Intelligent insurance business precise promotion and customer behavior analysis system and method

Abstract

The invention provides an intelligent insurance service precise promotion and customer behavior analysis system and method, and belongs to the technical field of data mining and insurance service promotion. The method comprises the steps of obtaining an interaction data set of a target customer in a current period; inputting the interaction data set into a pre-trained behavior prediction model, and predicting an interaction behavior mode of the target customer in the next period; based on the interaction behavior mode, determining an interaction mode of pushing the insurance service to the target client in the next period; the interaction mode comprises pause pushing or intelligent pushing; when it is determined that the interaction mode is intelligent pushing, a time period and a pushing mode for pushing the insurance service to the target client are determined, and the pushing mode comprises one of voice pushing, character pushing, picture pushing and video pushing. The method further comprises the step of generating insurance service promotion content corresponding to the intelligent push mode. According to the method, invalid promotion can be reduced, the promotion effect is continuously optimized through data analysis, and the conversion rate and customer satisfaction of insurance services are improved.

Description

Technical Field

[0001] This invention belongs to the field of data mining and insurance business promotion technology, and in particular relates to a system and method for precise promotion of intelligent insurance business and analysis of customer behavior. Background Technology

[0002] With the rapid development of artificial intelligence and big data technologies, the marketing model of the insurance industry is undergoing profound changes. Intelligent insurance business precision promotion, relying on data collection, user profiling, and large-scale model training, is gradually becoming the mainstream marketing method in the industry.

[0003] At the data collection level, after complying with relevant laws and regulations and obtaining user consent, insurance institutions can leverage multi-channel data collection technologies to integrate structured and unstructured data such as basic user information, historical insurance records, online browsing behavior, and geographical location. Through big data cleaning, noise reduction, and standardization, high-quality datasets are formed. Cluster analysis and association rule mining techniques are then used to extract core characteristics of users, such as risk preferences, consumption habits, and protection needs. For example, by analyzing a car owner's driving behavior data, their risk level can be determined, and targeted car insurance products can be recommended; or personalized insurance plans can be customized based on health insurance users' medical examination reports and medical records. Ultimately, this creates a visualized, multi-dimensional user profile, laying the foundation for accurately reaching target customers. Examples include the business recommendation method based on CRM customer profiles proposed in Chinese invention patent application CN202510389788, and the automatic push method for SOP private domain marketing information based on user behavior proposed in Chinese authorized invention patent CN116862592B.

[0004] However, in these technologies, the targeting of business marketing is weak, and high-frequency business marketing messages are prone to causing user resentment; while low-frequency promotional marketing messages often fail to achieve the desired effect. Furthermore, these technologies do not consider how users prefer to receive information. For example, users may not want to receive marketing messages at certain times; at other times, users may only want to view silent messages such as text messages and not voice or video messages. Sending the wrong type of message at the wrong time (e.g., sending a voice message during a silent period) will also cause user resentment and fail to elicit user feedback. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention proposes a system and method for precise promotion of intelligent insurance business and analysis of customer behavior.

[0006] In a first aspect of the invention, a customer behavior analysis method is proposed, which is applied to the target customer group of an insurance business.

[0007] The method includes:

[0008] Obtain the interaction dataset of the target customer within the current period;

[0009] The interaction dataset is input into a pre-trained behavior prediction model to predict the interaction behavior pattern of the target customer in the next cycle.

[0010] Based on the interaction behavior pattern, determine the interaction method for pushing insurance business to the target customer in the next cycle;

[0011] The interaction methods include: pausing push notifications or smart push notifications;

[0012] When the interaction method is determined to be intelligent push, the method further includes:

[0013] The time period and method for pushing insurance services to the target customers are determined, and the push method includes one of voice push, text push, image push, and video push.

[0014] The voice push notifications include human phone calls, robot phone calls, personal WeChat voice messages, and official WeChat voice messages; the text push notifications include APP text message push notifications, human SMS text message push notifications, robot SMS text message push notifications, personal WeChat text messages, and official WeChat text messages; the image push notifications include APP image push notifications, personal WeChat image push notifications, and official WeChat image push notifications; and the video push notifications include APP video push notifications, personal WeChat video push notifications, and official WeChat video push notifications.

[0015] The interaction dataset of the target customer within the current period includes:

[0016] The target customer's human-computer interaction patterns at different time periods within the current period, including human-computer text dialogue, human-computer voice dialogue, video browsing, APP activity, and APP muting.

[0017] In a second aspect of the present invention, a method for precise promotion of intelligent insurance business is proposed, the method being applied to the target customer group of insurance business, the method comprising the following steps:

[0018] Determine whether the data collection and update cycle has been reached; if so, collect the current insurance business interaction dataset of the target customer.

[0019] Determine whether the current insurance business interaction dataset has been updated. If so, update the current user profile of the target customer based on the updated insurance business interaction dataset to obtain the updated user profile.

[0020] Using the method described in the first aspect, determine the interaction method for pushing insurance business to the target customer in the next cycle;

[0021] When the interaction method is intelligent push, insurance business promotion content corresponding to the intelligent push method is generated based on the updated user profile.

[0022] The insurance business promotion content is pushed to the target customers using the aforementioned intelligent push method.

[0023] When the interaction method is to pause push notifications, the data collection and update cycle is increased.

