User loss early warning method, device, equipment and storage medium
By integrating multimodal data and time decay factors, the system accurately depicts changes in user behavior, solving the problem of delayed early warning in static threshold models in OTT platforms. This enables early identification and personalized intervention of user churn risks, thereby reducing user churn rates.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- FUTURE TV CO LTD
- Filing Date
- 2026-03-23
- Publication Date
- 2026-06-23
AI Technical Summary
In existing OTT platform user churn management technologies, static threshold models cannot identify dynamic trends in user behavior changes in a timely manner, resulting in delayed early warnings and an inability to intervene and take targeted measures in advance, leading to the loss of existing users.
The system integrates explicit and implicit behavioral data and device attribute data using a multimodal feature fusion module, and makes predictions through a churn warning model. It also incorporates a time decay factor to highlight the weight of recent behaviors, accurately characterizes changes in user behavior, and enables early identification of churn risks and generation of personalized intervention strategies.
It enables early identification and accurate attribution of user churn risks, supports the platform to intervene in advance with personalized measures, reduces user churn rate, and meets the needs of refined operations.
Smart Images
Figure CN122269085A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of user operation technology, and more specifically, to a user churn early warning method, device, equipment, and storage medium. Background Technology
[0002] The over-the-top (OTT) television industry has gradually moved beyond the period of rapid user growth, with user acquisition costs continuing to rise. Compared to acquiring the same number of new users, maintaining existing users and reducing churn rates can bring more significant economic benefits to the platform. Therefore, user churn management has become an important issue for OTT platform operations.
[0003] Among existing OTT platform user churn management technologies, static threshold models are a widely used approach. This approach determines a user's churn status by setting a fixed churn threshold time interval (e.g., setting 30 consecutive days of inactivity as the churn criterion) and combining it with single behavioral indicators (such as user viewing time, payment frequency, etc.). When a user meets the preset time interval condition or the single behavioral indicator reaches the set threshold, the system triggers generalized intervention measures, such as uniformly pushing coupons to the user, in an attempt to win them back.
[0004] However, such static threshold models suffer from significant early warning lag in practical applications. Because the models rely solely on fixed time thresholds and single behavioral indicators, they fail to effectively capture the dynamic trends of user behavior changes. In real-world scenarios, user churn is often a gradual process, not a sudden occurrence. For example, if some users' average daily viewing time drops sharply from 2 hours to 1 hour, this significant behavioral decline clearly indicates a high risk of churn. However, because the preset churn threshold time interval has not been reached, the static threshold model cannot promptly identify this risk and trigger an early warning. This makes it difficult for the platform to intervene in advance and take targeted measures, ultimately leading to unnecessary churn of existing users and failing to meet the needs of refined operations for OTT platforms. Summary of the Invention
[0005] This application addresses the shortcomings of the prior art by providing a user churn early warning method, apparatus, device, and storage medium to solve the problems existing in the prior art.
[0006] The technical solution adopted in the embodiments of this application is as follows: In a first aspect, embodiments of this application provide a user churn early warning method, including: Acquire target user behavior data and device attribute data for the Internet TV system over multiple historical periods; Based on behavioral data from multiple historical time periods and device attribute data, the user time series data of the target user is obtained; Based on the user time-series data, a churn prediction model is used to predict the churn probability and reasons for the churn of the target user.
[0007] In one embodiment, before predicting the churn probability and reasons for churn of the target user using a churn warning model based on the user time-series data, the method further includes: Obtain the text comments corresponding to the target user; A preset multimodal feature fusion module is used to semantically associate the text comments and the user time-series data to obtain the fused target time-series data; The step of using a churn prediction model based on the user time-series data to predict the churn probability and reasons for churn of the target user includes: Based on the target time-series data, the churn prediction model is used to predict the churn probability and reasons for the churn of the target user.
[0008] In one embodiment, the churn prediction model includes: a feature extraction module, a behavior capture module, a relationship analysis module, and an output module; the step of using the churn prediction model to predict the churn probability and reasons for churn of the target user based on the user time-series data includes: Based on the user time-series data, the feature extraction module is used to obtain a time-series feature vector; Based on the temporal feature vector, the behavior capture module is used to capture behavior and obtain the behavior pattern; Based on the time-series feature vector, the relationship analysis module is used to analyze and obtain the user equipment interaction relationship; Based on the behavioral patterns and the user device interaction relationships, the output module is used to obtain the churn probability and churn reasons of the target user.
