age, gender, marital status, education level, occupation, etc. These indicators can reflect the basic personal situation and stability of customers. (2) Financial status indicators: including income, expenditure, deposits, liabilities, etc. These indicators can reflect the financial status and debt repayment ability of customers. (3) Credit record indicators: including credit ratings, credit card usage, overdue situations, etc. These indicators can reflect the credit records and credit risks of customers. (4) Behavioral characteristic indicators: including consumption habits, online shopping situations, social media usage situations, etc. These indicators can reflect the behavioral characteristics and lifestyles of customers. (5) Personal preference indicators: including investment preferences, consumption preferences, lifestyles, etc. These indicators can reflect the personal preferences and needs of customers. (6) Loan history indicators: the number of loan applications in the previous year, the total loan amount in the previous year, overdue situations, etc. These indicators can reflect the historical loan situations of customers.

[0102] The construction unit 20 constructs a churned customer network and a passing customer network based on the above customer information, and uses a community partitioning algorithm to partition the above churned customer network to obtain multiple churned customer network modules.

[0103] Specifically, the community partitioning algorithm is an algorithm used to partition a network or graph into several subgraphs or communities. The purpose of the community partitioning algorithm is to discover subgraphs with high-density connections and close associations in the network, that is, communities. These communities usually have certain functional or structural meanings in the network, such as friend circles in social networks, groups for information dissemination, etc.

[0104] Specifically, commonly used community partitioning algorithms include, for example: the Louvain algorithm, the label propagation algorithm, the genetic algorithm, etc. Among them, the Louvain algorithm: measures the quality of the community structure based on modularity, and maximizes the modularity by iteratively optimizing the community membership of nodes in the network; the label propagation algorithm: each node is initially assigned a unique label, and then the labels are propagated through the relationships between nodes until convergence, forming communities; the genetic algorithm: searches for the optimal community partitioning scheme by simulating the process of biological evolution. Through the iterative optimization process of the genetic algorithm, a better community partitioning scheme can be found, enabling the nodes in the network to be reasonably partitioned into different communities, thereby better understanding the network structure and the relationships between nodes.

[0105] Specifically, this method is illustrated by taking the particle swarm optimization algorithm in the community division algorithm as an example. The basic idea of the particle swarm optimization algorithm is as follows: A group of particles is randomly initialized in a continuous space. Theoretically, the positions of the particles can take any values in the continuous search space. Moreover, each position can be used as a feasible solution to the target problem, and the quality of the solution is obtained by the fitness. The particles fly in the continuous search space to search for better solutions. Each movement is an iteration. The particle reaches a new position by adding the increment of the velocity to the old position. After each iteration is completed, the particle decides whether to update its own optimal solution Pbest according to the fitness of the new position. The particle swarm decides whether to update the global optimal solution Gbest of the particle swarm according to the optimal fitness of the current positions of all particles. Now, assume that the number of particles in the initialized particle swarm is P, and the dimension of the solution space is q. For a certain particle i (1 ≤ i ≤ q), the current velocity of the particle is v i =(v i1 ,v i2 ,v i3 ,…v iq ), the current optimal solution of the particle is pbest i =(pbest i1 ,pbest i2 ,pbest i3 ,…pbest iq ), the current optimal solution of the entire particle swarm is gbest i =(gbest i2 ,gbest i3 ,…gbest iq ). The particle swarm updates the position x i =(x i1 ,x i2 ,x i2 ,…x iq ) of the particle according to these three values. If the dimensions of the particle are too large, the particle will fluctuate severely. Therefore, it needs to be limited within a range [-v max ,v max . The update formulas for the velocity and position of the particle are as follows:

[0106]

[0107] Among them, c1 and c2 are acceleration coefficients, w is the inertia coefficient, r1 and r2 are random numbers between 0 and 1, and v max is the maximum velocity that the particle can reach.

[0108] Further, the steps of the particle swarm optimization algorithm are as follows: Step 1: Read the initialization parameters of the particle swarm, initialize the positions and velocities of each particle in the particle swarm, individual optimal solutions, global optimal solutions, etc.; Step 2: Calculate the fitness of each particle and update its own optimal solution Pbest; Step 3: Update the global optimal solution Gbest; Step 4: According to Update the velocity; Step 5: According to The position update formula updates the position; Step 6: Determine whether the iteration is completed. If the iteration is completed, go to Step 7. If the iteration is not completed, return to Step 2 to recalculate the current particle fitness and its own optimal solution Gbest; Step 7: The module division is completed, and the result is output.

[0109] The first analysis unit 30 performs a global characteristic analysis on the above-mentioned through the customer network to obtain a first analysis characteristic, and performs a local characteristic analysis on the above-mentioned through the customer network to obtain a second analysis characteristic.

[0110] The second analysis unit 40 performs the above-mentioned global characteristic analysis on the above-mentioned churned customer network to obtain a third analysis characteristic, and performs the above-mentioned local characteristic analysis on multiple above-mentioned churned customer network modules to obtain multiple fourth analysis characteristics.