[0024] The process of generating insurance business promotion content corresponding to the intelligent push method based on the updated user profile specifically includes:

[0025] When the intelligent push method is voice push, insurance business promotion content is generated in voice format;

[0026] When the intelligent push method is text push, it generates insurance business promotion content in text form;

[0027] When the intelligent push method is image push, insurance business promotion content is generated in the form of an image.

[0028] When the intelligent push method is video push, insurance business promotion content is generated in video format.

[0029] In a third aspect of the invention, a customer behavior analysis system is provided, the system comprising:

[0030] An interactive dataset acquisition unit is used to acquire the interactive dataset of the target customer within the current period according to a preset data collection period.

[0031] An interaction behavior pattern prediction unit is used to input the interaction dataset into a pre-trained behavior prediction model to predict the interaction behavior pattern of the target customer in the next cycle.

[0032] An interaction method determination unit is used to determine, based on the interaction behavior pattern, the interaction method for pushing insurance business to the target customer in the next cycle;

[0033] A feedback unit is used to detect feedback information received by the target customer from the insurance business, and to retrain the behavior prediction model based on the feedback information;

[0034] The interaction methods include: pausing push notifications or smart push notifications;

[0035] When the interaction method is determined to be intelligent push, the interaction method determination unit further determines the time period and push method for pushing insurance business to the target customer. The push method includes one of voice push, text push, image push, and video push.

[0036] The system also includes:

[0037] First cycle adjustment unit;

[0038] The first cycle adjustment unit is connected to the interaction behavior pattern prediction unit. When the similarity between the interaction behavior pattern of the target customer in the next cycle and the interaction behavior pattern in the previous cycle predicted by the interaction behavior pattern prediction unit is greater than a preset value, the first cycle adjustment unit adjusts the preset data collection cycle.

[0039] In a fourth aspect of the invention, a smart insurance business precision promotion system is also proposed, the system comprising:

[0040] The first judgment unit determines whether the data collection and update cycle has been reached. If so, it collects the current insurance business interaction dataset of the target customer.

[0041] The second judgment unit determines whether the current insurance business interaction dataset has been updated. If so, it updates the current user profile of the target customer based on the updated insurance business interaction dataset to obtain the updated user profile.

[0042] Using the interaction behavior pattern prediction unit in the customer behavior analysis system described in the third aspect, the interaction method for pushing insurance business to the target customer in the next cycle is determined;

[0043] The profile update unit generates insurance business promotion content corresponding to the intelligent push method based on the updated user profile when the interaction method is intelligent push method.

[0044] The push unit uses the intelligent push method to push the insurance business promotion content to the target customer.

[0045] The system also includes:

[0046] Second cycle adjustment unit;

[0047] The second cycle adjustment unit is connected to the interaction behavior pattern prediction unit;

[0048] When the interaction behavior pattern prediction unit predicts that the interaction mode is to pause push, the second cycle adjustment unit increases the data collection and update cycle.

[0049] Compared to existing technologies, the customer behavior analysis method proposed in this application, driven by dynamic data and intelligent decision-making, significantly improves the accuracy of insurance marketing and user experience. By acquiring the current period's interaction dataset and inputting it into the behavior prediction model, it captures the latest user needs in real time and proactively predicts the behavior patterns of the next period, avoiding strategy lag caused by relying on static historical data. A refined classification decision mechanism of "pause push / intelligent push" is introduced to reduce interference when users have a weak intention to receive information or are unable to receive it, while accurately reaching high-intent users and balancing promotion frequency with user acceptance. Furthermore, it dynamically matches push times and multimodal formats (voice, text, images, video) based on user behavior patterns and APP status rather than the device's own status, avoiding conflicts between information format and context.

[0050] Building upon this foundation, the present invention further proposes an intelligent insurance business precision promotion method that significantly optimizes insurance marketing efficiency through dynamic data-driven and intelligent decision-making mechanisms. On one hand, it sets a data collection and update cycle and monitors data changes in real time, promptly capturing the latest interactive behaviors of target customers. Based on the updated interaction dataset, it dynamically refreshes user profiles, ensuring the timeliness and accuracy of user characteristic and demand analysis, and avoiding the strategy lag caused by traditional static profiles. On the other hand, it constructs an intelligent tiered push strategy, determining the interaction method based on a behavior prediction model: intelligent pushes are triggered for high-intent customers, and multimodal promotional content (voice, text, etc.) is customized based on the updated user profile to achieve precise matching of "content-form-demand"; pushes are paused for low-intent customers with little data change, and the data collection cycle is extended to reduce ineffective outreach and lower the risk of user aversion.

[0051] Therefore, the technical solution of this invention combines real-time updated user profiles to generate insurance business promotion content adapted to different push formats, achieving deep coupling of "content-format-scenario". The entire process forms a closed loop of "data collection-behavior prediction-intelligent push-effect feedback", continuously optimizing strategies and solving problems such as data lag, rigid strategies, and poor user experience in traditional marketing, thereby improving marketing efficiency and conversion rate.

[0052] Further specific advantages and implementation principles of the present invention will be further detailed in the specific embodiments section in conjunction with the accompanying drawings. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0054] Figure 1This is a schematic diagram of the main execution flow of a customer behavior analysis method according to an embodiment of the present invention;

[0055] Figure 2 This is a schematic diagram outlining the principle of determining the interaction method for insurance business based on the interaction dataset and interaction behavior patterns of target customers.

