Marketing intelligent triggering and optimizing method and device based on rule engine, equipment and medium

Abstract

The invention relates to an intelligent marketing triggering and optimizing method and device based on a rule engine, equipment and a medium. The intelligent marketing triggering and optimizing method comprises the steps that marketing user information of a user in a marketing activity which is released through first release information and has first activity content is acquired; according to marketing user information of a user in a marketing activity released through the first release information, based on a reinforcement learning algorithm, with an optimal action value as a target, adjusting a trigger rule in the marketing activity trigger expression to obtain a new marketing activity trigger expression; according to the new marketing activity trigger expression, adjusting the first activity content of the marketing activity to obtain second activity content of the marketing activity, and according to the new marketing activity trigger expression and the second activity content, determining second delivery information of the marketing activity after the activity content is adjusted; and under the condition that the new marketing activity triggering expression is satisfied, re-putting the marketing activity after the activity content is adjusted according to the second putting information.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to a marketing intelligent triggering and optimization method, device, equipment and medium based on a rule engine. Background Art

[0002] In the field of digital marketing, accurate delivery and optimization of marketing activities are the key to improving conversion rates and user engagement. Traditional marketing methods often rely on manually set trigger rules and static activity content, which is not only inefficient but also difficult to adapt to rapidly changing market demands and user behaviors. With the development of big data and artificial intelligence technologies, data-driven marketing strategies have gradually become mainstream. Among them, rule engines, as a powerful tool, can dynamically trigger and execute marketing activities according to preset rules, while machine learning algorithms, especially reinforcement learning, make it possible to continuously optimize these rules.

[0003] In the prior art, although there have been some attempts to combine rule engines and machine learning algorithms to optimize marketing activities, these solutions often have limitations. For example, they may only rely on simple machine learning models to predict user behavior, while ignoring the complexity and diversity of the marketing activities themselves; or, they may lack the ability to dynamically adjust the triggering rules and content of marketing activities, resulting in limited optimization effects. Therefore, it is particularly important to develop a method that can comprehensively consider the characteristics of marketing activities, user behavior patterns, and changes in the market environment to achieve intelligent triggering and optimization. Summary of the invention

[0004] The purpose of the present invention is to provide a marketing intelligent triggering and optimization method, device, equipment and medium based on a rule engine, aiming to improve the accuracy and conversion rate of marketing activities and reduce the cost of manual intervention.

[0005] In order to achieve the above-mentioned purpose, the first aspect of the embodiment of the present disclosure provides a marketing intelligent triggering and optimization method based on a rule engine, comprising: Acquiring marketing user information of a user in a marketing activity with first activity content delivered through first delivery information, wherein the marketing activity with first activity content delivered through the first delivery information is based on a random forest algorithm, and after selecting a trigger rule from a rule library according to the first delivery information to construct a marketing activity trigger expression, the marketing activity is delivered to the user when the marketing activity trigger expression is satisfied; According to the marketing user information of the user in the marketing activity delivered by the first delivery information, based on the reinforcement learning algorithm, with the optimal action value as the goal, the trigger rule in the marketing activity trigger expression is adjusted to obtain a new marketing activity trigger expression; According to the new marketing activity trigger expression, the first activity content of the marketing activity is adjusted to obtain the second activity content of the marketing activity, and according to the new marketing activity trigger expression and the second activity content, second delivery information of the marketing activity after the activity content is adjusted is determined; When the new marketing activity trigger expression is satisfied, the marketing activity after adjusting the activity content is re-delivered according to the second delivery information.

[0006] In a possible implementation, the triggering rule in the marketing activity triggering expression is adjusted based on the marketing user information in the marketing activity delivered by the first delivery information, based on a reinforcement learning algorithm and with the optimal action value as the goal, to obtain a new marketing activity triggering expression, including: Construct a virtual agent corresponding to the user, and execute the following steps in a loop: using the marketing user information of the user in the marketing activity delivered by the first delivery information as the basis for selecting the movement path of the virtual agent, and determining the branch point of each movement from the forest tree corresponding to the marketing activity trigger expression; For the probability of the virtual agent selecting a movement path at any branch point, determine the action value assignment for each movement to the next branch point according to the movement path; Calculating the sum of the action values ​​of all the branch points along the path each time the virtual agent reaches a treetop of the forest tree; When the sum is smaller than the average value of the action value of the virtual agent corresponding to the user moving to all treetops, target trigger rules are selected from the rule library from the treetops in turn to replace the original trigger rules of the branch points, and the movement of the virtual agent is re-executed until the difference between the two adjacent average values ​​is smaller than the preset difference threshold, so as to obtain the marketing activity trigger expression corresponding to the optimal action value.

[0007] In a possible implementation, the target trigger rule is selected from the rule base in turn from the treetop to replace the original trigger rule of the branch point, and the movement of the virtual agent is re-executed until the difference between the two adjacent average values ​​is less than a preset difference threshold, and the marketing activity trigger expression corresponding to the optimal action value is obtained, including: Selecting target trigger rules from the rule base from the treetop in turn to replace the original trigger rules of the branch point, and re-executing the movement of the virtual agent until the difference between the two adjacent average values ​​is less than a preset difference threshold, and then stopping replacing the original trigger rules of the branch point; Determine the maximum action value that the virtual agent corresponding to the user can obtain when reaching a treetop of the forest tree under the condition that the virtual agent corresponding to the user stops changing the original trigger rule of the branch point; The branch point on the moving path corresponding to the maximum action value of the virtual agent corresponding to the user is used as a locked branch point, and when the original trigger rule of the branch point is replaced by executing the mobile agent corresponding to other users, the trigger rule corresponding to the locked branch point is not replaced; The virtual agents corresponding to all users are moved and the original triggering rules of the branch points are replaced until all users have completed the traversal and the marketing activity triggering expression corresponding to the optimal action value is obtained.

[0008] In a possible implementation, the step of selecting a target trigger rule from the rule base in turn from the treetop to replace the original trigger rule of the branch point, and re-executing the movement of the virtual agent until the difference between two adjacent average values ​​is less than a preset difference threshold, and then stopping replacing the original trigger rule of the branch point includes: When the treetop is used as the target branch point, the original trigger rule of the target branch point is replaced, and the movement of the virtual agent to reach the target branch point is re-executed to obtain the action value assignment of moving to the target branch point according to the movement path; Determine whether the difference between the average values ​​corresponding to the previous movement and the current movement is less than the preset difference threshold; When the difference is less than the preset difference threshold, stop replacing the original trigger rule of the target branch point, and take the parent node of the target branch point in the forest tree as the next target branch point, re-execute the movement of the virtual agent to reach the target branch point, and so on, until the last branch point on the movement path has been taken as the target branch point, re-execute the movement of the virtual agent to reach the target branch point, and stop replacing the original trigger rule of the branch point.

