Multi-touch-channel private domain user management method and system, medium and program product

Abstract

The invention discloses a private domain user management method and system for multiple touch channels, a medium and a program product, and the method comprises the steps: obtaining the historical behavior data of a user in each touch channel; identifying the behavior relevance of the user among the arriving channels; constructing an inter-channel behavior mapping relation of migration paths of the user among the touch channels based on behavior relevance; according to the behavior mapping relation between the channels, the overlapping ratio of user behaviors between the reaching channels, the difference of conversion effects and the overlapping ratio of time sequences are calculated; based on the overlapping rate, the difference and the coincidence degree, determining the reaching channel complementation degree; grouping and clustering the reaching channels according to the reaching channel complementation degree to obtain a plurality of channel collaboration units; and the optimal time difference of the arriving users in each channel cooperation unit is calculated, and the marketing resources are allocated among the arriving channels based on the optimal time difference, so that the accuracy of private domain operation management is improved.

Description

Technical Field

[0001] The present application relates to the field of Internet services, and in particular to a method, system, medium and program product for managing private domain users with multiple touch channels. Background Art

[0002] With the rise of private domain traffic operations, companies are paying more and more attention to establishing deep connections with users through multiple touch channels to improve user stickiness and marketing effectiveness. However, in the process of managing private domain users with multiple touch channels, due to the dispersion of user data on different channel platforms, user portraits are fragmented, marketing strategies are fragmented, and operational efficiency is low, making it difficult to achieve unified management and accurate reach of users.

[0003] In related technologies, a unified user data platform can be established to integrate user data from various channels, build a complete user portrait, and formulate differentiated marketing strategies based on the characteristics of the user portrait. This technology uses data labeling management to classify and organize user behavior data, transaction data, and social data, achieving unified management of user data and providing data support for precision marketing.

[0004] However, users may have changing preferences between different reach channels. Due to the lack of consideration of user channel migration behavior, it is difficult to determine the user's active trends in each reach channel, resulting in marketing resources being unable to reach users through the optimal channel at the best time, thereby reducing the accuracy of private domain management. Summary of the invention

[0005] The present application provides a multi-touch channel private domain user management method, system, medium and program product for improving the accuracy of private domain operation management.

[0006] In the first aspect, the present application provides a method for managing private domain users in multiple contact channels, which obtains historical behavior data of users in each contact channel, and the historical behavior data includes click records, browsing records, interactive operation records and transaction records; Perform two-way cross-validation on historical behavior data to identify the correlation between user behaviors across various contact channels; Based on behavioral relevance, build a cross-channel behavioral mapping relationship of the user's migration path between various contact channels; Based on the mapping relationship between channel behaviors, the overlap rate of user behaviors, the difference in conversion effects, and the overlap of time series between different channels are calculated; Determine the degree of complementarity of reach channels based on overlap, difference and coincidence; Group and cluster the various reach channels according to their complementarity levels to obtain a number of channel synergy units, where the reach channels in the channel synergy units are the reach channels whose corresponding complementarity levels are greater than a preset threshold; Calculate the optimal time difference for reaching users within each channel collaboration unit, and allocate marketing resources among various channels based on the optimal time difference.

[0007] By adopting the above technical solution, obtaining the historical behavior data of users in each contact channel and performing two-way cross-validation, the accuracy of identifying the behavioral correlation between users in different channels can be improved, and then the migration path mapping relationship between users in various contact channels can be constructed. By clustering channels based on the degree of complementarity as the basis for grouping, contact channels with strong complementarity can be grouped into channel coordination units to reduce the duplication of channel resources. By allocating marketing resources by calculating the optimal contact time difference within the coordination unit, it not only ensures the coordination of each contact channel in the time dimension, but also can reasonably allocate resources according to the characteristics of the channel, thereby improving the overall resource utilization efficiency. It realizes the complementary advantages and optimal resource allocation among multiple contact channels, and improves the accuracy of private domain operation management.

[0008] In conjunction with some embodiments of the first aspect, in some embodiments, calculating the optimal time difference for reaching users in each channel collaboration unit, and allocating marketing resources among the reaching channels based on the optimal time difference, specifically includes: Determine the customer reach time distribution of each reach channel based on the historical reach data of each reach channel in each channel collaboration unit, and obtain the high-frequency reach time period of each reach channel; Calculate the center of gravity of the high-frequency reach time period, and define the time difference between the centers of gravity of each reach channel as the optimal time difference; Based on the total amount of marketing resources of each channel collaboration unit, the importance coefficient of each reach channel is allocated to obtain the initial marketing resource volume of each reach channel. The importance coefficient is determined based on the user scale, user value and conversion effect of each reach channel.

[0009] By adopting the above technical solution and analyzing the historical reach data of the reach channels in each channel collaboration unit, the high-frequency reach time period of each reach channel can be identified, and the optimal time difference can be determined by calculating the center of gravity of these high-frequency time periods. It can objectively reflect the active patterns of users in different channels and reduce the deviation caused by subjective experience judgment. On this basis, the importance coefficient is determined according to key indicators such as the user scale, user value and conversion effect of the reach channel, and marketing resources are allocated accordingly, so that resource allocation is in line with the user's reach habits and matches the actual value of the channel, improving reach efficiency and reducing resource waste. At the same time, it ensures that each reach channel can reach users at the most appropriate time point, improving user response rate and conversion effect.

[0010] In combination with some embodiments of the first aspect, in some embodiments, after determining the degree of complementarity of the reach channels based on the overlap rate, the difference, and the coincidence, the method further includes: Determine the channel influence index based on the complementary degree of reach channels; Represent the migration path with a low-dimensional vector to obtain the migration behavior vector; Input the channel influence index and migration behavior vector into the preset deep neural network model to obtain the churn risk of each reach channel within the preset time period; For channel collaboration units whose churn risk is greater than a preset risk threshold, a churn warning model is generated through a transfer learning method. The churn warning model is used to issue a preset warning message when a user whose churn risk is greater than a preset risk threshold is identified.

[0011] By adopting the above technical solution, the channel influence index is determined based on the complementary degree of the reach channels, and the user migration path is converted into a low-dimensional vector representation, combined with the deep neural network model for analysis, which can improve the accuracy of predicting the user churn risk of each reach channel within a preset time period. For high-risk channel collaboration units, the churn warning model constructed by the transfer learning method has a stronger generalization ability, can quickly identify potential churn users and issue warning information in a timely manner, and can take intervention measures in advance to reduce the user churn rate.