[0056] Figure 3 This is a schematic diagram of the main execution flow of a method for precise promotion of intelligent insurance business according to an embodiment of the present invention;

[0057] Figure 4 This is a schematic diagram of the hardware unit composition of a customer behavior analysis system according to an embodiment of the present invention;

[0058] Figure 5 This is a schematic diagram of the hardware unit composition of an intelligent insurance business precision promotion system according to an embodiment of the present invention. Detailed Implementation

[0059] In the specific embodiments of this application, if the embodiments of the relevant technical solutions involve user-related data, then when the embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0060] As described in the background section, existing intelligent insurance business precision promotion relies on data collection, user profile construction, and large model training, and is gradually becoming the mainstream marketing method in the industry.

[0061] At the data collection level, after obtaining user permission or consent and ensuring that the collection, use, and processing of relevant data comply with the relevant laws, regulations, and standards of the country and region, insurance institutions leverage multi-channel data collection technologies to integrate structured and unstructured data such as basic user information, historical insurance records, online browsing behavior, and geographical location. These data sources are diverse, covering official websites, apps, and third-party platforms, providing rich material for customer behavior analysis. Through big data cleaning, noise reduction, and standardization, high-quality datasets are formed to ensure the accuracy of subsequent analysis.

[0062] User profiling is a crucial step in precision marketing. Based on cleaned data, techniques such as cluster analysis and association rule mining are used to extract core characteristics of users, including risk preferences, consumption habits, and protection needs. For example, by analyzing a driver's driving behavior data, their risk level can be determined, and targeted car insurance products can be recommended; or personalized insurance plans can be customized based on health insurance users' medical examination reports and medical records. Ultimately, this creates a visualized, multi-dimensional user profile, laying the foundation for accurately reaching target customers.

[0063] The establishment and application of large-scale models are the core driving force behind precision marketing. Through deep learning algorithms, such as Transformer and Graph Neural Networks (GNNs), massive amounts of data are trained to uncover potential correlations between user behavior patterns and insurance needs. These models possess powerful predictive capabilities, able to anticipate user purchase intentions, identify potential risks, and even simulate the effects of different marketing strategies.

[0064] The system can automatically match the most suitable insurance products, promotional copy, and outreach channels based on user profiles and model predictions, avoiding ineffective marketing. Simultaneously, by monitoring promotional effectiveness in real time and collecting user feedback, it continuously optimizes model parameters, forming a closed loop of "data-analysis-promotion-feedback-optimization," driving insurance marketing towards intelligence and precision.

[0065] However, the inventors discovered that the aforementioned technical methods still have the following problems:

[0066] (1) Existing user profiles mostly rely on static tags such as historical transaction data (e.g., insurance records) and basic attributes (e.g., age, occupation), lacking dynamic collection of real-time user scenario data (e.g., current geographical location, APP activity status, schedule). For example:

[0067] User calendar data (meeting times, vacation schedules) is not integrated, making it impossible to identify "APP silent periods" (such as office scenarios from 9:00 to 18:00 on weekdays).

[0068] In existing technologies, the methods for pushing relevant messages only consider the device itself (such as the phone being in silent mode or do not disturb mode), which is relatively simplistic. In reality, the phone being in silent mode only indicates that the user does not want to receive video or voice messages, but there is still a significant possibility that the user will continue to operate the phone and engage in text interaction.

[0069] (2) Existing models lack scenario-based training data and are mostly based on binary classification training of "whether to receive messages" without being refined to the combination of "time period-format" scenarios (such as "preferring text push at 8 am on weekdays and accepting video push at 7 pm on weekends").

[0070] For example, the health insurance promotion model did not distinguish between users who "prefer short texts during their commute" and "accept detailed video explanations during their rest periods," and uniformly pushed long videos, which caused resentment.

[0071] In response, this invention proposes a corresponding improved technical solution.

[0072] In various embodiments of the present invention, different expressions may be used for the same technical feature due to different descriptive scenarios. For example, from a business perspective, "target customer" may be used; from a neutral perspective, "target user" may be used. Therefore, "target customer" and "target user" have the same meaning. Those skilled in the art can accurately understand their meaning based on the context, and it will not cause any ambiguity in understanding. The same applies to other similar terms.

[0073] See Figure 1 , Figure 1 This is a schematic diagram of the main execution flow of a customer behavior analysis method according to an embodiment of the present invention.

[0074] Figure 1 The method includes the following steps:

[0075] Obtain the interaction dataset of the target customer within the current period;

[0076] The interaction dataset is input into a pre-trained behavior prediction model to predict the interaction behavior pattern of the target customer in the next cycle.

[0077] Based on the interaction behavior pattern, determine the interaction method for pushing insurance business to the target customer in the next cycle;

[0078] The interaction methods include: pausing push notifications or smart push notifications;

[0079] When the interaction method is determined to be intelligent push, the method further includes:

[0080] The time period and method for pushing insurance services to the target customers are determined, and the push method includes one of voice push, text push, image push, and video push.

[0081] In one specific embodiment, the acquired interaction dataset of the target customer within the current period includes:

[0082] The target customer's human-computer interaction patterns at different time periods within the current period, including human-computer text dialogue, human-computer voice dialogue, video browsing, APP activity, and APP muting.