[0111] The calculation unit 50 calculates the similarity coefficient between each of the above-mentioned churned customer network modules and the above-mentioned through the customer network at least according to the above-mentioned first analysis characteristic, the above-mentioned second analysis characteristic, the above-mentioned third analysis characteristic, and multiple above-mentioned fourth analysis characteristics.

[0112] Specifically, compare the analysis characteristic results of the churned customer network module with the analysis characteristic results of the through the customer network to study the similarities and differences therein.

[0113] The determination unit 60 determines whether each of the above-mentioned churned customer network modules is similar to the above-mentioned through the customer network by judging whether the above-mentioned similarity coefficient is within a preset similarity range. In the case where the above-mentioned churned customer network module is similar to the above-mentioned through the customer network, at least adjust the loading speed and response time of the above-mentioned target node of the above-mentioned product process. In the case where the above-mentioned churned customer network module is not similar to the above-mentioned through the customer network, at least simplify the operation steps of the above-mentioned target node of the above-mentioned product process, thereby realizing the customization of the above-mentioned product process.

[0114] Specifically, according to the characteristics of the above customer group, match a similar through the customer network for the churned customer network module. If the similarity is high, at least adjust the loading speed and response time of the target node of the product process. If the similarity is low, at least simplify the operation steps of the target node of the product process to realize the customization of the product process.

[0115] In the above embodiments, first, customer information at each stage of the product process is obtained; second, based on the churn customer information and passing customer information in the above customer information, a churn customer network and a passing customer network are constructed, and the above churn customer network is divided into multiple churn customer network modules by using a community partitioning algorithm; then, a global characteristic analysis and a local characteristic analysis are respectively performed on the passing customer network, a global characteristic analysis is performed on the churn customer network, and a local characteristic analysis is performed on the multiple churn customer network modules. According to the above analysis characteristic results, the similarity coefficient between each churn customer network module and the passing customer network is calculated; finally, it is determined whether each churn customer network module is similar to the passing customer network according to the similarity coefficient. When the churn customer network module is similar to the passing customer network, at least the loading speed and response time of the target node of the product process are adjusted. When the churn customer network module is not similar to the passing customer network, at least the operation steps of the target node of the product process are simplified. Compared with the prior art, the solution of the present application classifies customers into passing customers and churn customers, constructs a churn customer network and a passing customer network, divides the churn customer network into different churn customer network modules according to characteristics by using a community partitioning algorithm, and determines whether the churn customer network module is similar to the passing customer network according to the global characteristic analysis and local characteristic analysis of the customer network, the global characteristic analysis of the churn customer network, and the local characteristic analysis of the churn customer network module. When they are similar, at least the loading speed and response time of the target node of the product process are adjusted. When they are not similar, at least the operation steps of the target node of the product process are simplified, so as to realize the customization of the product process according to the different characteristics of churn customers, that is, the method formulates corresponding processes for different churn customers, meets the personalized needs of different customer groups, and thus solves the problem of poor optimization effect of the product process in the prior art.

[0116] In an embodiment of the present application, the construction unit includes a first mapping subunit, a first calculation subunit, and a first determination subunit. Among them, the first mapping subunit is used to map the above-mentioned multiple churn customer information into multiple first nodes in the initial churn customer network; the first calculation subunit is used to calculate the first weight between two of the above-mentioned first nodes respectively by using the Pearson correlation coefficient, and the above-mentioned first weight is the similarity between the two above-mentioned first nodes; the first determination subunit is used to determine the above-mentioned first nodes to be connected and the above-mentioned first nodes to be deleted by comparing the magnitudes of the above-mentioned multiple first weights and a first preset value, and connect the above-mentioned first nodes to be connected and delete the above-mentioned first nodes to be deleted, thereby obtaining the above-mentioned churn customer network. In this solution, by mapping customer information into nodes in the network and using the Pearson correlation coefficient to calculate the weights between nodes, the correlation and structure between churn customer groups can be revealed more effectively. By comparing the magnitudes of the weights and the preset value, it can be determined which nodes need to be connected and which nodes need to be deleted, so as to construct the churn customer network more accurately, and then adjust the product process more accurately, further improving the optimization effect of the product process.

[0117] Specifically, any two customers or any two churn customers are defined as two variables X and Y, and the Pearson correlation coefficient between X and Y is defined as the quotient of the covariance and standard deviation between the two variables:

[0118]

[0119] Estimating the covariance and standard deviation of the sample, the Pearson correlation coefficient r can be obtained:

[0120]

[0121] Further simplifying r in the above formula and expressing it in terms of the mean of the standard scores of the sample points (X i , Y i ), the following new expression is obtained:

[0122]

[0123] Among them, σ X and are the mean, standard deviation, and standard score of X i respectively.