[0009] In one embodiment, the method further includes: If the churn probability of the target user is greater than or equal to the preset churn probability threshold corresponding to the value level of the target user, then a churn intervention strategy for the target user is generated based on the value level of the target user and the reason for churn.
[0010] In one embodiment, generating a churn intervention strategy for the target user based on the target user's value level and the reason for churn includes: If the value level is the first level, then the preset subsidy amount information is pushed to the target user as the churn intervention strategy; If the value level is the second level, then a social sharing campaign is pushed to the target user as the churn intervention strategy. If the value level is level three, a limited-time trial activity will be pushed to the target user as the churn intervention strategy. If the value level is level four, a competitor content comparison report will be pushed to the target user as part of the churn intervention strategy.
[0011] In one embodiment, obtaining the target user's time-series data based on behavioral data from multiple historical time periods and device attribute data includes: Obtain the behavioral weights for multiple historical time periods; Based on the behavioral weights of the multiple historical time periods, the behavioral data of the multiple historical time periods are adjusted to obtain the adjusted behavioral data; Based on the adjusted behavioral data and the device attribute data, the user time series data of the target user is obtained.
[0012] In one embodiment, after generating the churn intervention strategy for the target user, the method further includes: Implement churn intervention strategies targeting the aforementioned user group; After the churn intervention strategy is implemented, the target user's Internet TV system usage data is obtained; The churn warning model is updated based on the usage data of the Internet TV system.
[0013] Secondly, embodiments of this application provide a user churn early warning device, comprising: The acquisition module is used to acquire behavioral data and device attribute data of the target user on the Internet TV system over multiple historical periods. The processing module is used to obtain the user time series data of the target user based on the behavioral data of multiple historical times and device attribute data; The prediction module is used to predict the churn probability and reasons for churn of the target user based on the user time series data using a churn warning model.
[0014] Thirdly, embodiments of this application provide an electronic device, including: a processor, a storage medium, and a bus. The storage medium stores program instructions executable by the processor. When the electronic device is running, the processor communicates with the storage medium via the bus, and the processor executes the program instructions to implement the user churn warning method described in any of the above embodiments.
[0015] Fourthly, embodiments of this application provide a readable storage medium storing program instructions, which, when executed by a processor, implement the user churn warning method described in any of the above embodiments.
[0016] The beneficial effects of this application are as follows: This application provides a user churn early warning method. First, it integrates multiple types of data, such as explicit and implicit behaviors and device attributes, to overcome the shortcomings of existing technologies that rely on only a single type of data, thus enriching the feature dimensions of churn prediction. Second, by constructing a time-series data model combined with a time decay factor, it highlights the weight of users' recent behaviors, accurately depicting the gradual process of user behavior from active to declining, and solving the problem that static models cannot capture dynamic changes. Finally, the churn early warning model enables early identification and accurate attribution of churn risks, supporting the platform to intervene in advance. Attached Figure Description
[0017] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 This is one of the flowcharts illustrating the user churn warning method provided in the embodiments of this application; Figure 2 The second flowchart illustrates the user churn warning method provided in this application embodiment; Figure 3 The third flowchart illustrating the user churn warning method provided in this application embodiment; Figure 4 The fourth flowchart illustrating the user churn warning method provided in this application embodiment; Figure 5 Fifth flowchart illustrating the user churn warning method provided in this application embodiment; Figure 6 The sixth flowchart illustrating the user churn warning method provided in this application embodiment; Figure 7 This is a schematic diagram of the user churn early warning device provided in the embodiments of this application; Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation
[0019] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of this application, but not all embodiments.
[0020] Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.
[0021] Furthermore, the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Additionally, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0022] It should be noted that, where there is no conflict, the features in the embodiments of this application can be combined with each other.
[0023] This application provides a user churn warning method, which can be generated by any electronic device with computing and processing capabilities. The electronic device can be, for example, a terminal-facing computer device or a backend server.
[0024] The following examples, in conjunction with the accompanying drawings, provide specific illustrations of the user churn early warning method provided in this application.
[0025] Figure 1 This is one of the flowcharts illustrating the user churn warning method provided in the embodiments of this application, such as... Figure 1 As shown, the method includes: S101. Obtain the target user's behavioral data and device attribute data for the Internet TV system at multiple historical times.