[0083] Unlike existing technologies that only consider the single mechanical processing of device muting (corresponding to no push messages), the technical solution of this invention is specifically refined to collecting human-computer interaction pattern data of target customers in different time periods within the current cycle, and then inputting the interaction dataset into a pre-trained behavior prediction model to predict the interaction behavior pattern of the target customer in the next cycle.

[0084] As an example, Figure 2This is a schematic diagram outlining the principle of determining the interaction method for insurance business based on the interaction dataset and interaction behavior patterns of target customers.

[0085] Figure 2 The target customer's (historical) interaction dataset includes user interaction behavior patterns data for different time periods within a preset period (e.g., weekly) and preset dates (e.g., Monday to Sunday);

[0086] Figure 2 It is shown that:

[0087] Monday: 9:00–10:30

[0088] Mute the app and enable text chat;

[0089] Tuesday: 10:00–12:30

[0090] Mute the app;

[0091] Friday: 14:00-17:00

[0092] Mute the app and enable voice chat.

[0093] Sunday: 20:00-23:00

[0094] App activity, voice conversations, video browsing

[0095] In the example above, during the current period, it was identified that the user silences the target app (including but not limited to WeChat, QQ, insurance apps, etc.) every Monday from 9:00 to 10:30. However, frequent text conversations still occur, and these text conversations may continue to occur on the target app. In other words, the target app itself is active (but silent), meaning that the user does not want to receive non-silent messages during this time, such as voice messages, phone messages, video messages, etc., and prefers to view text messages or image messages.

[0096] As an example, suppose that based on the interaction dataset of a target customer within a certain period, it is predicted that the target user will mute the target app (including but not limited to WeChat, QQ, insurance business apps, etc.) between 9:00 and 10:30 on Monday of the following week, but there are frequent text conversations that may continue to occur on the target app. Then, the interaction methods for pushing insurance business to the target customer in the next period are determined to include:

[0097] The interaction method for pushing insurance business to the target customers between 9:00 and 10:30 on Monday of next week will be intelligent push, and the push method will be either text push or video push.

[0098] exist Figure 2 The text illustrates the intelligent push method of "Monday: Text Promotional Content", which should be more specifically "Text Promotional Content from 9:00 to 10:30 on Mondays".

[0099] This adjustment method adapts to the user's commuting or work status. For example, "Monday 9:00-10:30" may be the user's meeting time after arriving at work. The user is obviously not in a convenient position to answer calls, voice or video calls, so they will mute the device; however, they can still operate the phone to have text conversations, including typing and browsing text.

[0100] Furthermore, in Figure 2 In this context, assuming that based on the interaction dataset of target customers within a certain period, it is predicted that the target user will mute the target app (including but not limited to WeChat, QQ, insurance business apps, etc.) between 10:00 and 12:30 on Tuesday of the following week and then perform no other operations, then the interaction methods for pushing insurance business to the target customer in the next period are determined to include:

[0101] Push notifications will be suspended from 10:00 to 12:30 on Tuesday next week.

[0102] exist Figure 2 The text illustrates the "Tuesday: Pause Push Notification" smart push notification method, which should more specifically mean "Pause Push Notification from 10:00 to 12:30 on Tuesdays".

[0103] This adjustment method adapts to users' training, learning, or other situations where they need to avoid device interference. For example, "Tuesday 10:00–12:30" may be a concentrated training period, exam period, or other situation where users need to avoid device interference.

[0104] Furthermore, in Figure 2 In this context, assuming that based on the interaction dataset of target customers within a certain period, it is predicted that the target user will mute the target app (including but not limited to WeChat, QQ, insurance business apps, etc.) between 14:00 and 17:00 on Friday of the following week, but frequent voice interaction will still occur, then the interaction methods for pushing insurance business to the target customer in the next period are determined to include:

[0105] The message will be sent via voice message between 2:00 PM and 5:00 PM next Friday.

[0106] exist Figure 2 The text illustrates the intelligent push method for "Friday: Voice Promotional Content," which should more specifically be "pushing content via voice dialogue from 14:00 to 17:00 on Fridays."

[0107] In this situation, even if the user has muted the target app, there are still voice interaction conversation records, indicating that the user is willing to receive voice conversations at this time.

[0108] Similarly, in Figure 2 In this context, assuming that based on the interaction dataset of target customers within a certain period, it is predicted that the target user will be active on a target app (including but not limited to WeChat, QQ, insurance business apps, etc.) between 20:00 and 23:00 on Sunday of the following week, and will exhibit frequent voice conversations and video browsing behavior, then the interaction methods for pushing insurance business to the target customer in the next period are determined to include:

[0109] Push notifications will be sent via voice and video between 20:00 and 23:00 next Sunday.

[0110] exist Figure 2 The text illustrates the intelligent push method for "Sunday: Video Promotional Content," which should more specifically mean "pushing content via video from 20:00 to 23:00 on Sundays."

[0111] In the specific execution of content delivery, as a further optimization, Figure 1 In the method, the voice push includes human phone push, robot phone push, personal WeChat voice push, and official account WeChat voice push; the text push includes APP text message push, human SMS text push, robot SMS text push, personal WeChat text push, and official account WeChat text push; the image push includes APP image push, personal WeChat image push, and official account WeChat image push; and the video push includes APP video push, personal WeChat video push, and official account video push.