[0124] In order to more accurately determine the first node to be connected and the first node to be deleted, so as to more accurately construct the churned customer network, in an embodiment of the present application, the first determination subunit includes a first comparison module and a first execution module. Among them, the first comparison module is used for the first comparison step. When the above first weight is greater than the above first preset value, connect the two above first nodes corresponding to the above first weight, and use the above first weight as the weight of the first connection edge between the two above first nodes. The above first connection edge between the above first nodes represents the similarity relationship between the two above first nodes, and the above weight of the above first connection edge between the above first nodes is the degree of similarity between the two above first nodes; the first execution module is used to compare multiple above first weights with the above first preset value, and repeat the above first comparison step at least once until it is determined whether to connect between any two above first nodes, and delete the above first nodes that are not connected to any other above first nodes, so as to obtain the above churned customer network.

[0125] Specifically, the selection of the preset value usually needs to be determined according to specific business requirements, data characteristics, and the goals of subsequent analysis. For example, if the goal is to identify highly similar customer groups, the preset value can be set relatively high; if the goal is to construct a comprehensive network structure to discover potential association patterns, the preset value can be set relatively low.

[0126] Specifically, when the first weight (i.e., the similarity between two first nodes) is greater than the first preset value, it indicates that there is a significant correlation or similar feature between these two churned customers. Therefore, a second connection edge is established in the network and the first weight is given as the weight value of this second connection edge. The magnitude of the weight value represents the degree of similarity or correlation between two nodes. On the contrary, if the first weight is less than or equal to the first preset value, no second connection edge is established, indicating that the correlation between these two churned customers is weak or not significant enough to form an effective connection in the network. After comparing all node pairs, there may be some isolated first nodes in the network, that is, nodes that are not connected to any other nodes. These isolated points may not contain sufficient information, or their churn reasons are significantly different from other customer groups.

[0127] In another embodiment, the construction unit includes a second mapping subunit, a second calculation subunit, and a second determination subunit. Among them, the second mapping subunit is used to map the multiple above-mentioned passing customers into multiple second nodes in the initial passing customer network; the second calculation subunit is used to calculate the second weight between two above-mentioned second nodes respectively by using the Pearson correlation coefficient, and the above-mentioned second weight is the similarity between the two above-mentioned second nodes; the second determination subunit is used to determine the above-mentioned second nodes to be connected and the above-mentioned second nodes to be deleted by comparing the magnitudes of the multiple above-mentioned second weights with a second preset value, and connect the above-mentioned second nodes to be connected and delete the above-mentioned second nodes to be deleted, so as to obtain the above-mentioned passing customer network. In this solution, by mapping customer information into nodes in the network and using the Pearson correlation coefficient to calculate the weights between nodes, the relevance and structure among the passing customer groups can be revealed more effectively. By comparing the magnitudes of the weights and the preset value, it can be determined which nodes need to be connected and which nodes need to be deleted, so as to construct the passing customer network more accurately, and further improve the optimization effect of the product process more effectively.

[0128] In order to more accurately determine the second nodes to be connected and the second nodes to be deleted, so as to more accurately construct the passing customer network, in an embodiment of the present application, the second determination subunit includes a second comparison module and a second execution module. Among them, the second comparison module is used for the second comparison step. When the above-mentioned second weight is greater than the above-mentioned second preset value, connect the two above-mentioned second nodes corresponding to the above-mentioned second weight, and use the above-mentioned second weight as the weight of the second connection edge between the two above-mentioned second nodes. The above-mentioned second connection edge between the above-mentioned second nodes represents the relationship of similarity between the two above-mentioned second nodes, and the above-mentioned weight of the above-mentioned second connection edge between the above-mentioned second nodes is the degree of similarity between the two above-mentioned second nodes; the second execution module is used to compare the multiple above-mentioned second weights with the above-mentioned second preset value, and repeat the above-mentioned second comparison step at least once until it is determined whether to connect between any two above-mentioned second nodes, and at least delete the above-mentioned second nodes without any connection, so as to obtain the above-mentioned passing customer network.

[0129] In still another embodiment, the first analysis unit includes a first conversion subunit, a third calculation subunit, a fourth calculation subunit, a fifth calculation subunit, a sixth calculation subunit, and a seventh calculation subunit. Among them, the first conversion subunit is used to convert the above-mentioned second weight between two above-mentioned second nodes into a first distance between the two above-mentioned second nodes by using a non-linear conversion method; the third calculation subunit is used to calculate according to calculate the first global efficiency GE1, where L 1ijFor the first distance between the second node i and the second node j, N1 is the number of the second nodes passing through the customer network; the fourth calculation subunit is used to calculate the first average shortest path length ASPL1 according to so as to obtain the first analysis feature [GE1, ASPL1]; the fifth calculation subunit is used to calculate the first local efficiency LE1 according to ; the sixth calculation subunit is used to calculate the first clustering coefficient C of the second node according to wherein, E 1i is the number of edges connected by the second nodes adjacent to the second node i, and k 1i is the total number of adjacent nodes of the second node i; the seventh calculation subunit is used to calculate the first node degree D of the customer network according to 1i wherein, a is obtained according to whether there is the second connection edge between the second node i and the second node j. When there is the second connection edge between the second node i and the second node j, a 1i is 1, and when there is no second connection edge between the second node i and the second node j, a 1ij is 0. The second analysis feature is [LE1, C 1ij …C 1ij , D 11i …D 1N1i . In this solution, the first global efficiency, the first average shortest path length, the first local efficiency, the first clustering coefficient, and the first node degree are combined to form the first analysis feature and the second analysis feature, so that the characteristics of the customer group in the customer network can be evaluated more accurately. 1i