[0012] In conjunction with some embodiments of the first aspect, in some embodiments, before calculating the optimal time difference for reaching the user in each channel collaboration unit, the method further includes: Obtain environmental noise data for each reach channel, including system failure records, network delay data, and user complaint records; A time window correction function is constructed based on the environmental noise data. The time window correction function is used to correct the time series data of each reach channel. Calibrate the historical reach time of each reach channel according to the time window correction function; The effective reach rate of each reach channel is calculated based on the calibrated reach time, and the effective reach rate is used as the weight factor for calculating the optimal time difference.

[0013] By adopting the above technical solution, collecting and analyzing the environmental noise data of each reach channel, and constructing a time window correction function to correct the time series data, the interference of external factors such as system failures and network delays on the reach effect can be reduced. The calibrated historical reach time more accurately reflects the real interaction between users and channels. The effective reach rate calculated can be used as a reliable weight factor to participate in the calculation of the optimal time difference. The accuracy of the time difference calculation is improved, making the subsequent resource allocation more reasonable.

[0014] In conjunction with some embodiments of the first aspect, in some embodiments, constructing a time window correction function specifically includes: Calculate the noise intensity coefficient of each reach channel in different time periods; Determine the dynamic expansion range of the time window according to the noise intensity coefficient, and construct an adaptive time window based on the dynamic expansion range; The historical reach data is segmented based on the adaptive time window to obtain the abnormal fluctuation range; The data in the abnormal fluctuation range is smoothed to obtain the time window correction coefficient.

[0015] By adopting the above technical solution and calculating the noise intensity coefficient of each reach channel in different time periods, the degree of influence of environmental noise on the reach effect can be quantitatively evaluated. Based on the noise intensity coefficient, the dynamic expansion range of the time window is determined and an adaptive time window is constructed, so that the time window can automatically adjust its size according to the actual noise situation, thereby enhancing the adaptability to noise interference. The historical reach data is segmented to obtain the abnormal fluctuation interval, which can accurately locate the time period that is most affected by noise. By smoothing the data in the abnormal fluctuation interval to obtain the time window correction coefficient, the impact of environmental noise on the accuracy of the reach time can be reduced, the reliability of the reach data can be improved, and the dynamic calibration of the reach time data can be realized, so that the reach time more accurately reflects the actual reception time of the user, thereby improving the scientificity and effectiveness of the subsequent reach strategy formulation.

[0016] In combination with some embodiments of the first aspect, in some embodiments, after calculating the effective reach rate of each reach channel, the method further includes: Build a reach effect evaluation index system, which includes user responsiveness and conversion timeliness; Calculate the effect scores of each reach channel in different time periods based on the reach effect evaluation index system, and establish the corresponding relationship between the effect score and the reach time; Generate reach time optimization suggestions based on the corresponding relationship.

[0017] By adopting the above technical solutions, a reach effect evaluation index system that includes user responsiveness and conversion timeliness is constructed, realizing a comprehensive quantitative evaluation of the reach effect. Based on this evaluation index system, the effect scores of each reach channel in different time periods are calculated, and the corresponding relationship between the effect score and the reach time is established, so that the system can accurately evaluate the impact of different reach times on the reach effect. The reach time optimization suggestions generated based on this correspondence can guide operators to choose the best reach time, improve the reach effect, and improve the efficiency of reach resource utilization.

[0018] In conjunction with some embodiments of the first aspect, in some embodiments, generating a reach time optimization suggestion according to the corresponding relationship specifically includes: Determine the periodic change characteristics of the effect score based on the time series data of the effect score of each reach channel; According to the periodic change characteristics, each time period is divided into an advantageous period, a stable period and an inferior period according to the effect score, and the duration of each period is calculated; Identify the complementary relationship between the reach time of various reach channels in the same time period; Calculate the optimal reach time interval for each reach channel based on the complementary relationship of reach time, and generate channel collaborative reach suggestions; Develop a plan to adjust the reach time based on channel collaborative reach recommendations and the operating costs and resource constraints of each reach channel; Convert the reach time adjustment plan into reach time optimization suggestion.

[0019] By adopting the above technical solution, the periodic change characteristics of the effect score are analyzed, the time periods are divided into advantageous periods, stable periods and disadvantageous periods, and the duration of each period is calculated, so as to accurately grasp the time distribution law of the reach effect. The complementary relationship between the reach time between the reach channels is identified, the optimal reach time interval is calculated, and channel collaborative reach suggestions are generated, so that multiple reach channels can cooperate and reduce the interference caused by the overlap of reach time, and the reach time is accurately optimized. While improving the reach effect, the economy and feasibility of the plan are guaranteed, and the reach cost is reduced.

[0020] In the second aspect, an embodiment of the present application provides a private domain user management system for multiple touch channels, which private domain user management system for multiple touch channels includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and one or more processors call the computer instructions to enable the system to execute the method described in the first aspect and any possible implementation method of the first aspect.

[0021] In a third aspect, an embodiment of the present application provides a computer-readable storage medium, comprising instructions, which, when executed on a system, causes the system to execute the method described in the first aspect and any possible implementation of the first aspect.

[0022] In a fourth aspect, an embodiment of the present application provides a computer program product, characterized in that when the computer program product runs on a system, the system executes the method described in any possible implementation manner in the first aspect.

[0023] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. This application provides a method for managing private domain users with multiple reach channels, which obtains the historical behavior data of users in each reach channel and performs two-way cross-validation, which can improve the accuracy of identifying the behavioral correlation between users in different channels, and then construct a mapping relationship between the migration paths of users between the reach channels. Channel clustering is performed based on the degree of complementarity as the basis for grouping, and reach channels with strong complementarity can be grouped into channel synergy units to reduce the duplication of channel resources. Marketing resources are allocated by calculating the optimal reach time difference within the synergy unit, which not only ensures the synergy of each reach channel in the time dimension, but also can reasonably allocate resources according to the characteristics of the channel, thereby improving the overall resource utilization efficiency. The complementary advantages and optimal resource allocation between multiple reach channels are realized, and the accuracy of private domain operation management is improved.

[0024] 2. This application provides a method for managing private domain users with multiple reach channels, collects and analyzes the environmental noise data of each reach channel, and constructs a time window correction function to correct the time series data, which can reduce the interference of external factors such as system failures and network delays on the reach effect. The calibrated historical reach time more accurately reflects the actual interaction between the user and the channel. The effective reach rate calculated in this way can be used as a reliable weight factor to participate in the calculation of the optimal time difference. The accuracy of the time difference calculation is improved, making the subsequent resource allocation more reasonable.