[0026] Acquire target users' behavioral data and device attribute data for Internet TV systems (Over-The-Top, OTT) over multiple historical periods. The behavioral data includes explicit behavioral data and implicit behavioral data. Explicit behavioral data specifically includes playback completion rate and payment frequency, while implicit behavioral data specifically includes interface hotspot click distribution and device switching interval.
[0027] Device attribute data includes, but is not limited to, the manufacturer, model, memory, and application version of the device used by the target user.
[0028] S102. Based on behavioral data from multiple historical time periods and device attribute data, obtain the user time series data of the target user.
[0029] Figure 2 This is a second flowchart illustrating the user churn warning method provided in this application embodiment, as shown below. Figure 2 As shown, S102 specifically includes: S201, Obtain the behavioral weights for multiple historical time periods.
[0030] The time decay factor formula α(t) = e λt calculates the behavioral weights at each historical time point, where t is the interval between the historical time and the current time, λ is the decay coefficient (the longer the time, the lower the behavioral weight), and e is the natural constant, approximately equal to 2.71828.
[0031] S202. Adjust the behavioral data from multiple historical time periods according to the behavioral weights of multiple historical time periods to obtain the adjusted behavioral data.
[0032] By multiplying the behavioral data at each historical time point with the corresponding weight α(t), a weighted adjustment of behavioral data at different time dimensions is achieved, highlighting the impact of recent behavior on user status.
[0033] S203. Based on the adjusted behavioral data and device attribute data, obtain the user time sequence data of the target user.
[0034] The weighted and adjusted explicit and implicit behavioral data are integrated with equipment attribute data such as equipment manufacturers and models in a structured manner, and arranged in chronological order to form user time-series data.
[0035] S103. Based on user time-series data, a churn prediction model is used to predict the churn probability and reasons for churn of target users.
[0036] Input user time-series data into a multimodal churn warning model, and output a churn probability with a confidence interval of 95%, as well as one or more churn reasons from content fatigue, price sensitivity, competitor attraction, operational obstacles, and natural churn.
[0037] In summary, this embodiment provides a user churn prediction method. First, it integrates multiple types of data, such as explicit and implicit behaviors and device attributes, to overcome the shortcomings of existing technologies that rely on only single data points and enrich the feature dimensions of churn prediction. Second, by constructing a time-series data model combined with a time decay factor, it highlights the weight of users' recent behaviors, accurately depicting the gradual process of user behavior from active to declining, and solving the problem that static models cannot capture dynamic changes. Finally, the churn prediction model enables early identification and accurate attribution of churn risks, supporting early intervention by the platform.
[0038] Figure 3 This is the third flowchart illustrating the user churn warning method provided in this application embodiment, as shown below. Figure 3 As shown, before executing step S103, which uses a churn prediction model based on user time-series data to predict the churn probability and reasons for churn of the target user, the method of this application further includes: S301. Obtain the text comments corresponding to the target user.
[0039] Collect text data such as content reviews and feedback messages posted by target users within the Internet TV system, including comments on picture quality, content satisfaction, and user experience complaints.
[0040] S302. Using a pre-set multimodal feature fusion module, semantic association is performed on text comments and user time-series data to obtain the fused target time-series data.
[0041] Feature fusion is achieved through a dual-tower network structure. The left side uses a BERT encoder to process text comments and extract NLP features; the right side uses an Informer model to process user time-series data and extract time-series features. The Cross-Attention module establishes semantic association between the two types of features, associating descriptions such as image quality stuttering in text comments with behaviors such as high-frequency buffering in behavioral data, thus obtaining the fused target time-series data.
[0042] Based on this, S103 includes: S303. Based on the target time series data, use the churn warning model to predict the churn probability and reasons for the churn of the target users.
[0043] The fused target time-series feature vector is input into the churn warning model. Through the collaborative processing of various modules in the model, the model outputs accurate churn probability and clearly categorized churn reasons.
[0044] In one embodiment, the churn warning model includes an input layer, a hidden layer, and an output layer. The input layer includes a feature extraction module, the hidden layer includes a behavior capture module and a relationship analysis module, and the output layer includes an output module.
[0045] Figure 4 This is the fourth flowchart illustrating the user churn warning method provided in this application embodiment, as shown below. Figure 4 As shown in S103, based on user time-series data, a churn prediction model is used to predict the churn probability and reasons for churn of the target user, including: S401. Based on the user's time series data, the feature extraction module is used to obtain the time series feature vector.