[0009] In one possible implementation, the method further includes: When it is greater than or equal to the preset difference threshold, the target trigger rule is selected from the rule library again to replace the original trigger rule of the target branch point, and the movement of the virtual agent is re-executed until the difference between the two adjacent average values ​​is less than the preset difference threshold.

[0010] In a possible implementation, adjusting the first activity content of the marketing activity according to the new marketing activity trigger expression to obtain the second activity content of the marketing activity, and determining the second delivery information of the marketing activity after adjusting the activity content according to the new marketing activity trigger expression and the second activity content, includes: According to the trigger condition and the corresponding trigger rule in the new marketing activity trigger expression, the first activity content of the marketing activity is adjusted to obtain the second activity content of the marketing activity, wherein the adjustment of the first activity content of the marketing activity includes at least one of the following: changing the preferential strength of the activity, adjusting the time range of the activity, adding or deleting activity goods or services, and modifying the promotional copy of the activity; According to the new marketing activity trigger expression and the second activity content, the second delivery information of the adjusted marketing activity is re-determined, and the second delivery information includes at least one of the following: delivery time, delivery channel, delivery frequency and delivery user group.

[0011] In a possible implementation, the marketing user information includes at least one of the following: The frequency of participation in the marketing activity, the duration of participation in the marketing activity, the time period of participation in the marketing activity, and basic information of users participating in the marketing activity.

[0012] According to a second aspect of the embodiments of the present disclosure, a marketing intelligent triggering and optimization device based on a rule engine is provided, the device comprising: an acquisition module, configured to acquire marketing user information of a user in a marketing activity with first activity content delivered through first delivery information, wherein the marketing activity with first activity content delivered through the first delivery information is based on a random forest algorithm, after selecting a trigger rule from a rule library according to the first delivery information to construct a marketing activity trigger expression, and delivering to the user when the marketing activity trigger expression is satisfied; an adjustment module configured to adjust the trigger rule in the marketing activity trigger expression based on the marketing user information in the marketing activity delivered by the first delivery information, based on the reinforcement learning algorithm and with the optimal action value as the goal, to obtain a new marketing activity trigger expression; a determination module configured to adjust the first activity content of the marketing activity according to the new marketing activity trigger expression to obtain the second activity content of the marketing activity, and determine the second delivery information of the marketing activity after the activity content is adjusted according to the new marketing activity trigger expression and the second activity content; The delivery module is configured to re-deliver the marketing activity after adjusting the activity content according to the second delivery information when the new marketing activity trigger expression is met.

[0013] According to a third aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the steps of any one of the methods described in the first aspect are implemented.

[0014] According to a fourth aspect of the embodiments of the present disclosure, an electronic device is provided, including: a memory having a computer program stored thereon; A processor is used to execute the computer program in the memory to implement the steps of any one of the methods in the first aspect.

[0015] The present invention provides a marketing intelligent triggering and optimization method, device, equipment and medium based on a rule engine. Compared with the prior art, it has the following beneficial effects: The random forest algorithm is used to conduct an in-depth analysis of the first delivery information, intelligently select the trigger rules that best match the current marketing activities from the rule library, and construct the initial marketing activity trigger expression. This not only improves the accuracy and pertinence of the trigger rules, but also ensures that the marketing activities can be delivered to the most appropriate user groups.

[0016] According to the marketing user information of users in marketing activities, the trigger rules are dynamically adjusted using reinforcement learning algorithms. Through continuous trial and error and learning, with the optimal action value as the goal, the trigger rules are gradually optimized, so that marketing activities can more accurately trigger potential interested users, thereby improving delivery efficiency and conversion rate.

[0017] On the basis of trigger rule optimization, the content of the marketing activity is further adjusted according to the new trigger expression, the second activity content is generated, and the corresponding second delivery information is determined, which not only enhances the flexibility and personalization of the marketing activity, but also enables the marketing activity to better adapt to changes in market demand and user preferences. Finally, when the new trigger expression is met, the adjusted marketing activity is re-delivered according to the second delivery information, achieving comprehensive optimization from trigger to content.

[0018] In summary, by integrating the rule engine and reinforcement learning algorithm, an efficient and intelligent marketing triggering and optimization method is realized. It not only improves the accuracy and conversion rate of marketing activities, but also reduces the cost of manual intervention, bringing significant technological innovation and economic benefits to the field of digital marketing.

[0019] Other features and advantages of the present disclosure will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings are used to provide a further understanding of the present disclosure and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the present disclosure but do not constitute a limitation of the present disclosure. In the accompanying drawings: Figure 1 It is a flowchart of a marketing intelligent triggering and optimization method based on a rule engine according to an embodiment of the specification.

[0021] Figure 2 It is a block diagram of a marketing intelligent triggering and optimization device based on a rule engine according to an embodiment of the specification.

[0022] Figure 3 It is a block diagram of another marketing intelligent triggering and optimization device based on a rule engine according to an embodiment of the specification. DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] The specific implementation of the present disclosure is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described herein is only used to illustrate and explain the present disclosure, and is not used to limit the present disclosure.

[0025] The present invention provides a marketing intelligent triggering and optimization method based on a rule engine. Figure 1 The present invention is a flow chart showing a marketing intelligent triggering and optimization method based on a rule engine according to an embodiment. The method comprises: In step S11, marketing user information of a user in a marketing activity with first activity content delivered through first delivery information is obtained, and the marketing activity with first activity content delivered through the first delivery information is based on a random forest algorithm, and after a trigger rule is selected from a rule library according to the first delivery information to construct a marketing activity trigger expression, delivery is performed to the user when the marketing activity trigger expression is satisfied; Among them, marketing user information refers to the data generated by users when participating in specific marketing activities, including but not limited to user behavior data, transaction data, preference data, etc. The first delivery information refers to a series of parameters and settings used when launching a marketing activity for the first time, such as delivery time, delivery channel, target user group, etc. The marketing activity trigger expression is a logical expression built based on preset rules and user data, which is used to determine whether to launch a marketing activity to a specific user.

[0026] In the disclosed embodiment, key features are extracted from the first delivery information as input to the random forest algorithm. These features should be able to reflect the characteristics of the target user group and the needs of the marketing campaign. The features are preprocessed, such as missing value filling, outlier processing, feature scaling, etc., to improve the performance and accuracy of the algorithm. The random forest model is trained using the extracted features and the trigger rules in the rule base. During the training process, the model learns how to construct a marketing campaign trigger expression based on user features and the rules in the rule base. Through multiple iterations and cross-validation, the parameters and performance of the model are optimized to ensure that the model can accurately predict which users meet the trigger conditions.