[0025] 3. This application provides a private domain user management method for multiple reach channels, constructs a reach effect evaluation index system that includes user responsiveness and conversion timeliness, and realizes a comprehensive quantitative evaluation of the reach effect. Based on this evaluation index system, the effect scores of each reach channel in different time periods are calculated, and a corresponding relationship between the effect score and the reach time is established, so that the system can accurately evaluate the impact of different reach times on the reach effect. The reach time optimization suggestions generated based on this correspondence can guide operators to choose the best reach time, improve the reach effect, and improve the efficiency of reach resource utilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a flow chart of a private domain user management method with multiple touch channels in an embodiment of the present application.

[0027] Figure 2 This is another flow chart of a method for managing private domain users through multiple touch channels in an embodiment of the present application.

[0028] Figure 3 It is a schematic diagram of the physical device structure of a private domain user management system with multiple reach channels provided in an embodiment of the present application; Figure 4 This is a functional page display diagram of a private domain user management system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0029] The terms used in the following embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to be used as limitations to the present application. As used in the specification and appended claims of the present application, the singular expressions "one", "a kind of", "said", "above", "the" and "this" are intended to also include plural expressions, unless there is a clear indication to the contrary in the context. It should also be understood that the term "and / or" used in the present application refers to any or all possible combinations comprising one or more listed items.

[0030] In the following, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as suggesting or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features, and in the description of the embodiments of the present application, unless otherwise specified, "plurality" means two or more.

[0031] It should be noted that a multi-touch channel private domain user management method of the present application is applied to a private domain user operation management system. The private domain user operation management system is first introduced below: The main function of the private domain user operation management system is to receive the intention data generated by each outbound calling platform, and then transmit the intention data to other platforms (such as corporate WeChat). Figure 4 Figure (a) in the figure is the main interface of the private domain user operation management system. The left column of the main interface contains several functional modules, including home page, system management, system monitoring, system tools and outbound call center. The outbound call center contains several sub-functional modules, including outbound call channels, reach channels and reach tasks. You can click the "Add button" to add outbound call channels. The outbound call channel information added in the past is displayed in the outbound call channel function module, and the outbound call channels can be synchronized, modified and deleted on this page.

[0032] See also Figure 4 Figure (b) in the figure is the page for adding an outbound channel in the private domain user operation management system, which contains the required system name, company ID, request URL, access ID, access key and callback URL. This initial page has the function of adding an outbound channel.

[0033] See also Figure 4 Figure (c) is the function detail page diagram of the reach channel. Similar to the outbound call channel, you can view the reach channels added in the history on this page.

[0034] See also Figure 4In Figure (d), you can also add a reach platform. First, click the "Add" button, and then enter Figure (d), which contains the required account name, request URL, access ID, access key and platform, as well as optional flexible parameters and extended parameters.

[0035] See also Figure 4 Figure (e) is the reach task page of the private domain user operation management system, where you can view the reach tasks added historically, and you can modify, delete, and export historical reach tasks.

[0036] See also Figure 4 As shown in Figure (f), when you need to add a new reach task, you can click "Add" and then enter the page shown in Figure (f), which contains the required project ID, project name and reach platform account.

[0037] It should be noted that Figure 4 All functions shown in are available, and the three pages of outbound channels, reach channels and reach tasks can exist at the same time.

[0038] The following uses an embodiment and combines Figure 1 , a method for managing private domain users in multiple touch channels in an embodiment of the present application is described: See also Figure 1 , which is a flow chart of a private domain user management method with multiple touch channels in an embodiment of the present application.

[0039] S101, obtaining historical behavior data of the user in each contact channel, performing two-way cross-validation on the historical behavior data, and identifying the behavioral correlation between the user in each contact channel; The system obtains the user's historical behavior data in each contact channel. The historical behavior data includes click records, browsing records, interactive operation records and transaction records, and performs two-way cross-validation on the historical behavior data to identify the behavioral correlation between users in each contact channel.

[0040] The system first obtains the historical behavior data of the target user from multiple contact channels. These historical behavior data include but are not limited to the user's click records, browsing records, interactive operation records, and transaction records. Among them, click records can reflect the user's interest in channel content, browsing records can reveal the user's access path and stay time in the channel, interactive operation records can reflect the depth of interaction between the user and the channel, and transaction records directly explain the user's consumption behavior and transaction amount in the channel. The system comprehensively collects the user's behavioral data on each contact channel, laying a data foundation for subsequent analysis.

[0041] After obtaining massive amounts of historical behavior data, the system needs to evaluate the reliability and consistency of data from each channel. One feasible way is to perform two-way cross-validation on the behavior data, that is, to use the user behavior data from two channels to verify each other and identify possible data deviations or anomalies. For example, if a user frequently browses products in one channel but has few purchase records in another channel, this may suggest that the data from the two channels are not well synchronized or that there are data omissions. Through two-way cross-validation, the system can discover potential data quality issues and perform targeted data cleaning and correction, thereby providing high-quality data support for subsequent analysis.

[0042] In addition, the system can also preliminarily determine the correlation between user behaviors in different channels based on the results of two-way cross-validation. For example, if a user browses a product in one channel and searches for the keywords of this product in another channel, it means that the two channels are related to the user. For example, if a user receives a coupon in one channel and then buys the corresponding product in another channel, this also reflects the behavioral correlation between channels. Through two-way cross-validation, the system can identify some correlation patterns of user behavior between channels and provide guidance for the subsequent construction of collaborative strategies for reaching channels. In general, two-way cross-validation of historical behavior data can help the system fully understand the user's behavioral performance in each reach channel and reveal the correlation between them, which is crucial for the subsequent optimization of marketing resource allocation.

[0043] S102, constructing an inter-channel behavior mapping relationship of the user's migration path between various contact channels based on the behavior correlation; After identifying the correlation of user cross-channel behaviors, the system needs to further explore the user's behavioral migration patterns between these related channels. Specifically, the system can track the user's access path from one channel to another, extract the recurring behavioral migration patterns, and build a behavioral mapping relationship between channels. This mapping relationship can be represented in the form of a directed graph, where the nodes represent the access channels, the edges represent the user's migration path between channels, and the edge weights can represent the frequency or probability of migration.