[0130] Specifically, the global efficiency describes the network from the perspective of the overall information flow, can better reflect the processing ability and transmission ability of the network global information, and can also reflect the integration degree of the network; the average shortest path length represents the ability of information transmission between network nodes, and can better reflect the functional integration level between regions. The lower this value is, the higher the functional integration level between regions is; the local efficiency is usually used to measure the separation degree of the network, and can better reflect the processing ability and transmission ability of local information; the clustering coefficient is usually used to measure the aggregation degree of nodes, and is positively correlated with the aggregation degree; the node degree is the sum of the out-degree and in-degree of nodes in the network, is related to the complexity of the network, and is mostly used in the research of network modules.

[0131] Specifically, by combining the first global efficiency, the first average shortest path length, the first local efficiency, the first clustering coefficient, and the first node degree to form the first analysis feature and the second analysis feature, the characteristics of the customer group can be evaluated more comprehensively. This method of combined feature analysis provides an in-depth understanding of the behavior and association patterns of customer groups in the customer network from both the global and local levels, laying a solid data foundation for subsequent customer group comparison analysis and process customization.

[0132] In order to more accurately evaluate the characteristics of customer groups in the churned customer network module, so as to more accurately adjust the product process and then more effectively improve the optimization effect of the product process, in an embodiment of the present application, the second analysis unit includes a second conversion subunit, an eighth calculation subunit, a ninth calculation subunit, a tenth calculation subunit, an eleventh calculation subunit, and a twelfth calculation subunit. Among them, the second conversion subunit is used to convert the first weight between the two first nodes into a second distance between the two first nodes by using a non-linear conversion method; the eighth calculation subunit is used to calculate the second global efficiency GE2 according to where L 2mn is the second distance between the first node m and the first node n, and N2 is the number of the first nodes in the churned customer network; the ninth calculation subunit is used to calculate the second average shortest path length ASPL2 according to and the third analysis feature is [GE2, ASPL2]; the tenth calculation subunit is used to calculate the second local efficiency LE2 according to ; the eleventh calculation subunit is used to calculate the second clustering coefficient C of each first node in each churned customer network module according to where E 2m is the number of edges connected by the first nodes adjacent to the first node m, and k 2m is the total number of adjacent nodes of the first node m; the twelfth calculation subunit is used to calculate the second node degree D of each churned customer network module according to 2m where a is obtained according to whether there is the first connection edge between the first node m and the first node n. When there is the first connection edge between the first node m and the first node n, a 2m is 1, and when there is no first connection edge between the first node m and the first node n, a 2mn is 0. The multiple fourth analysis features are [LE2, C 2mn , D 2mn …[LE2, C 21m , D 21m …[LE2, C 2Mm , D 2Mm, where M is the number of the above-mentioned lost customer network modules.

[0133] In yet another embodiment, the calculation unit includes a first vector fusion subunit, a first normalization subunit, a second vector fusion subunit, a second normalization subunit, and a thirteenth calculation subunit. Among them, the first vector fusion subunit is used to perform vector fusion on the above-mentioned first analysis feature and the above-mentioned second analysis feature to obtain a first feature vector; the first normalization subunit is used to normalize the above-mentioned first feature vector to obtain a second feature vector; the second vector fusion subunit is used to perform vector fusion on the above-mentioned third analysis feature and multiple above-mentioned fourth analysis features respectively to obtain multiple third feature vectors; the second normalization subunit is used to normalize multiple above-mentioned third feature vectors respectively to obtain multiple fourth feature vectors; the thirteenth calculation subunit is used to calculate the similarity between multiple above-mentioned fourth feature vectors and the above-mentioned second feature vector by using the KNN algorithm to obtain multiple above-mentioned similarity coefficients between the above-mentioned lost customer network modules and the above-mentioned passed customer network. In this solution, by performing feature analysis on the lost customer network module and the passed customer network, and fusing and normalizing different feature vectors, and finally calculating the similarity coefficient by using the KNN algorithm, the similarity degree between the lost customer network module and the passed customer network can be understood more accurately, and then the product process can be adjusted more accurately, and the optimization effect of the product process can be further improved.