[0046] The feature extraction module standardizes and normalizes user time-series data, extracts features from dimensions such as playback duration, payment interval, and device switching frequency, and generates a 128-dimensional time-series feature vector containing 20 core time-series features (such as 30-day moving average viewing duration).
[0047] S402. Based on the temporal feature vector, the behavior capture module is used to capture behavior and obtain the behavior pattern.
[0048] The behavior capture module is a Long Short-Term Memory (LSTM) network. By analyzing the sequential changes in user behavior in the temporal feature vector, it captures the behavioral evolution patterns of users, such as from high-frequency viewing to low-frequency login, and from paid activity to ceasing payment.
[0049] S403. Based on the time-series feature vector, the relationship analysis module is used to analyze and obtain the user equipment interaction relationship.
[0050] The relationship analysis module is a graph sampling and aggregation network (GraphSAGE). Based on data such as device switching intervals and multi-device login records in the time-series feature vector, it analyzes the relationship characteristics of users' interactions with different devices, such as interaction frequency and duration, as well as the impact of users' related user behaviors in social networks.
[0051] S404. Based on behavioral patterns and user device interaction, use the output module to obtain the churn probability and reasons for churn of the target user.
[0052] The output module performs comprehensive calculations on behavioral patterns and user device interaction relationships, outputs the churn probability in the range of 0-100% (confidence level 95%), and matches the corresponding churn reason classification results.
[0053] In one embodiment, if the churn probability of a target user is greater than or equal to a preset churn probability threshold corresponding to the value level of the target user, a churn intervention strategy for the target user is generated based on the value level of the target user and the reason for churn. Figure 5 This is the fifth flowchart illustrating the user churn warning method provided in the embodiments of this application, as shown below. Figure 5 As shown, it specifically includes: S501. If the value level is the first level, push the preset subsidy amount information to the target user as a churn intervention strategy.
[0054] If the value level is the first level (ARPU ≥ 200 yuan, representing high-value lost users), then push the preset subsidy amount information to the target users as a churn intervention strategy. Among them, the subsidy amount is calculated as 0.2 times the ARPU value of the target user, and at the same time, it is paired with a content customization package and exclusive customer service. The exclusive customer service will actively reach the user within 30 minutes after the intervention is triggered.
[0055] Among them, ARPU is the abbreviation of Average Revenue Per User, which means average revenue per user. It refers to the average revenue obtained by the Internet TV system from a single user within a certain statistical period (such as monthly or quarterly). The calculation method is the total platform revenue ÷ the total number of active users, and it is the core indicator for measuring user value and dividing user stratification.
[0056] S502: If the value level is the second level, then push a social裂变 activity to the target users as a churn intervention strategy.
[0057] If the value level is the second level (50 < ARPU < 200 yuan, medium-value lost users), then push a social裂变 promotion activity to the target users as a churn intervention strategy. For example, push personalized Top3 popular content to the target users, generate a dedicated sharing link for this user, and the user can get a 7-day membership reward after successfully裂变 new users.
[0058] S503: If the value level is the third level, then push a limited-time trial activity to the target users as a churn intervention strategy.
[0059] If the value level is the third level (ARPU ≤ 50 yuan, low-value lost users), then push a limited-time trial activity to the target users as a churn intervention strategy. For example, automatically activate a 3-day free trial permission for the user, simplify the ordering process, and support a one-click renewal function that skips complex payment steps.
[0060] S504: If the value level is the fourth level, then push a comparison report of competitor content to the target users as a churn intervention strategy.
[0061] If the value level is the fourth level (users with a tendency to churn to competitors), then push a comparison report of competitor content to the target users as a churn intervention strategy. The report clearly states the differentiated advantages of this platform and competitors in terms of content library, update frequency, membership rights and interests, etc., and at the same time activates cross-platform membership interoperability rights and interests (such as OTT membership附赠 by telecom broadband).
[0062] Among them, users with a tendency to churn to competitors refer to users in the OTT platform who have shown potential intentions or behavioral characteristics of transferring to competitor platforms and have a relatively high churn risk.
[0063] Figure 6This is the sixth flowchart illustrating the user churn warning method provided in the embodiments of this application, as shown below. Figure 6 As shown in step S102, after generating the churn intervention strategy for the target user, the method of this application further includes: S601. Implement churn intervention strategies targeting specific users.