[0044] For example, if a large number of users jump to the mini program to view product details after browsing the product recommendations in the APP, and complete the order in the official account, then the system can establish a weighted behavior migration path such as "APP-Mini Program-Official Account". This path reflects the user's usage habits and preferences for these three channels, indicating that these channels are interrelated at different stages of reaching users. By analyzing the behavior migration data of massive users, the system can obtain many such cross-channel migration paths, and then summarize the behavior mapping relationship between channels.

[0045] S103, calculating the overlap rate of user behaviors among various access channels, the difference in conversion effects, and the overlap of time series according to the inter-channel behavior mapping relationship; After having the behavior mapping relationship between channels, the system needs to further quantify and analyze the differences and connections between channels in reaching users. An important indicator is the overlap rate of user behavior, that is, what proportion of users have left behavior records on both channels. The higher the overlap rate, the more similar the user groups covered by the two channels are, and there may be a certain redundancy and duplication when reaching users. The system can use the behavior mapping relationship between channels to identify the key paths for users to migrate between channels, calculate the user overlap on different paths, and thus estimate the user overlap rate of each channel.

[0046] In addition to the user overlap rate, the system also needs to examine the differences in conversion effects of each reach channel. The conversion effect reflects the proportion of users who complete specific target behaviors (such as purchases, registrations, sharing, etc.) after the channel reaches the user. By comparing the conversion rates of different channels, the system can evaluate the differences in the effectiveness of each channel in facilitating user conversions. For example, if it is found that the conversion rate of social channels is much higher than that of search channels, it means that social channels may be more conducive to stimulating users' willingness to buy. The system can use the conversion data in the inter-channel behavior mapping to calculate the conversion rate of each migration path, and use this to reflect the overall differences in channel conversion effects.

[0047] The overlap of time series is also an indicator worthy of attention. It reflects the time distribution characteristics of different channels reaching users, as well as the overlap or misalignment between channels in the time dimension. For example, if a user mainly visits the APP in the morning, concentrates on using the mini-program in the afternoon, and prefers to browse information on the official account in the evening, then there will be obvious differences in the time distribution of these three channels reaching users. Understanding the time overlap between channels will help the system to reasonably arrange the marketing resources for different time periods and avoid conflicts in the reach time between channels. The system can use the timestamp information in the inter-channel behavior mapping to perform a sequence comparison of the user reach time of each channel and calculate its overlap.

[0048] S104. Determine the complementary degree of the reach channels based on the overlap rate, difference and coincidence; After calculating the overlap rate, conversion effect difference, and time overlap, the system needs to comprehensively evaluate the complementary relationship between the various channels. The complementary relationship reflects whether multiple channels can complement each other and cooperate with each other when reaching users. Generally speaking, the higher the degree of complementarity between channels, the more conducive it is to achieve multi-channel collaborative operations.

[0049] Specifically speaking of the overlap rate, if the user overlap rate of two channels is low, it means that the user groups they cover are quite different and they can complement each other when reaching different users. On the contrary, if the overlap rate is very high, the two channels may have more duplications when reaching users, and the complementarity is relatively weak. Therefore, the system can set a threshold for the overlap rate. When the overlap rate between channels is lower than the threshold, it is considered that their user reach is complementary.

[0050] From the perspective of conversion effect differences, if the conversion rates of two channels are significantly different, it means that they play different roles in promoting user conversion. At this time, you can consider using channels with higher conversion rates to stimulate user interest and increase purchasing intention; while channels with lower conversion rates are more suitable for early traffic generation and user education. By reasonably matching channels with different conversion effects, you can achieve complementary channel advantages and improve overall marketing efficiency. Therefore, the system can set a threshold for the conversion rate difference. When the conversion rate difference between two channels exceeds this threshold, they are considered to be complementary in conversion effect.

[0051] Time overlap reflects the complementarity between channels from the time dimension. If two channels stagger their peak hours for reaching users, they are complementary in time, and coverage efficiency can be improved by reaching them at different times. On the contrary, if the reach times of two channels overlap to a high degree, there may be problems of resource waste and poor user experience. Therefore, the system can also set a threshold for time overlap. When the time overlap between channels is lower than the threshold, they are considered complementary in terms of time distribution.

[0052] Combining the above three dimensions, the system can derive a complementarity matrix of reach channels. This matrix describes whether any two channels are complementary in terms of user overlap, conversion effect, and time distribution. The system can use the complementarity matrix to quickly determine which channels have greater synergy potential and provide a decision-making basis for channel portfolio optimization. In addition, the complementarity matrix can also be used as a new business discovery tool to help the system discover previously overlooked association patterns between channels and inspire more innovative marketing mix plans. In short, the determination of the degree of complementarity of reach channels can reveal the optimization path of multi-channel collaboration in a data-driven way and improve overall marketing efficiency.

[0053] S105, determining the channel influence index based on the complementary degree of the reaching channels, representing the migration path with a low-dimensional vector, and obtaining a migration behavior vector; After determining the degree of complementarity of the reach channels, the system needs to further evaluate the differences in the influence of each channel on user decisions. Here we introduce a new indicator - the channel influence index, which comprehensively considers factors such as the channel's user reach, conversion effect, and user stickiness, and quantitatively evaluates the relative importance of the channel in influencing user behavior. Specifically, channels with a higher influence index usually perform well in terms of the breadth, depth, and intensity of reaching users, and can more effectively guide users to complete conversion behaviors.

[0054] One way to calculate the channel influence index is to take the weighted sum of various indicators that reflect the channel influence. Among them, the scale of users reached can be measured by the number of daily active users or monthly covered users of the channel; the conversion effect can be measured by the conversion rate or GMV contribution of the channel; user stickiness can be measured by indicators such as the average usage time per person of the channel and the 30-day retention rate. The system can adjust the weight coefficients of various indicators according to the actual business situation to obtain a comprehensive channel influence index. It is worth noting that since the data calibers of different channels may be inconsistent, necessary standardization processing is required when comparing across channels.

[0055] The calculation of the channel influence index also needs to fully consider the regulatory effect of the channel complementarity relationship on the index. Specifically, if the degree of complementarity between two channels is very high, then their influence indexes should not be simply added together, but the multiplying effect brought by the synergy effect should be considered. For example, if a channel with a high influence index can divert a large number of users to another channel with a high conversion rate, then the overall influence of the two channels may far exceed the sum of their individual effects. Therefore, when calculating the channel influence index, the system needs to introduce the complementarity degree matrix as a regulating factor to quantitatively evaluate and correct the synergy between channels.