[0027] Build a marketing campaign trigger expression based on the trained random forest model, user characteristics and rules in the rule base. This expression is a logical judgment statement used to determine whether the user meets the trigger conditions of the marketing campaign. The trigger expression may contain multiple conditions and sub-conditions, which are combined through logical operations (such as AND, OR, NOT) to form a complete trigger logic.

[0028] Collect user data in real time and make judgments based on the marketing campaign trigger expression. When the user's real-time data meets the trigger expression, it is considered that the user meets the trigger condition of the marketing campaign. When the trigger condition is met, the marketing campaign with the first activity content is delivered to the user according to the preset marketing strategy and delivery information (such as delivery time, delivery channel, etc.). During the delivery process, continuously monitor user feedback and marketing campaign results to adjust the marketing strategy and optimize the trigger expression in a timely manner.

[0029] In the disclosed embodiment, the random forest algorithm is used to analyze this information, and the trigger rules that best match the current marketing activity are screened out from the preset rule library. These rules may involve multiple dimensions such as user portraits, behavior patterns, and geographic locations. Based on the screened trigger rules, a marketing activity trigger expression is constructed. When the user's real-time data meets the expression, a marketing activity with the first activity content is delivered to the user. Then, the marketing user information under the first delivery information is collected. This information includes but is not limited to the user's click behavior, browsing history, purchase history, etc.

[0030] For example, suppose a bank plans to push a new loan campaign to potential users. Using the random forest algorithm to analyze historical user data, it is found that users who have browsed related loans are more likely to be interested in this loan campaign. Therefore, the system selects these users as the target group and constructs a marketing campaign trigger expression, such as "users have browsed bank loan pages in the past 30 days and have not taken out loans." When users who meet this condition come online, the system pushes the new loan campaign to them. Collect the first delivery information, such as the delivery time is the weekend, and the delivery channel is social media and email.

[0031] In step S12, according to the marketing user information of the user in the marketing activity delivered by the first delivery information, based on the reinforcement learning algorithm, the trigger rule in the marketing activity trigger expression is adjusted with the optimal action value as the goal, so as to obtain a new marketing activity trigger expression; Among them, reinforcement learning algorithms can learn the optimal strategy through the interaction between the agent and the environment to maximize the cumulative reward. Action value refers to the expected return that can be obtained after performing an action in reinforcement learning.

[0032] In the disclosed embodiment, based on the user's feedback under the first delivery information (i.e., marketing user information), the trigger rules in the marketing activity trigger expression are adjusted using a reinforcement learning algorithm. The reinforcement learning algorithm optimizes the trigger rules gradually through continuous trial and error and learning, with the goal of maximizing the cumulative reward, that is, improving the conversion rate or user engagement of the marketing activity. By simulating the user response under different trigger rules, and adjusting the rule weights or introducing new rules according to actual feedback, a new marketing activity trigger expression is obtained.

[0033] Continuing with the above example, suppose the conversion rate of the pushed promotional activities is lower than expected. The actual feedback data of users is analyzed using reinforcement learning algorithms, and it is found that although users are interested in loans, they are not sensitive enough to the interest rate of loan activities. Therefore, the system adjusts the trigger rules and changes "users have browsed bank loan pages in the past 30 days and have not taken out loans" to "users have browsed bank loan pages in the past 30 days and have not taken out loans and are highly sensitive to interest rates." At the same time, other trigger conditions are also considered, such as user repayment information, overdue repayment, etc., to further optimize the trigger rules.

[0034] In step S13, the first activity content of the marketing activity is adjusted according to the new marketing activity trigger expression to obtain the second activity content of the marketing activity, and second delivery information of the marketing activity after the activity content is adjusted is determined according to the new marketing activity trigger expression and the second activity content; Among them, the activity content adjustment is to modify and optimize the specific content of the marketing activity (such as bank type, loan type, etc.) according to the user portrait and needs after the trigger rules are adjusted. The second delivery information is a series of parameters and settings used when re-delivering the marketing activity after adjusting the activity content.

[0035] In the disclosed embodiment, the first activity content is adjusted according to the new marketing activity trigger expression. This adjustment may involve multiple aspects such as adjustment of the discount strength, optimization of product recommendations, and modification of activity copy. At the same time, the system also determines the second delivery information based on the new trigger expression and the second activity content. This information may include the adjusted delivery time, delivery channel, target user group, etc. By comprehensively considering the changes in trigger rules and user needs, the system ensures that the adjusted marketing activities can reach the target users more accurately and improve the conversion rate.

[0036] Continuing with the previous example, suppose the trigger rules are adjusted and it is found that users are highly sensitive to loan interest rates. The first activity content is adjusted, the original bank interest rate is changed from "annual interest rate 3.75%" to "annual interest rate 3.55%", and some loan repayment activities that users may be interested in are added. At the same time, the system also determines the second delivery information based on the new trigger expression, such as adjusting the delivery time to the prime time in the evening, increasing the delivery ratio on social media platforms, etc.

[0037] In step S14, when the new marketing activity trigger expression is satisfied, the marketing activity with adjusted activity content is re-delivered according to the second delivery information.

[0038] In the disclosed embodiment, the marketing campaign after adjusting the activity content is re-launched when the new marketing campaign trigger expression is met. During the re-launch process, user feedback data, i.e., marketing user information, is continuously collected and analyzed, and the effect of the marketing campaign is evaluated. By comparing the changes in indicators such as conversion rate and user engagement before and after the re-launch, the system can determine whether the adjusted marketing campaign has achieved the expected effect. If the effect is still not ideal, it is possible to return to step S12 for further trigger rule adjustment and content optimization.

[0039] Continuing with the previous example, the adjusted marketing campaign is re-launched based on the new trigger expression and the second delivery information. During the delivery process, user feedback data is continuously collected and analyzed, and indicators such as conversion rate and user engagement are monitored and evaluated. Assuming that the conversion rate after re-delivery is significantly improved and user engagement is also increased, this indicates that the adjusted marketing campaign has achieved the expected results. At this point, the system can consider the process of triggering and optimizing this marketing campaign to be over, and use the relevant experience and data for future marketing campaign planning and optimization.

[0040] The above technical solution uses the random forest algorithm to conduct an in-depth analysis of the first delivery information, intelligently selects the trigger rules that best match the current marketing activities from the rule library, and constructs the initial marketing activity trigger expression. This not only improves the accuracy and pertinence of the trigger rules, but also ensures that the marketing activities can be delivered to the most appropriate user groups.