[0134] Specifically, the steps of calculating the similarity coefficient by using KNN: Select a suitable distance metric method to quantify the distance between two vectors. Commonly used ones include Euclidean distance, Manhattan distance, Minkowski distance, or cosine similarity, etc.; Calculate the similarity coefficient. Based on the KNN algorithm, define a similarity coefficient to evaluate the similarity between two vectors. In this method, the distance metric is transformed into a similarity metric. For example, the distance value is transformed into its reciprocal or a certain exponential function is used to weaken the influence of the distance.

[0135] In order to more accurately determine whether the lost customer network module is similar to the passed customer network, so as to more accurately adjust the product process and further improve the optimization effect of the product process. In an embodiment of the present application, the determination unit includes a third determination subunit and a fourth determination subunit. Among them, the third determination subunit is used to determine that the above-mentioned lost customer network module is similar to the above-mentioned passed customer network when the above-mentioned similarity coefficient is within the above-mentioned preset similarity range; the fourth determination subunit is used to determine that the above-mentioned lost customer network module is not similar to the above-mentioned passed customer when the above-mentioned similarity coefficient is not within the above-mentioned preset similarity range.

[0136] In order to more accurately adjust the product process when the lost customer network module is similar to the passed customer network, so as to further improve the optimization effect of the product process, in an embodiment of the present application, the determination unit includes a first adjustment subunit, where the first adjustment subunit is used to, when the lost customer network module is similar to the passed customer network, at least use asynchronous loading technology to enable the user not to load the content of other nodes of the product process when browsing the target node of the product process, so as to adjust the loading speed and the response time of the target node of the product process.

[0137] Specifically, in web development, asynchronous loading technology allows a page to load only the specified parts when loading, while other parts are postponed to be loaded under specific conditions. For example, using JavaScript's AJAX (Asynchronous JavaScript and XML) or related frameworks (such as React, Vue, etc.), asynchronous loading of specific nodes can be achieved. When a customer accesses the target node, only the data and elements related to that node are loaded, while the elements of other nodes remain in a lazy loading state and are loaded only when they are actually needed. This can significantly reduce the initial loading time and improve the response speed.

[0138] Specifically, the adjustment process of asynchronous loading technology is as follows: resource grouping, according to the analysis results of the lost customer network and the passed customer network similarity module, the nodes in the product process are divided into multiple groups, and each group contains nodes with similar characteristics; loading strategy customization, different asynchronous loading strategies are formulated for each group of nodes. For the target node, depth-first asynchronous loading is implemented, that is, the target node is loaded first to ensure fast response. For non-target nodes, deferred loading or on-demand loading is adopted to reduce unnecessary resource loading; optimizing the front-end code, by optimizing the front-end HTML, CSS, and JavaScript code, ensuring that the elements and resources of the target node can be quickly recognized and loaded. This may include using preloading, prefetching, and resource lazy loading technologies; back-end optimization, in cooperation with front-end asynchronous loading, the back-end needs to provide support, including on-demand data provision (that is, only return data related to the target node when requested), data compression, and caching strategies to further improve the loading speed and response time; testing and adjustment, after implementing asynchronous loading, strict testing is required, including user behavior testing, performance testing, and compatibility testing, to ensure that the optimized process not only improves the loading speed and response time but also does not affect the user experience and the integrity of the process. According to the test results, it may be necessary to adjust the parameters and strategies of asynchronous loading to achieve the best effect.

[0139] In order to more accurately adjust the product process when the lost customer network module is not similar to the customer network passed through, so as to further improve the optimization effect of the product process, in an embodiment of the present application, the determination unit includes a second adjustment subunit. Among them, the second adjustment subunit is used to at least simplify the form tax design of the target node and reduce the required fields of the form when the lost customer network module is not similar to the customer network passed through, so as to simplify the operation steps of the target node of the product process.

[0140] Specifically, the form design optimization mainly targets the target node (for example, the product application form), and redesigns the form structure to make it more intuitive and easy to fill out. This may include reducing the required fields of the form, splitting the long form into multiple short steps, using the pre-fill function, or providing more friendly user interface prompts.

[0141] Specifically, the required field evaluation is to re-evaluate the necessity of the required fields in the form. Consider whether some items can be changed to optional, or whether they can be pre-filled in other ways (such as user account information), so as to reduce the burden on users when filling out the form.

[0142] The adjustment device of the above product process includes a processor and a memory. The above acquisition unit, construction unit, first analysis unit, second analysis unit, calculation unit, and determination unit are all stored in the memory as program units, and the processor executes the above program units stored in the memory to implement corresponding functions. The above modules are all located in the same processor; or, the above modules are respectively located in different processors in any combination form.

[0143] The processor contains a kernel, and the kernel retrieves the corresponding program unit from the memory. One or more kernels can be set, and the problem of poor optimization effect of the product process in the prior art is solved by adjusting the kernel parameters.

[0144] Figure 5 It is a flowchart of the application method of the product process according to the embodiment of the present application. As Figure 5 shown, the method includes the following steps:

[0145] Step S501, adjust the product process by using any one of the above product process adjustment methods to obtain a target product process.