[0056] Corresponding to the channel influence index, the system also needs to reasonably characterize and measure the behavior migration path. Since the original behavior migration path usually contains a large number of nodes and edges, direct analysis of it may face problems such as high computational complexity and poor interpretability. Therefore, the system can consider using technologies such as embedded learning to map the migration path into a low-dimensional vector space to obtain a set of compact and efficient migration behavior vector representations.

[0057] Specifically, the system can treat the migration path as a kind of sequence data and encode and compress it using algorithms such as Word2Vec and DeepWalk. Through embedded learning, each migration path can be represented as a fixed-length real number vector, and each element in the vector reflects the performance of the path on a certain potential feature dimension. This vectorized representation not only greatly reduces storage and computing overhead, but also reveals the deep associations and similarity structures between migration paths.

[0058] S106, inputting the channel influence index and the migration behavior vector into a preset deep neural network model to obtain the churn risk of each reach channel within a preset time period; The system generates a churn warning model through transfer learning method for channel collaboration units whose churn risk is greater than the preset risk threshold. The churn warning model is used to issue preset warning information when users with churn risk greater than the preset risk threshold are identified.

[0059] After obtaining the channel influence index and migration behavior vector, the system needs to further predict the user churn risk faced by each reach channel. User churn refers to the phenomenon that users stop using a channel or service, which poses a challenge to the stability and sustainability of channel operations. If the churn risk can be identified and quantified in advance, the system can formulate corresponding response and retention strategies for high-risk channels, thereby reducing the losses caused by user churn.

[0060] A feasible method for predicting churn risk is to use deep learning technology to build an end-to-end prediction model. This model takes the channel influence index and migration behavior vector as input, and finally outputs the churn risk probability of each channel in the future after nonlinear transformation and abstraction of multi-layer neural networks. Deep neural networks can automatically learn the complex interactions and high-order associations between input features and capture the deep laws and patterns of churn risk. Therefore, compared with traditional statistical models and machine learning methods, deep learning models can often provide more accurate and stable churn risk prediction results.

[0061] When actually building a churn risk prediction model, the system needs to first label churn users and retained users based on historical data to form a supervised learning training set. Then, the system can select an appropriate deep neural network structure (such as MLP, CNN, RNN, etc.) to design and tune the model according to the specific input feature type and business scenario. In order to make full use of the information provided by the channel influence index and migration behavior vector, the system can add some special processing mechanisms to the model, such as attention mechanisms, gating units, etc., to help the model better capture key churn risk signals.

[0062] During model training and inference, the system also needs to reasonably set some hyperparameters and thresholds to balance the complexity of the model, the risk of overfitting, and the interpretability of the prediction results. For example, by adjusting the number of hidden layers and neurons in the model, the abstract level and expressiveness of the model learning can be controlled; by setting the threshold of the probability of churn risk, a trade-off can be made between accuracy and recall according to business goals. In addition, the system can also use cross-validation, early stopping and other techniques to improve the generalization performance and robustness of the model.

[0063] When predicting churn risk, the system needs to clearly define the time granularity and scope of the prediction. Usually, according to the actual business needs, churn risk prediction can be performed on different time scales such as the next week, month, or quarter. For different prediction time windows, the system can adjust the time span and aggregation method of the input features accordingly to obtain the best prediction effect. For example, if predicting the churn risk in the next week, the input features can be the channel influence index and migration behavior vector of the past week; if predicting the churn risk in the next month, the input features can be extended to the data of the past month or even longer.

[0064] It should be noted that steps S104-S106 and steps S107-S108 are parallel steps and there is no precedence relationship in terms of time sequence and logical order.

[0065] S107. For channel collaboration units whose churn risk is greater than a preset risk threshold, a churn warning model is generated by a transfer learning method; After identifying the channel collaboration units with high churn risk, the system needs to further build a set of targeted early warning models to dynamically monitor and timely warn of user churn in the unit. Since the user characteristics, behavior patterns, and churn mechanisms of different channel collaboration units may vary greatly, it is unrealistic to train a separate early warning model for each high-risk unit. This not only requires a large amount of labeled data and computing resources, but may also face problems such as sparse training samples and model overfitting.

[0066] One way to solve this problem is to use transfer learning technology to generate a personalized early warning model for specific channel collaboration units based on the overall churn risk prediction model. Transfer learning refers to a type of machine learning method that uses existing models and knowledge to quickly adapt to and learn new tasks or scenarios. Its core idea is to find the correlation and commonality between different tasks or fields, and use this prior knowledge to guide and accelerate the learning process of the new model.

[0067] Specifically for the generation of the churn warning model, the system can adopt a transfer learning method based on feature representation transfer. First, the system can use the trained overall churn risk prediction model to extract the feature representation of the middle layer as the general churn risk semantics. These features contain the general laws and high-level patterns of churn risk learned by the model from large-scale multi-channel data, and have certain guiding significance for different channel collaboration units. Next, the system can use a small amount of annotated data of high-risk channel collaboration units, fine-tune and adapt on the basis of the general churn risk semantics, and generate a specialized warning model for this unit.

[0068] During the training process of the specialized early warning model, the system can make full use of the prior knowledge of general semantics to avoid overfitting and improve the generalization performance of the model by freezing some model parameters and introducing regularization terms. At the same time, the system can also use active learning, few-sample learning and other technologies to further reduce the number of labeled samples required and accelerate the generation and iteration of the early warning model. For example, the system can perform a preliminary churn risk ranking of users in high-risk units based on general semantics, and then focus on labeling the top-ranked high-risk users, thereby obtaining higher information gain at a lower labeling cost.

[0069] After generating the early warning model, the system needs to set a reasonable early warning threshold. When the probability of user loss exceeds the threshold, the corresponding early warning mechanism and retention action will be triggered. The setting of the early warning threshold needs to balance the timeliness and accuracy of the early warning. If the threshold is set too high, the best retention opportunity may be missed; if the threshold is set too low, it may cause too many false alarms and waste of resources. Therefore, the system needs to dynamically adjust and optimize the early warning threshold according to the business characteristics and risk preferences of the channel collaboration unit to improve the effectiveness of the early warning mechanism.