[0041] According to the marketing user information of users in marketing activities, the trigger rules are dynamically adjusted using reinforcement learning algorithms. Through continuous trial and error and learning, with the optimal action value as the goal, the trigger rules are gradually optimized, so that marketing activities can more accurately trigger potential interested users, thereby improving delivery efficiency and conversion rate.

[0042] On the basis of trigger rule optimization, the content of the marketing activity is further adjusted according to the new trigger expression, the second activity content is generated, and the corresponding second delivery information is determined, which not only enhances the flexibility and personalization of the marketing activity, but also enables the marketing activity to better adapt to changes in market demand and user preferences. Finally, when the new trigger expression is met, the adjusted marketing activity is re-delivered according to the second delivery information, achieving comprehensive optimization from trigger to content.

[0043] In summary, by integrating the rule engine and reinforcement learning algorithm, an efficient and intelligent marketing triggering and optimization method is realized. It not only improves the accuracy and conversion rate of marketing activities, but also reduces the cost of manual intervention, bringing significant technological innovation and economic benefits to the field of digital marketing.

[0044] In a possible implementation, in step S12, the triggering rule in the marketing activity triggering expression is adjusted based on the marketing user information in the marketing activity delivered by the first delivery information, based on the reinforcement learning algorithm and with the optimal action value as the goal, to obtain a new marketing activity triggering expression, including: In step S121, a virtual agent corresponding to the user is constructed and executed cyclically: the marketing user information of the user in the marketing activity delivered by the first delivery information is used as the basis for selecting the movement path of the virtual agent, and the branch point of each movement is determined from the forest tree corresponding to the marketing activity trigger expression; Among them, the virtual agent is an entity that performs actions and receives feedback from the environment in reinforcement learning. In this case, the virtual agent represents the user and is used to select paths in the decision tree (or forest tree, for simplicity of understanding, it is collectively referred to as a decision tree below) of the marketing campaign trigger expression.

[0045] In the disclosed embodiment, a virtual agent is constructed for each user, and the agent selects a path in the decision tree according to the marketing user information (such as browsing history, browsing time, etc.) of the user under the first delivery information. The movement path of the agent reflects the process of the user meeting the trigger condition.

[0046] For example: Assume that there are multiple branch points in the decision tree, each branch point represents a trigger rule (such as "the user has browsed bank loan pages in the past 30 days and has not applied for a loan and is highly sensitive to interest rates"). The virtual agent selects the corresponding branch to move based on the user's actual behavior data (such as the user has indeed not applied for a loan in the past 30 days).

[0047] In step S122, for the probability of the virtual agent selecting a movement path at any branch point, determine the action value assignment for each movement to the next branch point according to the movement path; Action Value is the expected value of the cumulative reward obtained after performing an action in reinforcement learning. In this case, the action value is used to evaluate the effect of the trigger rule in the marketing campaign.

[0048] In the disclosed embodiment, for each branch point in the decision tree, the probability of the virtual agent selecting a path at that point is calculated, and the action value is assigned according to the selected path. This assignment reflects the effect of the triggering rule in the current user group.

[0049] For example: If at a branch point, the virtual agent has an 80% probability of choosing "yes" (i.e. the user meets the trigger rule), then the action value of this path may be higher. On the contrary, if the agent has a 20% probability of choosing "no", then the action value of this path may be lower.

[0050] In step S123, each time the virtual agent reaches a treetop of the forest tree, the sum of the action values ​​of all the branch points along the path is calculated; The leaf node is a leaf node in the decision tree, which represents the final decision result or output value. In this example, the leaf node represents the final judgment result of the marketing activity trigger expression.

[0051] In the disclosed embodiment, when the virtual agent reaches the top of the decision tree, the sum of the action values ​​of all branch points on its path is calculated. This sum reflects the effect of the entire trigger expression in the current user group.

[0052] For example: suppose the virtual agent starts from the root node, passes through multiple branch points in sequence, and finally reaches a treetop. In this process, the action value of each branch point is accumulated to form the final sum value.

[0053] In step S124, when the sum is less than the average value of the action value of the virtual agent corresponding to the user moving to all treetops, target trigger rules are selected from the rule library from the treetops in turn to replace the original trigger rules of the branch points, and the movement of the virtual agent is re-executed until the difference between the two adjacent average values ​​is less than the preset difference threshold, so as to obtain the marketing activity trigger expression corresponding to the optimal action value.

[0054] In the disclosed embodiment, if the calculated action value and value are less than the average value of the action value of the virtual agent moving to all treetops, it means that the current trigger expression is not effective. At this time, starting from the treetop, select the target trigger rule from the rule library to replace the original trigger rule, and re-execute the movement of the virtual agent. This process will be iterated continuously until the difference between the two adjacent average values ​​is less than the preset difference threshold, and the trigger expression obtained at this time is the optimal one.

[0055] For example: Assume that in the first iteration, the calculated action value and value are low. At this point, you can start from the top of the tree and check whether the trigger rules of each branch point are appropriate. If you find that a trigger rule is not effective (such as the rule that "the user has browsed bank loan pages in the past 30 days and has not taken out a loan and is highly sensitive to interest rates" is not common among the user group), you can select a more suitable rule from the rule library to replace it (such as "the user has browsed bank loan pages in the past 30 days and has not taken out a loan and has a low repayment period requirement"). Then re-execute the movement process of the virtual agent and calculate the new action value and value. This process will be repeated until the optimal trigger expression is found.

[0056] In this way, the reinforcement learning algorithm can be used to adjust and optimize the trigger rules in the marketing campaign trigger expression, thereby improving the conversion rate and user engagement of the marketing campaign.

[0057] In a possible implementation, in step S124, the target trigger rule is selected from the rule base in turn from the treetop to replace the original trigger rule of the branch point, and the movement of the virtual agent is re-executed until the difference between the two adjacent average values ​​is less than a preset difference threshold, and the marketing activity trigger expression corresponding to the optimal action value is obtained, including: In step S1241, target trigger rules are selected from the rule base from the treetop to replace the original trigger rule of the branch point, and the movement of the virtual agent is re-executed until the difference between the two adjacent average values ​​is less than a preset difference threshold, and the replacement of the original trigger rule of the branch point is stopped; In the disclosed embodiment, starting from the top of the decision tree, the trigger rules of each branch point are checked in turn, and the target trigger rule is selected from the rule base for replacement. Then the movement of the virtual agent is re-executed, and the new action value and average value are calculated. If the difference between two adjacent average values ​​is less than the preset difference threshold, the trigger rule replacement of the branch point is stopped.