[0146] Step S502, push the above target product process to the target customer, and the above target customer includes at least the lost customers of the above product process.

[0147] Specifically, implement a refined push strategy, and push the target product process to lost customers or customers with high value to the enterprise through multiple channels such as email, text message, and APP push notification.

[0148] The memory may include non-permanent memory in a computer-readable medium, in the form of random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one storage chip.

[0149] Embodiments of the present invention provide a computer-readable storage medium, and the computer-readable storage medium includes a stored program, wherein when the program runs, it controls the device where the computer-readable storage medium is located to execute the adjustment method of the above product process.

[0150] Embodiments of the present invention provide a processor, and the processor is used to run a program, wherein when the program runs, it executes the adjustment method of the above product process.

[0151] Embodiments of the present invention provide a device, the device includes a processor, a memory, and a program stored on the memory and executable on the processor, and when the processor executes the program, it implements at least the steps in the above method.

[0152] The device herein may be a server, a PC, a PAD, a mobile phone, etc.

[0153] The present application also provides a computer program product, which is suitable for executing the initialization with at least the steps in the above method when executed on a data processing device.

[0154] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general-purpose computing device. They can be concentrated on a single computing device or distributed on a network composed of multiple computing devices. They can be implemented by program codes executable by the computing device, so that they can be stored in a storage device and executed by the computing device. And in some cases, the steps shown or described can be executed in a different order from here, or they can be separately made into individual integrated circuit modules, or multiple modules or steps among them can be made into a single integrated circuit module to implement. Thus, the present invention is not limited to any specific combination of hardware and software.

[0155] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0156] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, and combinations of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processors of general-purpose computers, special-purpose computers, embedded processors, or other programmable data processing devices to generate a machine, such that the instructions executed by the processors of the computer or other programmable data processing devices produce means for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0157] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory produce a manufacture including instruction means that implement the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0158] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operational steps are executed on the computer or other programmable device to generate a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0159] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and memory.

[0160] The memory may include non-permanent memory in the form of computer-readable media, random access memory (RAM), and / or non-volatile memory such as read-only memory (ROM) or flash RAM. The memory is an example of computer-readable media.

[0161] A computer-readable medium includes permanent and non-permanent, removable and non-removable media that can implement information storage by any method or technology. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette tapes or other magnetic storage devices, or any other non-transitory media that can be used to store information that can be accessed by a computing device. As defined herein, a computer-readable medium does not include transitory computer-readable media, such as modulated data signals and carrier waves.

[0162] It should also be noted that the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or apparatus comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or apparatus. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or apparatus comprising the element.

[0163] From the above description, it can be seen that the above embodiments of the present application achieve the following technical effects:

[0164] In the above embodiments, first, customer information at each stage of the product process is obtained; second, based on the churn customer information and passing customer information in the above customer information, a churn customer network and a passing customer network are constructed, and the churn customer network is divided into multiple churn customer network modules by using a community partitioning algorithm; then, a global feature analysis and a local feature analysis are respectively performed on the passing customer network, a global feature analysis is performed on the churn customer network, and a local feature analysis is performed on the multiple churn customer network modules. According to the above analysis feature results, the similarity coefficient between each churn customer network module and the passing customer network is calculated; finally, it is determined whether each churn customer network module is similar to the passing customer network according to the similarity coefficient. When the churn customer network module is similar to the passing customer network, at least the loading speed and response time of the target node of the product process are adjusted. When the churn customer network module is not similar to the passing customer network, at least the operation steps of the target node of the product process are simplified. Compared with the prior art, the solution of the present application classifies customers into passing customers and churn customers, constructs a churn customer network and a passing customer network, divides the churn customer network into different churn customer network modules according to characteristics by using a community partitioning algorithm, and determines whether the churn customer network module is similar to the passing customer network according to the global feature analysis and local feature analysis of the customer network, the global feature analysis of the churn customer network, and the local feature analysis of the churn customer network module. When they are similar, at least the loading speed and response time of the target node of the product process are adjusted. When they are not similar, at least the operation steps of the target node of the product process are simplified, so as to achieve customization of the product process according to different characteristics of churn customers, that is, the method formulates corresponding processes for different churn customers, meets the personalized needs of different customer groups, and thus solves the problem of poor optimization effect of the product process in the prior art.