[0112] It is understandable that the sending (pushing) entities for "human telephone push", "personal WeChat voice push", "human SMS text push", "personal WeChat text push", "personal WeChat image push", and "personal WeChat video push" are human customer service representatives, while the sending (pushing) entities for other methods are robots or automated system backends.

[0113] This further distinction is made as one of the improved embodiments of the present invention.

[0114] In practical applications, the inventors further discovered that some (introverted) users are accustomed to receiving promotional messages in a one-way manner, that is, they only receive promotional messages but do not want to connect with a specific person or be disturbed by calls / messages from a specific contact person. For such users, the system prioritizes sending corresponding messages automatically in the background using robots or the system, and embeds automatic feedback options in the messages, such as clicking links or automatically replying with keywords, to help users trigger subsequent feedback.

[0115] At this time, the voice push is delivered via robot phone call or WeChat official account voice push; the text push is delivered via APP text message push, robot SMS text push, or WeChat official account text push; the image push is delivered via APP image push or WeChat official account image push; and the video push is delivered via APP video push or WeChat official account video push.

[0116] In summary, the method further includes: inputting the interaction dataset into a pre-trained behavior prediction model to predict the interaction behavior pattern of the target customer in the next cycle, wherein the interaction behavior pattern includes the user's message interaction habits, and the message interaction habits are robot interaction or human interaction;

[0117] When the message interaction habit is robot interaction, the voice push is implemented through robot phone calls or WeChat official account voice push; the text push is implemented through APP text message push, robot SMS text push, or WeChat official account text push; the image push is implemented through APP image push or WeChat official account image push; and the video push is implemented through APP video push or WeChat official account video push.

[0118] Correspondingly, some (extroverted) users may find the mechanical interaction of robots averse and prefer direct communication with actual people. For these users, priority will be given to sending messages via dedicated personnel (human customer service). After receiving the message, the user can reply directly using natural language interaction habits to learn more details.

[0119] At this time, the voice push is a manual phone call or a personal WeChat voice push; the text push is a manual SMS text push or a personal WeChat text push; the image push is a personal WeChat image push; and the video push is a personal WeChat video push.

[0120] Therefore, the method further includes:

[0121] When the message interaction habit is human interaction, the voice push is a human telephone push or a personal WeChat voice push; the text push is a human SMS text push or a personal WeChat text push; the image push is a personal WeChat image push; and the video push is a personal WeChat video push.

[0122] In summary, the method further includes: determining the interaction method for pushing insurance business to the target customer in the next cycle based on the interaction behavior pattern and the message interaction habits.

[0123] therefore, Figure 1 A further preferred embodiment of the method includes:

[0124] Obtain the interaction dataset of the target customer within the current period;

[0125] The interaction dataset is input into a pre-trained behavior prediction model to predict the target customer's interaction behavior patterns and habits in the next cycle.

[0126] Based on the aforementioned interaction patterns and habits, the interaction method for pushing insurance services to the target customer in the next cycle is determined.

[0127] The interaction methods include: pausing push notifications or smart push notifications;

[0128] When the interaction method is determined to be intelligent push, the method further includes:

[0129] The time period and method for pushing insurance services to the target customers are determined, and the push method includes one of voice push, text push, image push, and video push.

[0130] Specifically, when the interaction method is determined to be intelligent push and the message interaction habit is human interaction, the voice push is a human telephone push or a personal WeChat voice push; the text push is a human SMS text push or a personal WeChat text push; the image push is a personal WeChat image push; and the video push is a personal WeChat video push.

[0131] When the interaction method is determined to be intelligent push and the message interaction habit is robot interaction, the voice push is implemented through robot phone calls or WeChat official account voice push; the text push is implemented through APP text message push, robot SMS text push, or WeChat official account text push; the image push is implemented through APP image push or WeChat official account image push; and the video push is implemented through APP video push or WeChat official account video push.

[0132] In summary, Figure 1The customer behavior analysis method described above significantly enhances the personalization and user experience of insurance marketing by deepening the analysis of user interaction habits and multimodal intelligent adaptation. Based on a behavior prediction model, it accurately identifies user message interaction preferences (robot interaction or human interaction) and constructs a differentiated push strategy: for introverted users who prefer automated interaction, non-contact channels such as robot calls and app messages are used, with embedded automatic feedback options (such as clicking links and keyword replies) to meet their one-way reception needs and reduce human interference; for extroverted users who prefer natural communication, human customer service representatives are matched to proactively connect through personalized channels such as WeChat and telephone, supporting natural language interaction and avoiding mechanical responses that may cause resentment. This "human-machine collaborative" intelligent distribution mechanism respects user communication habits while ensuring information transmission efficiency, achieving a balance between "technological accuracy" and "human warmth," improving user acceptance and conversion rates, while reducing the consumption of human customer service resources and optimizing the marketing cost structure.

[0133] exist Figure 1 The method accurately identifies interaction behavior patterns and message interaction habits. Figure 3 The diagram further illustrates the main execution flow of a method for precise promotion of intelligent insurance business according to an embodiment of the present invention.

[0134] Figure 3 The method includes the following steps:

[0135] Determine whether the data collection and update cycle has been reached; if so, collect the current insurance business interaction dataset of the target customer.

[0136] Determine whether the current insurance business interaction dataset has been updated. If so, update the current user profile of the target customer based on the updated insurance business interaction dataset to obtain the updated user profile.