[0070] S108, grouping and clustering the various reach channels according to their complementarity levels to obtain a number of channel collaboration units; The system groups and clusters the reach channels according to their complementarity levels to obtain several channel collaboration units. The reach channels in the channel collaboration units are the reach channels whose corresponding complementarity levels are greater than a preset threshold.

[0071] After calculating the complementary degree matrix of the reach channels, the system needs to further group and cluster the channels to identify several channel combinations with close internal synergy and obvious external differences, namely the so-called channel synergy units. Dividing channels with strong complementarity into the same synergy unit can maximize the synergy between channels; and the differences between different synergy units can help achieve differentiated operations and provide personalized services for different user groups. Therefore, the identification and division of channel synergy units is an important basis for multi-channel refined operations.

[0072] The key here is how to design a reasonable grouping clustering algorithm based on the complementary relationship between channels. Common clustering algorithms such as K-means and hierarchical clustering are all based on the distance or similarity measurement between samples for division. In the context of channel complementary relationships, we need to define the similarity between channels based on the complementarity matrix. A simple and intuitive idea is to regard the complementarity matrix as a channel similarity matrix. The larger the element value in the matrix, the stronger the complementarity between the corresponding two channels, and the more likely they should be classified into the same category when clustering.

[0073] S109, calculating the optimal time difference for reaching users in each channel collaboration unit, and allocating marketing resources among the various reaching channels based on the optimal time difference.

[0074] In the scenario of reaching users through multiple channels, there is a certain time difference between different channels, that is, after a user interacts on a certain channel, he or she may generate subsequent behavior on another related channel after a certain time interval. Optimizing the time difference between channels can help improve the overall reach effect and save marketing costs.

[0075] In actual calculations, the system can use the behavioral data of each channel within the channel collaboration unit to construct a cross-channel user behavior time series. By analyzing the temporal sequence and interval distribution of different behaviors, the average time difference and optimal time difference between channels can be obtained. Among them, the optimal time difference can be defined by specific business goals (such as conversion rate, revisit cycle, etc.), which represents the best time point to trigger subsequent channel behaviors. After obtaining the optimal time difference, the system can automatically generate the best timetable for channel reach and guide the rhythm of marketing resource delivery in each channel. For example, if a user browses a product on a public account and generally places an order on a mini program one hour later, the system can place relevant coupons in the mini program about one hour after the user browses to promote conversion.

[0076] When allocating marketing resources, the system can also take into account the weight differences between channels. Specifically, the reach effect of each channel can be attribution analyzed to calculate the contribution of each channel to the final conversion. Channels with high contribution can be allocated more marketing resources, such as advertising budgets, number of pushes, etc.; while channels with low contribution can reduce resource investment accordingly. At the same time, this dynamic resource allocation based on attribution analysis can be adjusted according to the real-time changes in the effects of each channel, thereby achieving dynamic optimization of marketing resources. For example, if the system finds that the conversion effect of a certain channel is much higher than that of other channels, it can increase the resource allocation ratio of that channel in real time to maximize the overall marketing performance.

[0077] In the above embodiment, obtaining the historical behavior data of users in each reach channel and performing two-way cross-validation can improve the accuracy of identifying the behavioral correlation between users in different channels, and then construct a mapping relationship of the migration path of users between the reach channels. By clustering channels based on the degree of complementarity as the basis for grouping, reach channels with strong complementarity can be grouped into channel coordination units to reduce the duplication of channel resources. By allocating marketing resources by calculating the optimal reach time difference within the coordination unit, it not only ensures the coordination of each reach channel in the time dimension, but also can reasonably allocate resources according to the characteristics of the channel, thereby improving the overall resource utilization efficiency. The complementary advantages and optimal resource allocation between multiple reach channels are realized, and the accuracy of private domain operation management is improved.

[0078] In order to further improve the accuracy of the calculation of the reach time difference, this application also provides another implementation scheme. Based on the previous embodiment, considering that each reach channel may be affected by environmental noise factors such as system failures and network delays during actual operation, these factors will interfere with the accuracy of the reach time, thereby affecting the rationality of marketing resource allocation. Figure 2 , another method for managing private domain users in multiple channels in an embodiment of the present application is described: See also Figure 2 , which is another flow chart of a private domain user management method with multiple touch channels in an embodiment of the present application.

[0079] S201, obtaining environmental noise data of each reach channel, and constructing a time window correction function based on the environmental noise data; Obtain the environmental noise data of each reach channel, which includes system failure records, network delay data, and user complaint records, and build a time window correction function based on the environmental noise data. The time window correction function is used to correct the time series data of each reach channel. The construction of the time window correction function specifically includes: Calculate the noise intensity coefficient of each reach channel in different time periods; Determine the dynamic expansion range of the time window according to the noise intensity coefficient, and construct an adaptive time window based on the dynamic expansion range; The historical reach data is segmented based on the adaptive time window to obtain the abnormal fluctuation range; The data in the abnormal fluctuation range is smoothed to obtain the time window correction coefficient.

[0080] In this step, the system first obtains the environmental noise data of each reach channel. These environmental noise data may include but are not limited to system failure records, network delay data, and user complaint records. By collecting and analyzing these data, the system can fully understand the interference factors of each reach channel during the actual operation process, and provide data support for the subsequent construction of the time window correction function.

[0081] After obtaining the environmental noise data, the system will construct a time window correction function based on this data. Specifically, the system can first calculate the noise intensity coefficient of each reach channel in different time periods to quantify the impact of environmental noise on the reach effect. Then, the system determines the dynamic expansion range of the time window based on the noise intensity coefficient, and constructs an adaptive time window based on this range. Next, the system uses the adaptive time window to segment the historical reach data and identify abnormal fluctuation intervals in the data. Finally, the system smoothes the data in these abnormal fluctuation intervals to obtain the time window correction coefficient.

[0082] Furthermore, in another embodiment, after calculating the effective reach rate of each reach channel, the method further includes: constructing a reach effect evaluation index system, wherein the reach effect evaluation index system includes user responsiveness and conversion timeliness; Calculate the effect scores of each reach channel in different time periods based on the reach effect evaluation index system, and establish the corresponding relationship between the effect score and the reach time; Generate optimization suggestions for reach time based on the corresponding relationship. Specifically: determine the periodic change characteristics of the effect score based on the time series data of the effect score of each reach channel; According to the periodic change characteristics, each time period is divided into an advantageous period, a stable period and an inferior period according to the effect score, and the duration of each period is calculated; Identify the complementary relationship between the reach time of various reach channels in the same time period; Calculate the optimal reach time interval for each reach channel based on the complementary relationship of reach time, and generate channel collaborative reach suggestions; Develop a plan to adjust the reach time based on channel collaborative reach recommendations and the operating costs and resource constraints of each reach channel; Convert the reach time adjustment plan into reach time optimization suggestion.