[0165] The foregoing is only a preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for adjusting a product process, characterized in that, Including: Obtain customer information at each stage of the product process. Among them, the customer information includes churn customer information and passed customer information. The churn customer information is the information of customers who have not completed the product process of the target node. The passed customer information is the information of the customers who have completed the product process of the target node. The product process is at least composed of a user registration or login node, a data input node, a product selection node, and a transaction completion node. The target node is any node in the product process. According to the customer information, construct a churn customer network and a passed customer network, and use a community division algorithm to divide the churn customer network to obtain multiple churn customer network modules. Conduct a global characteristic analysis on the passed customer network to obtain a first analysis characteristic, and conduct a local characteristic analysis on the passed customer network to obtain a second analysis characteristic. Conduct the global characteristic analysis on the churn customer network to obtain a third analysis characteristic, and conduct the local characteristic analysis on multiple churn customer network modules to obtain multiple fourth analysis characteristics. Calculate the similarity coefficient between each churn customer network module and the passed customer network at least according to the first analysis characteristic, the second analysis characteristic, the third analysis characteristic, and multiple fourth analysis characteristics. By determining whether the similarity coefficient is within a preset similarity range, determine whether each churn customer network module is similar to the passed customer network. When the churn customer network module is similar to the passed customer network, at least adjust the loading speed and response time of the target node of the product process. When the churn customer network module is not similar to the passed customer network, at least simplify the operation steps of the target node of the product process, so as to realize the customization of the product process.

2. The method for adjusting the product process according to claim 1, wherein According to the customer information, construct a churn customer network, including: Map multiple pieces of the churn customer information to multiple first nodes in the initial churn customer network. Use the Pearson correlation coefficient to calculate the first weight between two of the first nodes respectively. The first weight is the similarity between the two first nodes. By comparing the magnitude relationship between multiple first weights and a first preset value, determine the first nodes to be connected and the first nodes to be deleted, and connect the first nodes to be connected and delete the first nodes to be deleted, so as to obtain the churn customer network.

3. The method for adjusting the product process according to claim 2, wherein By comparing the magnitude relationship between multiple first weights and a first preset value, determine the first nodes to be connected and the first nodes to be deleted, and connect the first nodes to be connected and delete the first nodes to be deleted, so as to obtain the churn customer network, including: The first comparison step: when the first weight is greater than the first preset value, connect the two first nodes corresponding to the first weight, and use the first weight as the weight of the first connection edge between the two first nodes. The first connection edge between the first nodes represents the similarity relationship between the two first nodes, and the weight of the first connection edge between the first nodes is the degree of similarity between the two first nodes. Compare multiple first weights with the first preset value, and repeat the first comparison step at least once until it is determined whether to connect between any two first nodes. Delete the first nodes that are not connected to any other first nodes, thereby obtaining the churned customer network.

4. The method for adjusting the product process according to claim 1, characterized in that Construct a passed customer network according to the customer information, including: Map multiple pieces of passed customer information to multiple second nodes in the initial passed customer network; Use the Pearson correlation coefficient to calculate the second weight between two second nodes respectively. The second weight is the similarity between the two second nodes; By comparing the magnitude relationship between multiple second weights and a second preset value, determine the second nodes to be connected and the second nodes to be deleted, and connect the second nodes to be connected and delete the second nodes to be deleted, thereby obtaining the passed customer network.

5. The method for adjusting the product process according to claim 4, wherein, By comparing the magnitude relationship between multiple second weights and a second preset value, determine the second nodes to be connected and the second nodes to be deleted, and connect the second nodes to be connected and delete the second nodes to be deleted, thereby obtaining the passed customer network, including: The second comparison step: when the second weight is greater than the second preset value, connect the two second nodes corresponding to the second weight, and use the second weight as the weight of the second connection edge between the two second nodes. The second connection edge between the second nodes represents the similarity relationship between the two second nodes, and the weight of the second connection edge between the second nodes is the degree of similarity between the two second nodes; Compare multiple second weights with the second preset value, and repeat the second comparison step at least once until it is determined whether to connect between any two second nodes. Delete the second nodes that are not connected to any other second nodes, thereby obtaining the passed customer network.

6. The method for adjusting the product process according to claim 5, characterized in that Conduct global characteristic analysis and local characteristic analysis on the passed customer network to obtain a first analysis characteristic and a second analysis characteristic, including: Use a non-linear transformation method to transform the second weight between two second nodes into a first distance between the two second nodes; According to calculate the first global efficiency GE1, where L 1ij is the first distance between the second node i and the second node j, and N1 is the number of the second nodes passing through the customer network; According to calculate the first average shortest path length ASPL1, so as to obtain the first analysis feature [GE1, ASPL1]; According to calculate the first partial efficiency LE1; According to calculate the first clustering coefficient C of the second node 1i , where E 1i is the number of edges connected by the second nodes adjacent to the second node i, and k 1i is the total number of adjacent nodes of the second node i; According to calculate the first node degree D of the customer network 1i , where a 1ij is obtained based on whether there is the second connection edge between the second node i and the second node j. When there is the second connection edge between the second node i and the second node j, a 1ij is 1. When there is no second connection edge between the second node i and the second node j, a 1ij is 0. The second analysis feature is [LE1, C 11i …C 1N1i , D 1i .