[0137] use Figure 1 The method described above determines the interaction method for pushing insurance business to the target customer in the next cycle;

[0138] When the interaction method is intelligent push, insurance business promotion content corresponding to the intelligent push method is generated based on the updated user profile.

[0139] The insurance business promotion content is pushed to the target user using the aforementioned intelligent push method.

[0140] When the interaction method is to pause push notifications, the data collection and update cycle is increased.

[0141] Preferably, the step of generating insurance business promotion content corresponding to the intelligent push method based on the updated user profile specifically includes:

[0142] When the intelligent push method is voice push, insurance business promotion content is generated in voice format;

[0143] When the intelligent push method is text push, it generates insurance business promotion content in text form;

[0144] When the intelligent push method is image push, insurance business promotion content is generated in the form of an image.

[0145] When the intelligent push method is video push, insurance business promotion content is generated in video format.

[0146] As can be seen, this intelligent insurance business precision promotion solution, driven by dynamic data and intelligent decision-making, possesses three significant advantages. First, by periodically collecting and monitoring insurance business interaction data in real time, it dynamically updates user profiles, ensuring that the profiles promptly reflect the latest customer needs and behavioral changes, significantly improving data timeliness and accuracy compared to static profiles. Second, it constructs an intelligent tiered push mechanism, triggering intelligent pushes for high-intent customers based on interaction behavior prediction results. This is combined with customized multimodal (voice, text, image, video) promotional content based on user profiles, achieving precise matching of "needs-format-content." Pushes are paused for low-intent customers or those with long data collection and update cycles, and the data collection cycle is extended to reduce unnecessary interruptions and balance promotion efficiency with user experience. Finally, the entire solution forms a closed-loop optimization system of "data collection-profile update-strategy execution-cycle adjustment," which can automatically iterate strategies, continuously improving marketing conversion rates and customer satisfaction while reducing operating costs.

[0147] exist Figures 1-3 Based on the method implementation examples, Figures 4-5 A schematic diagram of the structural unit composition of a system (product) embodiment is provided.

[0148] For details, see Figure 4 This paper illustrates a customer behavior analysis system, the system comprising:

[0149] An interactive dataset acquisition unit is used to acquire the interactive dataset of the target customer within the current period according to a preset data collection period.

[0150] An interaction behavior pattern prediction unit is used to input the interaction dataset into a pre-trained behavior prediction model to predict the interaction behavior pattern of the target customer in the next cycle.

[0151] An interaction method determination unit is used to determine, based on the interaction behavior pattern, the interaction method for pushing insurance business to the target customer in the next cycle;

[0152] A feedback unit is used to detect feedback information received by the target customer from the insurance business, and to retrain the behavior prediction model based on the feedback information;

[0153] The interaction methods include: pausing push notifications or smart push notifications;

[0154] When the interaction method is determined to be intelligent push, the interaction method determination unit further determines the time period and push method for pushing insurance business to the target customer. The push method includes one of voice push, text push, image push, and video push.

[0155] The system also includes:

[0156] First cycle adjustment unit;

[0157] The first cycle adjustment unit is connected to the interaction behavior pattern prediction unit. When the similarity between the interaction behavior pattern of the target customer in the next cycle and the interaction behavior pattern in the previous cycle predicted by the interaction behavior pattern prediction unit is greater than a preset value, the first cycle adjustment unit adjusts the preset data collection cycle.

[0158] The interactive dataset includes:

[0159] The target customer's human-computer interaction patterns at different time periods within the current period, including human-computer text dialogue, human-computer voice dialogue, video browsing, APP activity, APP muting, and message interaction habits, including human interaction or robot interaction;

[0160] Preferably, the interaction behavior pattern prediction unit is used to input the interaction dataset into a pre-trained behavior prediction model to predict the target customer's interaction behavior pattern and message interaction habits in the next cycle.

[0161] The interaction method determination unit is used to determine, based on the interaction behavior pattern and message interaction habits, the interaction method for pushing insurance business to the target customer in the next cycle, specifically including:

[0162] When the interaction method is determined to be intelligent push and the message interaction habit is human interaction, the voice push is a human telephone push or a personal WeChat voice push; the text push is a human SMS text push or a personal WeChat text push; the image push is a personal WeChat image push; and the video push is a personal WeChat video push.

[0163] When the interaction method is determined to be intelligent push and the message interaction habit is robot interaction, the voice push is implemented through robot phone calls or WeChat official account voice push; the text push is implemented through APP text message push, robot SMS text push, or WeChat official account text push; the image push is implemented through APP image push or WeChat official account image push; and the video push is implemented through APP video push or WeChat official account video push.

[0164] Preferably, the feedback unit detects feedback information received by the target customer from the insurance business, and retrains the behavior prediction model based on the feedback information, specifically including:

[0165] (1) Multi-source feedback information acquisition and structured processing

[0166] Explicit feedback: Directly collect user actions (such as clicking links, replying with keywords, and online inquiries), analyze the inquiry intent through natural language processing (NLP), and extract key demand characteristics (such as coverage amount and insurance type).

[0167] Implicit Feedback: Analyze data that reflects users' implicit attitudes but not their direct responses (such as message reading time and push notification open rates) to identify behavioral pattern shifts through time series analysis.