[0083] In this embodiment, after calculating the effective reach rate of each reach channel, the system will further construct a reach effect evaluation index system to evaluate the reach effect of the channel from multiple angles. The index system mainly includes two dimensions: user responsiveness and conversion timeliness. User responsiveness reflects the user's acceptance of the channel reach content and willingness to interact, which can be measured by indicators such as click-through rate and open rate; while conversion timeliness examines the time interval from the user receiving the reach information to completing the conversion behavior, reflecting the influence of channel reach on user decision-making.

[0084] After building the evaluation index system, the system will calculate the effect scores of each reach channel in different time periods based on the system. Specifically, the system can collect channel reach data within a certain time range, such as the number of reaches, user response data, and conversion data in each time period, and then quantitatively evaluate the reach effect of each channel in each time period according to the calculation rules in the index system to obtain a series of effect score data. At the same time, the system will also establish a mapping relationship between the effect score and the corresponding reach time to form a "time-effect" data set.

[0085] Next, the system will generate recommendations for optimizing reach time based on the corresponding relationship between the effect score and reach time. First, the system will analyze the time series data of the effect scores of each channel to find out the periodic change characteristics of the scores, such as whether there are obvious peaks and troughs, and the regularity of the peaks. Then, based on these periodic characteristics, the system will divide each time period into three categories: advantageous time period, stable time period, and disadvantageous time period, and count the duration of each time period.

[0086] After completing the time period division, the system will further analyze the differences in the reach effects of different channels in the same time period and identify the complementary relationship between the reach time between channels. For example, the reach effect of channel A in the morning period is significantly better than that of channel B, while channel B performs better in the evening period. The advantageous time periods of the two channels are complementary. Based on these complementary relationships, the system can calculate the optimal reach time interval for each channel and formulate a recommended plan for channel collaborative reach, such as "reach users on channel A around 10 am, and reach them a second time on channel B at 3 pm".

[0087] S202, calibrating the historical reach time of each reach channel according to the time window correction function; In this step, the system will use the constructed time window correction function to calibrate the historical reach time data of each reach channel. Through this calibration process, the system can effectively eliminate the impact of environmental noise on reach time, improve the accuracy and reliability of reach time data, and lay the foundation for subsequent optimal time difference calculation.

[0088] In specific implementation, the system can use a sliding time window to calibrate the touch time of each timestamp one by one. For each touch time data in the window, the system multiplies it by the corresponding time window correction coefficient to obtain the calibrated touch time. After completing the calibration of a single timestamp, the time window slides backward to start the calibration of the next timestamp until all historical touch time data are calibrated.

[0089] During the calibration process, some historical data of the reach channels may be missing or the timestamps may be recorded incorrectly, which makes it impossible to directly apply the time window correction function. To address this issue, the system can pre-process the original data through data cleaning and repair technology, interpolate missing values, and correct incorrect timestamps. In addition, the system can also introduce a data quality monitoring mechanism to monitor abnormal fluctuations in calibration results in real time, and issue timely warnings when encountering data with excessive deviations to ensure the smooth operation of the calibration process.

[0090] S203. Calculate the effective reach rate of each reach channel based on the calibrated reach time, and use the effective reach rate as a weight factor for calculating the optimal time difference.

[0091] After completing the noise calibration of the reach time, the system will calculate the effective reach rate of each reach channel based on the calibrated data. The effective reach rate can measure the actual effect of each channel in reaching users and is an important indicator reflecting the reach ability of the channel. By calculating the effective reach rate, the system can quantitatively evaluate the marketing effectiveness of different reach channels and provide a decision-making basis for the rational allocation of marketing resources.

[0092] When calculating the effective reach rate, the system needs to first define the standard for effective reach, such as whether the user completes a specific conversion behavior within a specified time after reaching out. Then, the system performs statistical analysis on the calibrated reach time data, calculates the number of effective reaches and the total number of reaches for each channel within a certain time period, and divides the two to get the effective reach rate.

[0093] It is worth noting that the effective reach rates of different reach channels may vary greatly. In order to reasonably evaluate the reach time difference between channels, the system introduces the effective reach rate into the calculation of the optimal time difference as an important weighting factor. The higher the reach rate of a channel, the greater the weight given to its reach time in the time difference calculation, while the weight of a channel with poor reach effect is relatively low. Through this weighted calculation, the system can fully consider the actual reach capabilities of each channel, making the final channel collaboration plan more optimized and efficient.

[0094] At the same time, in setting the weight factor, the system can also take into account other factors such as the channel's user coverage and marketing costs, balance timeliness and economy, and maximize the comprehensive benefits of multi-channel collaboration. In addition, the setting of the weight factor should also have a certain degree of dynamic adjustment capabilities to adapt to changes in the market environment and user habits.

[0095] In the above embodiment, the environmental noise data of each reach channel is collected and analyzed, and a time window correction function is constructed to correct the time series data, which can reduce the interference of external factors such as system failures and network delays on the reach effect. The calibrated historical reach time more accurately reflects the real interaction between users and channels, and the effective reach rate calculated in this way can be used as a reliable weight factor to participate in the calculation of the optimal time difference. The accuracy of the time difference calculation is improved, making the subsequent resource allocation more reasonable.

[0096] The following describes the system in the embodiment of the present invention from the perspective of hardware processing. Figure 3 , which is a schematic diagram of the physical device structure of a private domain user management system with multiple reach channels provided in an embodiment of the present application.

[0097] It should be noted that Figure 3 The structure of the system shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0098] like Figure 3 As shown, the system includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 302 or the program loaded from the storage part 308 to the random access memory (RAM) 303, such as executing the method in the above embodiment. In RAM 303, various programs and data required for system operation are also stored. CPU 301, ROM 302 and RAM 303 are connected to each other through a bus 304. Input / output (I / O) interface 305 is also connected to bus 304.

[0099] The following components are connected to the I / O interface 305: an input section 306 including a camera, an infrared sensor, etc.; an output section 307 including a liquid crystal display (LCD) and a speaker, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I / O interface 305 as needed. A removable medium 311, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 310 as needed so that a computer program read therefrom is installed into the storage section 308 as needed.