[0058] For example: Assume that at a certain branch of the decision tree, the action value and average value of the corresponding user group under the original trigger rule are low. At this time, a new trigger rule can be selected from the rule library to replace the original rule, and the action value and average value can be recalculated. If the difference between the new average value and the old average value is less than the difference threshold (for example, 0.01), it is considered that the trigger rule of the branch point has been sufficiently optimized and no replacement is required.

[0059] In step S1242, determining the maximum action value that the virtual agent corresponding to the user can obtain when reaching a treetop of the forest tree under the condition that the virtual agent corresponding to the user stops changing the original trigger rule of the branch point; In the disclosed embodiment, after the trigger rule is stopped from being changed, the maximum action value that the virtual agent can obtain when reaching the top of the decision tree is determined. This maximum action value represents the optimal effect of the current trigger expression in the user group.

[0060] For example: suppose that after stopping changing the trigger rule, the virtual agent obtains a higher action value (for example, 0.8) when it reaches a certain treetop. This value may be gradually optimized during multiple iterations and represents the optimal effect of the current trigger expression in the user group.

[0061] In step S1243, the branch point on the moving path corresponding to the maximum action value of the virtual agent corresponding to the user is used as a locked branch point, and when the original trigger rule of the branch point is replaced by executing the replacement of the trigger rule of the branch point through other mobile agents corresponding to the users, the trigger rule corresponding to the locked branch point is not replaced; In the disclosed embodiment, the branch points of the virtual agent on the moving path corresponding to the maximum action value are used as locked branch points. In the subsequent optimization process, these locked branch points will no longer change the trigger rules.

[0062] For example: Assume that the virtual agent obtains the maximum action value when it reaches a certain treetop, and the branch points A, B, and C on the path all make important contributions to the final action value. At this time, A, B, and C can be used as locked branch points, and their triggering rules will not be changed in the subsequent optimization process.

[0063] In step S1244, the virtual agents corresponding to all users are moved and the original trigger rules of the branch points are replaced until all users have completed the traversal and the marketing activity trigger expression corresponding to the optimal action value is obtained.

[0064] In the disclosed embodiment, the virtual agents of all users are traversed to perform the operations of moving and changing the trigger rules. During the traversal process, the rules of locking branch points are followed, and the trigger rules of locked branch points are not changed. This continues until all users have completed the traversal and the marketing activity trigger expression corresponding to the optimal action value is obtained.

[0065] For example: suppose there are 1,000 users, each with a corresponding virtual agent. During the traversal process, the virtual agent of each user is moved and the trigger rule is replaced. At the same time, the rule of locking branch points is followed, and the trigger rule of locked branch points is not replaced. After multiple iterations and traversals, a marketing campaign trigger expression that performs well in all user groups is finally obtained.

[0066] Through the above steps, the marketing campaign trigger expression can be optimized based on the reinforcement learning algorithm to obtain the trigger rule combination that performs best in the user group. This method not only improves the conversion rate and user engagement of marketing activities, but also enhances the targeting and personalization of marketing activities.

[0067] In a possible implementation, in step S1241, the treetop sequentially selects target trigger rules from the rule base to replace the original trigger rule of the branch point, and re-executes the movement of the virtual agent until the difference between two adjacent average values ​​is less than a preset difference threshold, and stops replacing the original trigger rule of the branch point, including: In step S12411, the original trigger rule of the target branch point is replaced by taking the treetop as the target branch point, and the movement of the virtual agent to reach the target branch point is re-executed to obtain the action value assignment of moving to the target branch point according to the movement path; In the disclosed embodiment, starting from the top of the decision tree, the currently considered branch point is taken as the target branch point. A new trigger rule is selected from the rule library to replace the original trigger rule of the target branch point. Then, the movement of the virtual agent is re-executed, and the action value assignment of moving to the target branch point according to the new movement path is calculated.

[0068] For example: suppose that the user group corresponding to a certain treetop of the decision tree has a low action value under the original trigger rule. At this time, select the branch point corresponding to the treetop as the target branch point, and select a new trigger rule from the rule library to replace it. Then, re-execute the movement of the virtual agent and calculate the new action value assignment.

[0069] In step S12412, determine whether the difference between the average values ​​corresponding to the previous movement and the current movement is less than the preset difference threshold; In the disclosed embodiment, after the trigger rule is changed and the movement of the virtual agent is re-executed, the difference between the average values ​​corresponding to the last movement and the current movement is calculated, and it is determined whether the difference is less than a preset difference threshold.

[0070] For example: suppose that after changing the trigger rule, the difference between the average value corresponding to the new action value assignment and the old average value is 0.02, and the preset difference threshold is 0.01. At this time, it is judged that the difference is greater than the preset difference threshold, and it is necessary to continue iterative optimization.

[0071] In step S12413, when the difference is less than the preset difference threshold, stop replacing the original trigger rule of the target branch point, and use the parent node of the target branch point in the forest tree as the next target branch point, re-execute the movement of the virtual agent to reach the target branch point, and so on, until the last branch point on the movement path has been used as the target branch point, re-execute the movement of the virtual agent to reach the target branch point, and stop replacing the original trigger rule of the branch point.

[0072] In the disclosed embodiment, if the average difference is less than the preset difference threshold, the trigger rule for the current target branch point is stopped from being replaced. The parent node of the current target branch point is used as the next target branch point, and steps S12411 and S12412 are repeatedly performed until the last branch point (i.e., the root node) on the moving path is used as the target branch point, and the trigger rule replacement is stopped.

[0073] For example: If the average difference of a branch point is less than the preset difference threshold during the iteration, the trigger rule of the branch point will be stopped from being replaced. Then, the parent node of the branch point will be used as the new target branch point, and the iteration process will continue until the root node is also considered and the trigger rule replacement is stopped.

[0074] In one possible implementation, the method further includes: In step S12414, when the value is greater than or equal to the preset difference threshold, the target trigger rule is selected from the rule base again to replace the original trigger rule of the target branch point, and the movement of the virtual agent is re-executed until the difference between the two adjacent average values ​​is less than the preset difference threshold.

[0075] In the disclosed embodiment, if the average difference is greater than or equal to the preset difference threshold, the target trigger rule is selected from the rule library again to replace the original trigger rule of the target branch point, and the movement of the virtual agent is re-executed. This process is repeated until the difference between two adjacent average values ​​is less than the preset difference threshold.