[0168] Heterogeneous data fusion: Linking user feedback with external data sources (such as weather, holidays, and industry events) to construct multi-dimensional feature vectors;

[0169] (2) Real-time model training and incremental learning framework

[0170] Online learning engine: A real-time data stream processing platform built with Flink / Kafka is used to achieve minute-level model updates;

[0171] Incremental training strategy: Only fine-tune the model parameters that are relevant to the feedback, and retain the historical training results.

[0172] (3) Reinforcement learning-driven decision optimization

[0173] Reward function design: Quantify metrics such as user conversion rate, dwell time, and NPS score into model reward signals.

[0174] This feedback mechanism enables the system to "self-evolve".

[0175] Data testing shows that after a large property insurance company implemented a similar solution, user click-through rates increased by 32%, manual intervention rates decreased by 41%, and model prediction accuracy improved by 0.5-1% weekly. By continuously learning about the evolution of user preferences, the system has upgraded from "passive response" to "proactive guidance," significantly improving customer acquisition and retention efficiency in an increasingly complex market environment for insurance products.

[0176] See also Figure 5 , Figure 5 This paper presents an intelligent insurance business precision promotion system, the system comprising:

[0177] The first judgment unit determines whether the data collection and update cycle has been reached. If so, it collects the current insurance business interaction dataset of the target customer.

[0178] The second judgment unit determines whether the current insurance business interaction dataset has been updated. If so, it updates the current user profile of the target customer based on the updated insurance business interaction dataset to obtain the updated user profile.

[0179] use Figure 4 The interaction behavior pattern prediction unit in the customer behavior analysis system determines the interaction method for pushing insurance business to the target customer in the next cycle.

[0180] The profile update unit generates insurance business promotion content corresponding to the intelligent push method based on the updated user profile when the interaction method is intelligent push method.

[0181] The push unit uses the intelligent push method to push the insurance business promotion content to the target user.

[0182] The system also includes:

[0183] Second cycle adjustment unit;

[0184] The second cycle adjustment unit is connected to the interaction behavior pattern prediction unit;

[0185] When the interaction behavior pattern prediction unit predicts that the interaction mode is to pause push, the second cycle adjustment unit increases the data collection and update cycle.

[0186] Similarly, this system achieves precision and intelligence in insurance marketing through modular design and dynamic feedback mechanisms. A dual-judgment unit periodically collects and monitors insurance business interaction data in real time, ensuring that user profiles are dynamically adjusted according to customer behavior, avoiding analytical biases caused by data lag. Relying on an interaction behavior pattern prediction unit, it accurately predicts push interaction methods based on profiles, distinguishing between intelligent push and paused push, balancing promotion efficiency and user experience. Multimodal promotional content is customized based on updated profiles, with the push unit achieving precise reach. When push is paused, the data collection cycle is automatically extended to reduce ineffective resource consumption. All units collaborate to form a closed-loop system of "data collection - profile update - strategy execution - cycle optimization," enabling continuous iteration of marketing strategies, improving customer conversion rates and satisfaction, while reducing operating costs.

[0187] Although not shown in the accompanying drawings, a preferred and more common product embodiment may also be an electronic device comprising: a memory and one or more processors. The memory stores one or more application programs adapted to be executed by the one or more processors using the aforementioned customer behavior analysis method or intelligent insurance business precision promotion method.

[0188] Although not shown in the accompanying drawings, more embodiments also include a computer-readable storage medium storing a computer program that, when executed, implements the steps of the aforementioned customer behavior analysis method or the intelligent insurance business precision promotion method.

[0189] It is understood that the system, product, equipment, and media implementation examples and method implementations correspond to each other and can be referenced by each other, and their principles are similar or the same, so they will not be elaborated again.

[0190] Other technologies, principles, algorithms, or models not elaborated in detail in this application can be found in the prior art.

[0191] Based on the above description of the embodiments and the test data of the technical solutions of this application in practical applications, the advantages of the present invention compared to the prior art include at least the following:

[0192] (1) Dynamic data-driven, accurate user profile: By setting a data collection and update cycle, the system monitors changes in the target customer's insurance business interaction data in real time. Once the data is updated, the user profile is optimized synchronously to ensure that the profile can reflect the latest customer behavior and needs in a timely manner. Compared with traditional static profiles, this solution effectively avoids the problem of data lag and lays a solid foundation for precision marketing.

[0193] (2) Intelligent hierarchical push, balancing efficiency and experience: Utilize behavioral prediction models to predict customer interaction patterns in the next cycle, and divide push into "pause push" and "intelligent push" accordingly. Pause push for low-intent customers and extend the data collection cycle to reduce disturbance; trigger intelligent push for high-intent customers, and combine user interaction habits to match robot or human customer service, multimodal message format, to achieve precise adaptation of "demand-format-reach method", taking into account both marketing efficiency and user experience.

[0194] (3) Closed-loop feedback optimization to continuously improve model performance: The feedback unit collects explicit (such as clicks and replies) and implicit (such as reading time) feedback information from users on insurance business promotion in real time, and uses it to retrain the behavior prediction model after structured processing. With the help of technologies such as online learning and incremental training, the model can be updated at the minute level, enabling the system to have the ability to self-evolve and continuously improve the prediction accuracy and the effectiveness of marketing strategies.