[0100] In particular, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication part 309, and / or installed from a removable medium 311. When the computer program is executed by the central processing unit (CPU) 301, various functions defined in the present invention are performed.

[0101] It should be noted that the computer-readable medium shown in the embodiment of the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable computer program. Such a propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing.

[0102] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present invention. Among them, each box in the flowchart or block diagram can represent a module, a program segment, or a part of the code, and the above-mentioned module, program segment, or a part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0103] As another aspect, the present invention further provides a computer-readable storage medium, which may be included in the system described in the above embodiment; or may exist independently without being assembled into the system. The above storage medium carries one or more computer programs, and when the above one or more computer programs are executed by a processor of a system, the system implements the method provided in the above embodiment.

[0104] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

[0105] As used in the above embodiments, the term "when..." may be interpreted to mean "if..." or "after..." or "in response to determining..." or "in response to detecting...", depending on the context. Similarly, the phrases "upon determining..." or "if (the stated condition or event) is detected" may be interpreted to mean "if determining..." or "in response to determining..." or "upon detecting (the stated condition or event)" or "in response to detecting (the stated condition or event)", depending on the context.

[0106] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state hard disk), etc.

[0107] Those skilled in the art can understand that to implement all or part of the processes in the above-mentioned embodiments, the processes can be completed by computer programs to instruct related hardware, and the programs can be stored in computer-readable storage media. When the programs are executed, they can include the processes of the above-mentioned method embodiments. The aforementioned storage media include: ROM or random access memory RAM, magnetic disk or optical disk and other media that can store program codes.

Claims

1. A multi-touch channel private domain user management method, characterized in that: include: Obtain the user's historical behavior data on each contact channel, the historical behavior data includes click records, browsing records, interactive operation records and transaction records; Performing two-way cross-validation on the historical behavior data to identify the behavioral correlation of the user between the various contact channels; Constructing an inter-channel behavior mapping relationship of the user's migration path between the contact channels based on the behavior correlation; Calculate the overlap rate of user behaviors among the contact channels, the difference in conversion effects, and the overlap of time series according to the inter-channel behavior mapping relationship; Determine the complementarity of the reach channels based on the overlap rate, the difference and the coincidence; Grouping and clustering the reach channels according to the complementarity of the reach channels to obtain a number of channel coordination units, wherein the reach channels in the channel coordination units are reach channels whose corresponding complementarity of the reach channels is greater than a preset threshold; The optimal time difference for reaching users in each channel collaboration unit is calculated, and marketing resources are allocated among the reaching channels based on the optimal time difference.

2. The method according to claim 1, characterized in that: The calculating the optimal time difference for reaching users in each channel collaboration unit and allocating marketing resources among the reaching channels based on the optimal time difference specifically includes: Determine the customer reach time distribution of each reach channel according to the historical reach data of each reach channel in each channel collaboration unit, and obtain the high-frequency reach time period of each reach channel; Calculate the center of gravity of the high-frequency reach time period, and define the time difference between the centers of gravity of each reach channel as the optimal time difference; According to the total amount of marketing resources of each channel collaboration unit, the initial amount of marketing resources of each reach channel is obtained by allocating them according to the importance coefficient of each reach channel. The importance coefficient is determined according to the user scale, user value and conversion effect of each reach channel.

3. The method according to claim 1, characterized in that After determining the complementarity of the reach channels based on the overlap rate, the difference and the coincidence, the method further includes: Determine a channel influence index based on the complementary degree of the reach channels; Representing the migration path as a low-dimensional vector to obtain a migration behavior vector; Input the channel influence index and the migration behavior vector into a preset deep neural network model to obtain the churn risk of each of the reach channels within a preset time period; For the channel collaboration unit whose churn risk is greater than the preset risk threshold, a churn warning model is generated by a transfer learning method. The churn warning model is used to issue a preset warning message when a user whose churn risk is greater than the preset risk threshold is identified.

4. The method according to claim 1, characterized in that: Before calculating the optimal time difference for reaching users in each channel collaboration unit, the method further includes: Obtaining environmental noise data of each of the contact channels, wherein the environmental noise data includes system failure records, network delay data, and user complaint records; Constructing a time window correction function based on the environmental noise data, wherein the time window correction function is used to correct the time series data of each of the reaching channels; Calibrate the historical reach time of each of the reach channels according to the time window correction function; The effective reach rate of each reach channel is calculated based on the calibrated reach time, and the effective reach rate is used as a weight factor for calculating the optimal time difference.

5. The method according to claim 4, characterized in that The constructing of the time window correction function specifically includes: Calculate the noise intensity coefficient of each of the reach channels in different time periods; Determining a dynamic scaling range of a time window according to the noise intensity coefficient, and constructing an adaptive time window based on the dynamic scaling range; Based on the adaptive time window, the historical reach data is processed in segments to obtain an abnormal fluctuation range; The data in the abnormal fluctuation range is smoothed to obtain a time window correction coefficient.

6. The method according to claim 4 or 5, characterized in that: After calculating the effective reach rate of each of the reach channels, the method further includes: Constructing a reach effect evaluation index system, wherein the reach effect evaluation index system includes user responsiveness and conversion timeliness; Calculate the effect score of each of the reach channels in the different time periods based on the reach effect evaluation index system, and establish a corresponding relationship between the effect score and the reach time; Generate a reach time optimization suggestion based on the corresponding relationship.

7. The method according to claim 6, characterized in that Generating a reach time optimization suggestion according to the corresponding relationship specifically includes: Determine the periodic variation characteristics of the effect score based on the time series data of the effect score of each of the reach channels; According to the periodic change characteristics, each time period is divided into an advantageous time period, a stable time period and an inferior time period according to the effect score, and the duration of each time period is calculated; Identify the complementary relationship of reach time between the reach channels in the same time period; Calculate the optimal reach time interval of each reach channel based on the reach time complementary relationship, and generate channel collaborative reach suggestions; Formulate a reach time adjustment plan based on the channel collaborative reach suggestions and the operating costs and resource constraints of each reach channel; The reach time adjustment plan is converted into a reach time optimization suggestion.

8. A multi-channel private domain user management system, characterized in that: The system comprises: One or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to cause the system to execute the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that: When the instructions are executed on a system, the system is caused to execute the method according to any one of claims 1 to 7.

10. A computer program product, characterized in that When the computer program product is run on a system, the system is caused to execute the method according to any one of claims 1 to 7.

Images