[0064] In accordance with the tiered intervention strategies corresponding to S501~S504, target users are reached through APP push, SMS notification, customer service phone calls, etc., and various intervention measures are implemented.
[0065] S602. Obtain the target user's Internet TV system usage data after implementing the churn intervention strategy.
[0066] Collect user data such as login frequency, viewing time, paid conversion rate, and sharing behavior after intervention to form an intervention effect dataset.
[0067] S603. Update the churn warning model based on internet TV system usage data.
[0068] Based on the A / B testing framework, the Thompson Sampling algorithm was used to allocate traffic between the control group and the experimental group in a 1:3 ratio, and the SHAP value was used to quantify the contribution of each intervention measure. Combined with the intervention effect data, the parameters of the churn warning model were iteratively updated weekly to optimize feature weights and warning thresholds, thereby improving the model's prediction accuracy and the adaptability of intervention strategies.
[0069] The following will continue to explain the apparatus, device and storage medium for implementing the user churn warning method provided in any of the above embodiments of this application. The specific implementation process and the resulting technical effects are the same as those in the corresponding method embodiments. For the sake of brevity, the parts not mentioned in the following embodiments can be referred to the corresponding content in the method embodiments.
[0070] Figure 7 This is a schematic diagram of the user churn warning device provided in the embodiments of this application, as shown below. Figure 7 As shown, this application also provides a user churn early warning device, including: The acquisition module 10 is used to acquire the target user's behavioral data and device attribute data for the Internet TV system at multiple historical times.
[0071] Processing module 20 is used to obtain user time series data of the target user based on behavioral data from multiple historical times and device attribute data.
[0072] The prediction module 30 is used to make predictions based on the user time series data using a churn warning model to obtain the churn probability and churn reasons of the target user.
[0073] Optionally, the acquisition module 10 is further configured to acquire the text comments corresponding to the target user; the processing module 20 is further configured to use a preset multimodal feature fusion module to perform semantic association between the text comments and the user time series data to obtain fused target time series data; the prediction module 30 is further configured to use the churn warning model to make predictions based on the target time series data to obtain the churn probability and churn reasons of the target user.
[0074] Optionally, the churn warning model includes: a feature extraction module, a behavior capture module, a relationship analysis module, and an output module; the prediction module 30 is further configured to: obtain a time-series feature vector using the feature extraction module based on the user's time-series data; capture behavior patterns using the behavior capture module based on the time-series feature vector; analyze user device interaction relationships using the relationship analysis module based on the time-series feature vector; and obtain the churn probability and churn reasons of the target user using the output module based on the behavior patterns and the user device interaction relationships.
[0075] Optionally, the device further includes a generation module, configured to generate a churn intervention strategy for the target user based on the target user's value level and the reason for churn if the churn probability of the target user is greater than or equal to a preset churn probability threshold corresponding to the value level of the target user.
[0076] Optionally, the generation module is further configured to: if the value level is the first level, push preset subsidy information to the target user as the churn intervention strategy; if the value level is the second level, push social sharing activities to the target user as the churn intervention strategy; if the value level is the third level, push limited-time trial activities to the target user as the churn intervention strategy; and if the value level is the fourth level, push competitor content comparison reports to the target user as the churn intervention strategy.
[0077] Optionally, the processing module 20 is further configured to acquire behavioral weights for multiple historical time periods; adjust the behavioral data of the multiple historical time periods according to the behavioral weights of the multiple historical time periods to obtain adjusted behavioral data; and obtain the user time series data of the target user according to the adjusted behavioral data and the device attribute data.
[0078] Optionally, the device further includes an update module for executing a churn intervention strategy for the target user; acquiring internet TV system usage data of the target user after executing the churn intervention strategy; and updating the churn warning model based on the internet TV system usage data.
[0079] The above-described device is used to execute the method provided in the foregoing embodiments, and its implementation principle and technical effect are similar, so they will not be described again here.
[0080] These modules can be one or more integrated circuits configured to implement the above methods, such as one or more Application Specific Integrated Circuits (ASICs), one or more microprocessors, or one or more Field Programmable Gate Arrays (FPGAs). Alternatively, when a module is implemented using processing element scheduler code, the processing element can be a general-purpose processor, such as a Central Processing Unit (CPU) or other processor capable of calling program code. Furthermore, these modules can be integrated together as a system-on-a-chip (SOC).