[0195] (4) Modular collaborative operation enhances system flexibility: The system's units have clear division of labor and work closely together. The first and second judgment units ensure data timeliness, the profile update unit optimizes profiles based on data, the interactive behavior pattern prediction unit drives intelligent decision-making, the push unit is responsible for accurate outreach, and the second cycle adjustment unit adjusts the collection cycle as needed. The modular design facilitates system function expansion and maintenance, and adapts to the complex and ever-changing marketing scenarios of insurance business.

[0196] The foregoing has shown and described the method embodiments and systems of the present invention, but it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A customer behavior analysis method, said method being applied to the target customer group of insurance business, characterized in that, The method includes the following steps: Obtain the interaction dataset of the target customer within the current period; The interaction dataset is input into a pre-trained behavior prediction model to predict the interaction behavior pattern of the target customer in the next cycle. Based on the interaction behavior pattern, determine the interaction method for pushing insurance business to the target customer in the next cycle; The interaction methods include: pausing push notifications or smart push notifications; When the interaction method is determined to be intelligent push, the method further includes: Determine the time period and method for pushing insurance services to the target customers, wherein the push method includes one of voice push, text push, image push, and video push.

2. The customer behavior analysis method as described in claim 1, characterized in that, The voice push includes human phone push, robot phone push, personal WeChat voice push, and official account WeChat voice push; The text push includes text message push from the APP, text push from human SMS, text push from robot SMS, text push from personal WeChat, and text push from official WeChat accounts; The image push includes APP image push, personal WeChat image push, and official account WeChat image push; The video push includes video push from the app, video push from personal WeChat accounts, and video push from official accounts.

3. The customer behavior analysis method as described in claim 1, characterized in that, The interaction dataset of the target customer within the current period includes: The target customer's human-computer interaction patterns at different time periods within the current period, including human-computer text dialogue, human-computer voice dialogue, video browsing, APP activity, and APP muting.

4. A method for precise promotion of intelligent insurance business, wherein the method is applied to the target customer group of insurance business, characterized in that, The method includes the following steps: Determine whether the data collection and update cycle has been reached; if so, collect the current insurance business interaction dataset of the target customer. Determine whether the current insurance business interaction dataset has been updated. If so, update the current user profile of the target customer based on the updated insurance business interaction dataset to obtain the updated user profile. Using the method described in any one of claims 1-3, determine the interaction method for pushing insurance business to the target customer in the next cycle; When the interaction method is intelligent push, insurance business promotion content corresponding to the intelligent push method is generated based on the updated user profile. The insurance business promotion content is pushed to the target customers using the aforementioned intelligent push method.

5. The method for precise promotion of intelligent insurance business as described in claim 4, characterized in that, When the interaction method is to pause push notifications, the data collection and update cycle is increased.

6. The method for precise promotion of intelligent insurance business as described in claim 4, characterized in that, The process of generating insurance business promotion content corresponding to the intelligent push method based on the updated user profile specifically includes: When the intelligent push method is voice push, insurance business promotion content is generated in voice format; When the intelligent push method is text push, it generates insurance business promotion content in text form; When the intelligent push method is image push, insurance business promotion content is generated in the form of an image. When the intelligent push method is video push, insurance business promotion content is generated in video format.

7. A customer behavior analysis system, characterized in that, The system includes: The interactive dataset acquisition unit is used to acquire the interactive dataset of the target customer in the current period according to a preset data collection period. An interaction behavior pattern prediction unit is used to input the interaction dataset into a pre-trained behavior prediction model to predict the interaction behavior pattern of the target customer in the next cycle. An interaction method determination unit is used to determine, based on the interaction behavior pattern, the interaction method for pushing insurance business to the target customer in the next cycle; A feedback unit is used to detect feedback information received by the target customer from the insurance business, and to retrain the behavior prediction model based on the feedback information; The interaction methods include: pausing push notifications or smart push notifications; When the interaction method is determined to be intelligent push, the interaction method determination unit further determines the time period and push method for pushing insurance business to the target customer. The push method includes one of voice push, text push, image push, and video push.

8. A customer behavior analysis system as described in claim 7, characterized in that, The system also includes: First cycle adjustment unit; The first cycle adjustment unit is connected to the interaction behavior pattern prediction unit. When the similarity between the interaction behavior pattern of the target customer in the next cycle and the interaction behavior pattern in the previous cycle predicted by the interaction behavior pattern prediction unit is greater than a preset value, the first cycle adjustment unit adjusts the preset data collection cycle.

9. A smart insurance business precision promotion system, characterized in that, The system includes: The first judgment unit determines whether the data collection and update cycle has been reached. If so, it collects the current insurance business interaction dataset of the target customer. The second judgment unit determines whether the current insurance business interaction dataset has been updated. If so, it updates the current user profile of the target customer based on the updated insurance business interaction dataset to obtain the updated user profile. Using the interaction behavior pattern prediction unit in the customer behavior analysis system of claim 7, the interaction method for pushing insurance business to the target customer in the next cycle is determined; The profile update unit generates insurance business promotion content corresponding to the intelligent push method based on the updated user profile when the interaction method is intelligent push method. The push unit uses the intelligent push method to push the insurance business promotion content to the target customer.

10. The intelligent insurance business precision promotion system as described in claim 9, characterized in that, The system also includes: Second cycle adjustment unit; The second cycle adjustment unit is connected to the interaction behavior pattern prediction unit; When the interaction behavior pattern prediction unit predicts that the interaction mode is to pause push, the second cycle adjustment unit increases the data collection and update cycle.

Images