[0076] For example: Assume that after the first iteration, the average difference is still greater than the preset difference threshold. At this time, a new trigger rule is selected from the rule library to replace the original trigger rule of the target branch point, and the movement of the virtual agent is re-executed. Then, the average difference is calculated again, and it is determined whether to continue iterating. This process will be repeated until the stop condition is met.

[0077] Through the above steps, the trigger rules in the trigger expression of the marketing activity can be iteratively optimized based on the reinforcement learning algorithm. This method not only improves the conversion rate and user engagement of the marketing activity, but also enhances the pertinence and personalization of the marketing activity by continuously iterating and approaching the optimal solution.

[0078] In a possible implementation, in step S13, adjusting the first activity content of the marketing activity according to the new marketing activity trigger expression to obtain the second activity content of the marketing activity, and determining the second delivery information of the marketing activity after adjusting the activity content according to the new marketing activity trigger expression and the second activity content, includes: In step S131, the first activity content of the marketing activity is adjusted according to the trigger condition and the corresponding trigger rule in the new marketing activity trigger expression to obtain the second activity content of the marketing activity, wherein the adjustment of the first activity content of the marketing activity includes at least one of the following: changing the preferential strength of the activity, adjusting the time range of the activity, adding or deleting activity products or services, and modifying the promotional copy of the activity; In the disclosed embodiment, the new marketing activity trigger expression is first parsed to identify the trigger conditions and corresponding trigger rules. Then, the original first activity content is adjusted according to these conditions and rules. The adjustment may involve the following aspects: Change the discount intensity: Adjust the discount range of the activity based on trigger conditions, such as user historical purchasing behavior, user level or specific holidays, to better attract target users.

[0079] Adjust activity time range: Based on trigger rules, such as high user activity during a specific time period, adjust the start and end time of the activity to maximize user engagement.

[0080] Add or remove products or services for an event: Adjust the list of products or services involved in the event based on user preferences, market demand or inventory conditions.

[0081] Modify promotional copy: Based on the triggering conditions and characteristics of the target user group, adjust the style, content and appeal of the promotional copy to increase the exposure and participation of the event.

[0082] For example, suppose a new loan activity trigger expression defines that when a user has taken out a certain number of loans and has repaid all of them, an additional interest rate reduction will be provided. After the system recognizes this trigger condition, it automatically reduces the interest rate in the first activity content from 3.55% to 3.25% to attract these users with a high loan frequency.

[0083] In step S132, the second delivery information of the adjusted marketing activity is re-determined according to the new marketing activity trigger expression and the second activity content, and the second delivery information includes at least one of the following: delivery time, delivery channel, delivery frequency and delivery user group.

[0084] In the disclosed embodiment, the second delivery information of the adjusted marketing activity is re-determined according to the new marketing activity trigger expression and the second activity content. This step is intended to ensure that the activity can reach the user group most likely to participate at the most appropriate time, through the most appropriate channel, and at the most appropriate frequency.

[0085] Delivery time: Select the best delivery time or time period based on user activity and the time distribution of trigger conditions.

[0086] Distribution channels: Choose appropriate distribution platforms or media based on the preferences and habits of the target user group.

[0087] Delivery frequency: Adjust the delivery frequency based on user feedback, campaign performance, and market changes to ensure continued exposure and engagement of the campaign.

[0088] Target user group: A set of target users selected based on user attributes, behavioral data or trigger conditions to ensure accurate delivery of the campaign.

[0089] For example, suppose a new marketing campaign trigger expression defines that when a user has visited a specific product page in the past month, a campaign notification is sent via email. After the system recognizes this trigger condition, it automatically selects the user group that meets this condition and sends the campaign notification via email at the appropriate time. At the same time, based on user feedback and campaign results, the system can dynamically adjust the delivery frequency and time to optimize the campaign results.

[0090] In a possible implementation, the marketing user information includes at least one of the following: The frequency of participation in the marketing activity, the duration of participation in the marketing activity, the time period of participation in the marketing activity, and basic information of users participating in the marketing activity.

[0091] In the disclosed embodiment, the activity participation frequency is the number of times a user participates in a specific marketing activity during the activity. This indicator reflects the user's interest and activity in the activity, and is an important basis for evaluating the attractiveness of the activity and user stickiness.

[0092] Activity participation duration refers to the time users spend in a single marketing activity. This data helps us understand the user's level of involvement in the activity and the attractiveness of the content, and is instructive for optimizing activity design and improving user experience.

[0093] Participation time period refers to the time distribution of users participating in marketing activities, including specific dates, times or periods. By analyzing participation time periods, we can identify the peak and trough periods of user activity, providing data support for accurate advertising and arranging activity rhythms.

[0094] User basic information refers to basic personal information provided by users during registration, participation in activities or interactions, such as age, gender, region, occupation, interests and hobbies, etc. This information helps to build user portraits, achieve personalized marketing, and improve the relevance and conversion rate of activities. User basic information is an indispensable basic data for formulating marketing strategies and evaluating the effectiveness of activities.

[0095] The present disclosure also provides a marketing intelligent triggering and optimization device based on a rule engine, see Figure 2 As shown, the device comprises: The acquisition module 210 is configured to acquire marketing user information of a user in a marketing activity with a first activity content delivered through the first delivery information, wherein the marketing activity with the first activity content delivered through the first delivery information is based on a random forest algorithm, and after selecting a trigger rule from a rule library according to the first delivery information to construct a marketing activity trigger expression, the marketing activity is delivered to the user when the marketing activity trigger expression is satisfied; The adjustment module 220 is configured to adjust the trigger rule in the marketing activity trigger expression based on the marketing user information in the marketing activity delivered by the first delivery information, based on the reinforcement learning algorithm and with the optimal action value as the goal, to obtain a new marketing activity trigger expression; The determination module 230 is configured to adjust the first activity content of the marketing activity according to the new marketing activity trigger expression to obtain the second activity content of the marketing activity, and determine the second delivery information of the marketing activity after the activity content is adjusted according to the new marketing activity trigger expression and the second activity content; The delivery module 240 is configured to re-deliver the marketing activity after adjusting the activity content according to the second delivery information when the new marketing activity trigger expression is met.

[0096] In a possible implementation, the adjustment module 220 is configured to: Construct a virtual agent corresponding to the user, and execute the following steps in a loop: using the marketing user information of the user in the marketing activity delivered by the first delivery information as the basis for selecting the movement path of the virtual agent, and determining the branch point of each movement from the forest tree corresponding to the marketing activity trigger expression; For the probability of the virtual agent selecting a movement path at any branch point, determine the action value assignment for each movement to the next branch point according to the movement path; Calculating the sum of the action values ​​of all the branch points along the path each time the virtual agent reaches a treetop of the forest tree; When the sum is smaller than the average value of the action value of the virtual agent corresponding to the user moving to all treetops, target trigger rules are selected from the rule library from the treetops in turn to replace the original trigger rules of the branch points, and the movement of the virtual agent is re-executed until the difference between the two adjacent average values ​​is smaller than the preset difference threshold, so as to obtain the marketing activity trigger expression corresponding to the optimal action value.