[0081] Figure 8 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application, such as... Figure 8 As shown, this application also provides an electronic device, including a processor 100, a storage medium 200, and a bus 300. The storage medium stores program instructions executable by the processor. When the electronic device is running, the processor communicates with the storage medium via the bus, and the processor executes the program instructions to implement the user churn warning method described in any of the above embodiments.
[0082] This application also provides a readable storage medium storing program instructions, which, when executed by a processor, implement the user churn warning method described in any of the above embodiments.
[0083] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.
[0084] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0085] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in a combination of hardware and software functional units.
[0086] The integrated units implemented as software functional units described above can be stored in a computer-readable storage medium. These software functional units, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute some steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0087] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A user churn early warning method, characterized in that, include: Acquire target user behavior data and device attribute data for the Internet TV system over multiple historical periods; Based on behavioral data from multiple historical time periods and device attribute data, the user time series data of the target user is obtained; Based on the user time-series data, a churn prediction model is used to predict the churn probability and reasons for the churn of the target user.
2. The method according to claim 1, characterized in that, Before using the churn prediction model based on the user time-series data to predict the churn probability and reasons for churn of the target user, the method further includes: Obtain the text comments corresponding to the target user; A preset multimodal feature fusion module is used to semantically associate the text comments and the user time-series data to obtain the fused target time-series data; The step of using a churn prediction model based on the user time-series data to predict the churn probability and reasons for churn of the target user includes: Based on the target time-series data, the churn prediction model is used to predict the churn probability and reasons for the churn of the target user.
3. The method according to claim 1, characterized in that, The churn prediction model includes: a feature extraction module, a behavior capture module, a relationship analysis module, and an output module; the step of using the churn prediction model to predict the churn probability and reasons for churn of the target user based on the user time-series data includes: Based on the user time-series data, the feature extraction module is used to obtain a time-series feature vector; Based on the temporal feature vector, the behavior capture module is used to capture behavior and obtain the behavior pattern; Based on the time-series feature vector, the relationship analysis module is used to analyze and obtain the user equipment interaction relationship; Based on the behavioral patterns and the user device interaction relationships, the output module is used to obtain the churn probability and churn reasons of the target user.
4. The method according to claim 1, characterized in that, The method further includes: If the churn probability of the target user is greater than or equal to the preset churn probability threshold corresponding to the value level of the target user, then a churn intervention strategy for the target user is generated based on the value level of the target user and the reason for churn.
5. The method according to claim 4, characterized in that, The step of generating a churn intervention strategy for the target user based on the target user's value level and the reason for churn includes: If the value level is the first level, then the preset subsidy amount information is pushed to the target user as the churn intervention strategy; If the value level is the second level, then a social sharing campaign is pushed to the target user as the churn intervention strategy. If the value level is level three, a limited-time trial activity will be pushed to the target user as the churn intervention strategy. If the value level is level four, a competitor content comparison report will be pushed to the target user as part of the churn intervention strategy.
6. The method according to claim 1, characterized in that, The step of obtaining the target user's time-series data based on behavioral data from multiple historical time periods and device attribute data includes: Obtain the behavioral weights for multiple historical time periods; Based on the behavioral weights of the multiple historical time periods, the behavioral data of the multiple historical time periods are adjusted to obtain the adjusted behavioral data; Based on the adjusted behavioral data and the device attribute data, the user time series data of the target user is obtained.
7. The method according to claim 4, characterized in that, After generating the churn intervention strategy for the target user, the method further includes: Implement churn intervention strategies targeting the aforementioned user group; After the churn intervention strategy is implemented, the target user's Internet TV system usage data is obtained; The churn warning model is updated based on the usage data of the Internet TV system.
8. A user churn early warning device, characterized in that, include: The acquisition module is used to acquire behavioral data and device attribute data of the target user on the Internet TV system over multiple historical periods. The processing module is used to obtain the user time series data of the target user based on the behavioral data of multiple historical times and device attribute data; The prediction module is used to predict the churn probability and reasons for churn of the target user based on the user time series data using a churn warning model.
9. An electronic device, characterized in that, include: The device includes a processor, a storage medium, and a bus, wherein the storage medium stores program instructions executable by the processor. When the electronic device is running, the processor communicates with the storage medium via the bus, and the processor executes the program instructions to implement the user churn warning method according to any one of claims 1 to 7.
10. A readable storage medium, characterized in that, The readable storage medium stores program instructions, which, when executed by a processor, implement the user churn warning method according to any one of claims 1 to 7.