7. The method for adjusting the product process according to claim 3, wherein Conduct the global characteristic analysis on the churned customer network and conduct the local characteristic analysis on multiple churned customer network modules to obtain a third analysis characteristic and multiple fourth analysis characteristics, including: Use a non-linear transformation method to transform the first weight between two first nodes into a second distance between the two first nodes; According to calculate the second global efficiency GE2, where L 2mn is the second distance between the first node m and the first node n, and N2 is the number of the first nodes in the churned customer network; According to calculate the second average shortest path length ASPL2, and the third analysis feature is [GE2, ASPL2]; According to calculate the second partial efficiency LE2; According to calculate the second clustering coefficient C of the first node in each of the churn customer network modules 2m , where E 2m is the number of edges connected by the first nodes adjacent to the first node m, and k 2m is the total number of adjacent nodes of the first node m; According to calculate the second node degree D of each of the churn customer network modules 2m , where a 2mn is obtained based on whether there is the first connection edge between the first node m and the first node n. When there is the first connection edge between the first node m and the first node n, a 2mn is 1. When there is no first connection edge between the first node m and the first node n, a 2mn is 0. The multiple fourth analysis features are [LE2, C 21m , D 21m …[LE2, C 2Mm , D 2Mm , where M is the number of the churn customer network modules.

8. The method for adjusting the product process according to claim 1, characterized in that, Calculate the similarity coefficient between each of the churned customer network modules and the passed customer network based on at least the first analysis characteristic, the second analysis characteristic, the third analysis characteristic, and multiple fourth analysis characteristics, including: Perform vector fusion on the first analysis characteristic and the second analysis characteristic to obtain a first characteristic vector; Normalize the first characteristic vector to obtain a second characteristic vector; Perform vector fusion on the third analysis characteristic and multiple fourth analysis characteristics respectively to obtain multiple third characteristic vectors; Normalize multiple third characteristic vectors respectively to obtain multiple fourth characteristic vectors; Use the KNN algorithm to calculate the similarity between multiple fourth characteristic vectors and the second characteristic vector respectively, and obtain the similarity coefficient between each of the churned customer network modules and the passed customer network.

9. The method for adjusting the product process according to claim 1, characterized in that Determine whether each of the churned customer network modules is similar to the passed customer network by determining whether the similarity coefficient is within a preset similarity range, including: When the similarity coefficient is within the preset similarity range, determine that the churned customer network module is similar to the passed customer network; When the similarity coefficient is not within the preset similarity range, determine that the churned customer network module is not similar to the passed customer.

10. The method for adjusting the product process according to claim 1, wherein, When the churned customer network module is similar to the passed customer network, at least adjust the loading speed and response time of the target node of the product process, including: When the churned customer network module is similar to the passed customer network, at least use asynchronous loading technology so that when the user browses the target node of the product process, the content of other nodes of the product process is not loaded, thereby adjusting the loading speed and response time of the target node of the product process.

11. The method for adjusting the product process according to claim 1, wherein When the churned customer network module is not similar to the passed customer network, at least simplify the operation steps of the target node of the product process, including: When the churned customer network module is not similar to the passed customer network, at least simplify the form tax design of the target node and reduce the required items in the form, thereby simplifying the operation steps of the target node of the product process.

12. An application method for a product process, characterized in that, Including: Adjust the product process using the product process adjustment method according to any one of claims 1 to 11 to obtain a target product process; Push the target product process to target customers, where the target customers at least include the churned customers of the product process.

13. An adjustment device for a product process, characterized in that, Including: An acquisition unit that acquires customer information at each stage of the product process, where the customer information includes churned customer information and passed customer information, the churned customer information is the information of customers who have not completed the product process of the target node, the passed customer information is the information of the customers who have completed the product process of the target node, the product process is at least composed of a user registration or login node, a data input node, a product selection node, and a transaction completion node, and the target node is any node in the product process; A construction unit, which constructs a churned customer network and a passing customer network according to the customer information, and divides the churned customer network by using a community division algorithm to obtain a plurality of churned customer network modules; A first analysis unit, which performs a global characteristic analysis on the passing customer network to obtain a first analysis characteristic, and performs a local characteristic analysis on the passing customer network to obtain a second analysis characteristic; A second analysis unit, which performs the global characteristic analysis on the churned customer network to obtain a third analysis characteristic, and performs the local characteristic analysis on the plurality of churned customer network modules to obtain a plurality of fourth analysis characteristics; A calculation unit, which calculates a similarity coefficient between each of the churned customer network modules and the passing customer network at least according to the first analysis characteristic, the second analysis characteristic, the third analysis characteristic, and the plurality of fourth analysis characteristics; A determination unit, which determines whether each of the churned customer network modules is similar to the passing customer network by determining whether the similarity coefficient is within a preset similarity range. When a churned customer network module is similar to the passing customer network, at least the loading speed and response time of the target node of the product process are adjusted. When a churned customer network module is not similar to the passing customer network, at least the operation steps of the target node of the product process are simplified, so as to implement the customization of the product process.

14. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein when the program runs, it controls the device where the computer-readable storage medium is located to execute the product process adjustment method according to any one of claims 1 to 11.

15. An electronic device, characterized in that, Comprising: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include a method for executing the product process adjustment method according to any one of claims 1 to 11.

Images