[0097] In a possible implementation, the adjustment module 220 is configured to: Selecting target trigger rules from the rule base from the treetop in turn to replace the original trigger rules of the branch point, and re-executing the movement of the virtual agent until the difference between the two adjacent average values ​​is less than a preset difference threshold, and then stopping replacing the original trigger rules of the branch point; Determine the maximum action value that the virtual agent corresponding to the user can obtain when reaching a treetop of the forest tree under the condition that the virtual agent corresponding to the user stops changing the original trigger rule of the branch point; The branch point on the moving path corresponding to the maximum action value of the virtual agent corresponding to the user is used as a locked branch point, and when the original trigger rule of the branch point is replaced by executing the mobile agent corresponding to other users, the trigger rule corresponding to the locked branch point is not replaced; The virtual agents corresponding to all users are moved and the original triggering rules of the branch points are replaced until all users have completed the traversal and the marketing activity triggering expression corresponding to the optimal action value is obtained.

[0098] In a possible implementation, the adjustment module 220 is configured to: When the treetop is used as the target branch point, the original trigger rule of the target branch point is replaced, and the movement of the virtual agent to reach the target branch point is re-executed to obtain the action value assignment of moving to the target branch point according to the movement path; Determine whether the difference between the average values ​​corresponding to the last movement and the current movement is less than the preset difference threshold; When the difference is less than the preset difference threshold, stop replacing the original trigger rule of the target branch point, and take the parent node of the target branch point in the forest tree as the next target branch point, re-execute the movement of the virtual agent to reach the target branch point, and so on, until the last branch point on the movement path has been taken as the target branch point, re-execute the movement of the virtual agent to reach the target branch point, and stop replacing the original trigger rule of the branch point.

[0099] In a possible implementation, the adjustment module 220 is configured to: When it is greater than or equal to the preset difference threshold, the target trigger rule is selected from the rule library again to replace the original trigger rule of the target branch point, and the movement of the virtual agent is re-executed until the difference between the two adjacent average values ​​is less than the preset difference threshold.

[0100] In a possible implementation, the determining module 230 is configured to: According to the trigger condition and the corresponding trigger rule in the new marketing activity trigger expression, the first activity content of the marketing activity is adjusted to obtain the second activity content of the marketing activity, wherein the adjustment of the first activity content of the marketing activity includes at least one of the following: changing the preferential strength of the activity, adjusting the time range of the activity, adding or deleting activity goods or services, and modifying the promotional copy of the activity; According to the new marketing activity trigger expression and the second activity content, the second delivery information of the adjusted marketing activity is re-determined, and the second delivery information includes at least one of the following: delivery time, delivery channel, delivery frequency and delivery user group.

[0101] In a possible implementation, the marketing user information includes at least one of the following: The frequency of participation in the marketing activity, the duration of participation in the marketing activity, the time period of participation in the marketing activity, and basic information of users participating in the marketing activity.

[0102] The present disclosure also provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of the method described in any one of the aforementioned embodiments are implemented.

[0103] The present disclosure also provides an electronic device, including: a memory having a computer program stored thereon; A processor is used to execute the computer program in the memory to implement the steps of the method in any one of the aforementioned embodiments.

[0104] Figure 3 The marketing intelligent triggering and optimization device 100 based on the rule engine shown includes: a processor 1001 and a memory 1003. The processor 1001 and the memory 1003 are connected, such as through a bus 1002. Optionally, the marketing intelligent triggering and optimization device 100 based on the rule engine may also include a communication component 1004, which may be used for data interaction between the device 100 and other devices, such as data sending and / or data receiving. It should be noted that the communication component 1004 is not limited to one in actual scheduling, and the structure of the marketing intelligent triggering and optimization device 100 based on the rule engine does not constitute a limitation on the embodiments of the present application.

[0105] Processor 1001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. Processor 1001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0106] The bus 1002 may include a path to transmit information between the above components. The bus 1002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus 1002 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0107] The memory 1003 may be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), magnetic disk storage medium, other magnetic storage devices, or any other medium that can be used to carry or store program code and can be read by a computer, without limitation herein.

[0108] The memory 1003 is used to store program codes for executing the embodiments of the present disclosure, and the execution is controlled by the processor 1001. The processor 1001 is used to execute the program codes stored in the memory 1003 to implement the steps shown in the above-mentioned marketing intelligent triggering and optimization method embodiment based on the rule engine.

[0109] The disclosed embodiment also provides a computer-readable storage medium having program code stored thereon. When the program code is executed by a processor, the steps and corresponding contents of the aforementioned marketing intelligent triggering and optimization method embodiment based on a rule engine can be implemented.

[0110] The preferred embodiments of the present disclosure are described in detail above in conjunction with the accompanying drawings; however, the present disclosure is not limited to the specific details in the above embodiments; within the technical concept of the present disclosure, various changes, modifications, substitutions and variations may be made to these embodiments, and these changes, modifications, substitutions and variations all fall within the protection scope of the present disclosure.

[0111] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any appropriate manner without contradiction, and they should also be regarded as the contents disclosed in this disclosure. In order to avoid unnecessary repetition, this disclosure will not further describe various possible combinations. The technical scope of this application is not limited to the contents in the specification, and its technical scope must be determined according to the scope of the claims.

Claims

1. A marketing intelligent triggering and optimization method based on rule engine, characterized in that: include: Acquiring marketing user information of a user in a marketing activity with first activity content delivered through first delivery information, wherein the marketing activity with first activity content delivered through the first delivery information is based on a random forest algorithm, and after selecting a trigger rule from a rule library according to the first delivery information to construct a marketing activity trigger expression, the marketing activity is delivered to the user when the marketing activity trigger expression is satisfied; According to the marketing user information of the user in the marketing activity delivered by the first delivery information, based on the reinforcement learning algorithm, with the optimal action value as the goal, the trigger rule in the marketing activity trigger expression is adjusted to obtain a new marketing activity trigger expression; According to the new marketing activity trigger expression, the first activity content of the marketing activity is adjusted to obtain the second activity content of the marketing activity, and according to the new marketing activity trigger expression and the second activity content, second delivery information of the marketing activity after the activity content is adjusted is determined; When the new marketing activity trigger expression is satisfied, the marketing activity after adjusting the activity content is re-delivered according to the second delivery information.

2. The marketing intelligent triggering and optimization method based on rule engine according to claim 1, characterized in that: The method of adjusting the triggering rules in the marketing activity triggering expression based on the marketing user information in the marketing activity delivered by the first delivery information and taking the optimal action value as the goal based on the reinforcement learning algorithm to obtain a new marketing activity triggering expression includes: Construct a virtual agent corresponding to the user, and execute the following steps in a loop: using the marketing user information of the user in the marketing activity delivered by the first delivery information as the basis for selecting the movement path of the virtual agent, and determining the branch point of each movement from the forest tree corresponding to the marketing activity trigger expression; For the probability of the virtual agent selecting a movement path at any branch point, determine the action value assignment for each movement to the next branch point according to the movement path; Calculating the sum of the action values ​​of all the branch points along the path each time the virtual agent reaches a treetop of the forest tree; When the sum is smaller than the average value of the action value of the virtual agent corresponding to the user moving to all treetops, target trigger rules are selected from the rule library from the treetops in turn to replace the original trigger rules of the branch points, and the movement of the virtual agent is re-executed until the difference between the two adjacent average values ​​is smaller than the preset difference threshold, so as to obtain the marketing activity trigger expression corresponding to the optimal action value.

3. The marketing intelligent triggering and optimization method based on rule engine according to claim 2, characterized in that: The target trigger rule is selected from the rule base in turn from the treetop to replace the original trigger rule of the branch point, and the movement of the virtual agent is re-executed until the difference between the two adjacent average values ​​is less than the preset difference threshold, and the marketing activity trigger expression corresponding to the optimal action value is obtained, including: Selecting target trigger rules from the rule base from the treetop in turn to replace the original trigger rules of the branch point, and re-executing the movement of the virtual agent until the difference between the two adjacent average values ​​is less than a preset difference threshold, and then stopping replacing the original trigger rules of the branch point; Determine the maximum action value that the virtual agent corresponding to the user can obtain when reaching a treetop of the forest tree under the condition that the virtual agent corresponding to the user stops changing the original trigger rule of the branch point; The branch point on the moving path corresponding to the maximum action value of the virtual agent corresponding to the user is used as a locked branch point, and when the original trigger rule of the branch point is replaced by executing the mobile agent corresponding to other users, the trigger rule corresponding to the locked branch point is not replaced; The virtual agents corresponding to all users are moved and the original triggering rules of the branch points are replaced until all users have completed the traversal and the marketing activity triggering expression corresponding to the optimal action value is obtained.

4. The marketing intelligent triggering and optimization method based on rule engine according to claim 2 or 3, characterized in that: The step of selecting a target trigger rule from the rule base in turn from the treetop to replace the original trigger rule of the branch point, and re-executing the movement of the virtual agent until the difference between two adjacent average values ​​is less than a preset difference threshold, and stopping replacing the original trigger rule of the branch point, includes: When the treetop is used as the target branch point, the original trigger rule of the target branch point is replaced, and the movement of the virtual agent to reach the target branch point is re-executed to obtain the action value assignment of moving to the target branch point according to the movement path; Determine whether the difference between the average values ​​corresponding to the previous movement and the current movement is less than the preset difference threshold; When the difference is less than the preset difference threshold, stop replacing the original trigger rule of the target branch point, and take the parent node of the target branch point in the forest tree as the next target branch point, re-execute the movement of the virtual agent to reach the target branch point, and so on, until the last branch point on the movement path has been taken as the target branch point, re-execute the movement of the virtual agent to reach the target branch point, and stop replacing the original trigger rule of the branch point.

5. The marketing intelligent triggering and optimization method based on rule engine according to claim 4, characterized in that: The method further comprises: When it is greater than or equal to the preset difference threshold, the target trigger rule is selected from the rule library again to replace the original trigger rule of the target branch point, and the movement of the virtual agent is re-executed until the difference between the two adjacent average values ​​is less than the preset difference threshold.

6. The marketing intelligent triggering and optimization method based on rule engine according to claim 1, characterized in that: The step of adjusting the first activity content of the marketing activity according to the new marketing activity trigger expression to obtain the second activity content of the marketing activity, and determining the second delivery information of the marketing activity after adjusting the activity content according to the new marketing activity trigger expression and the second activity content, includes: According to the trigger condition and the corresponding trigger rule in the new marketing activity trigger expression, the first activity content of the marketing activity is adjusted to obtain the second activity content of the marketing activity, wherein the adjustment of the first activity content of the marketing activity includes at least one of the following: changing the preferential strength of the activity, adjusting the time range of the activity, adding or deleting activity goods or services, and modifying the promotional copy of the activity; According to the new marketing activity trigger expression and the second activity content, the second delivery information of the adjusted marketing activity is re-determined, and the second delivery information includes at least one of the following: delivery time, delivery channel, delivery frequency and delivery user group.

7. The marketing intelligent triggering and optimization method based on rule engine according to claim 1, characterized in that: The marketing user information includes at least one of the following: The frequency of participation in the marketing activity, the duration of participation in the marketing activity, the time period of participation in the marketing activity, and basic information of users participating in the marketing activity.

8. A marketing intelligent triggering and optimization device based on a rule engine, characterized in that: The device comprises: an acquisition module, configured to acquire marketing user information of a user in a marketing activity with first activity content delivered through first delivery information, wherein the marketing activity with first activity content delivered through the first delivery information is based on a random forest algorithm, after selecting a trigger rule from a rule library according to the first delivery information to construct a marketing activity trigger expression, and delivering to the user when the marketing activity trigger expression is satisfied; an adjustment module configured to adjust the trigger rule in the marketing activity trigger expression based on the marketing user information in the marketing activity delivered by the first delivery information, based on the reinforcement learning algorithm and with the optimal action value as the goal, to obtain a new marketing activity trigger expression; a determination module configured to adjust the first activity content of the marketing activity according to the new marketing activity trigger expression to obtain the second activity content of the marketing activity, and determine the second delivery information of the marketing activity after the activity content is adjusted according to the new marketing activity trigger expression and the second activity content; The delivery module is configured to re-deliver the marketing activity after adjusting the activity content according to the second delivery information when the new marketing activity trigger expression is met.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method described in any one of claims 1 to 7 are implemented.

10. An electronic device, characterized in that: include: a memory having a computer program stored thereon; A processor, configured to execute the computer program in the memory to implement the steps of the method according to any one of claims 1 to 7.

Images