A strategy generation method, device and equipment based on counterfactual reasoning and a storage medium

Abstract

This invention discloses a strategy generation method, apparatus, device, and storage medium based on counterfactual deduction. The method includes: during virtual interaction between a user agent and an interactive agent, determining the user's acceptance probability for each interaction strategy under the corresponding psychological state parameters based on the total attractiveness of the interactive agent's various interaction strategies and the user agent's corresponding psychological state parameters; determining the counterfactual conversion rate of each interaction strategy based on the user's acceptance probability under each set of psychological state parameters; determining the maximum counterfactual conversion rate among the counterfactual conversion rates of each interaction strategy, and determining the interaction strategy corresponding to the maximum counterfactual conversion rate as the target interaction strategy. This technical solution, through counterfactual deduction, determines a target interaction strategy that matches the user, increasing the probability of the user accepting the interactive content and avoiding user churn due to inappropriate strategies.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of artificial intelligence technology, and in particular to a strategy generation method, apparatus, device and storage medium based on counterfactual deduction. Background Technology

[0002] For product transaction scenarios with long decision-making cycles and high average order values, such as auto finance, real estate leasing, and high-end insurance, determining interaction strategies to increase product transaction volume has become an urgent technical problem to be solved.

[0003] In existing technologies, targeted information pushes are typically sent to users with high product intent based on their product intent tags. When the number of high-intent users is limited, a rating model can be used to determine the score corresponding to each interaction strategy, and then information interaction can be performed on the user based on the interaction strategy with the highest score.

[0004] However, the interaction strategy determined in this way is difficult to adapt to the complex and ever-changing market environment. Summary of the Invention

[0005] This invention provides a strategy generation method, apparatus, device, and storage medium based on counterfactual deduction to determine an interaction strategy that matches the user.

[0006] In a first aspect, embodiments of the present invention provide a strategy generation method based on counterfactual deduction, comprising: During the virtual interaction between the user agent and the interactive agent, the user's acceptance probability for each interactive strategy under the psychological state parameters is determined based on the total attractiveness of each interactive strategy adopted by the interactive agent and the psychological state parameters corresponding to the user agent. The user agent is constructed based on the user's user characteristics and psychological state parameters, and the interactive agent is constructed based on multiple interactive strategies. The counterfactual conversion rate of each interaction strategy is determined based on the user's acceptance probability for each interaction strategy under each set of psychological state parameters. Determine the maximum counterfactual conversion rate among the counterfactual conversion rates of each of the interaction strategies, and determine the interaction strategy corresponding to the maximum counterfactual conversion rate as the target interaction strategy.

[0007] The technical solution of this invention provides a strategy generation method based on counterfactual deduction, comprising: during a virtual interaction between a user agent and an interactive agent, determining the user's acceptance probability for each interactive strategy under the psychological state parameters based on the total attractiveness of each interactive strategy adopted by the interactive agent and the psychological state parameters corresponding to the user agent, wherein the user agent is constructed based on the user's user characteristics and psychological state parameters, and the interactive agent is constructed based on multiple interactive strategies; determining the counterfactual conversion rate of each interactive strategy based on the user's acceptance probability for each interactive strategy under each set of psychological state parameters; determining the maximum counterfactual conversion rate among the counterfactual conversion rates of each interactive strategy, and determining the interactive strategy corresponding to the maximum counterfactual conversion rate as the target interactive strategy. The above technical solution, during the virtual interaction between the user agent and the interactive agent, determines the user's acceptance probability for each interaction strategy under the corresponding psychological state parameters based on the total attractiveness of the interaction strategy adopted by the interactive agent and the corresponding psychological state parameters of the user agent. This achieves the determination of the user's acceptance probability for each interaction strategy under each set of psychological state parameters. Based on the user's acceptance probability for each interaction strategy under each set of psychological state parameters, the counterfactual conversion rate of each interaction strategy is determined, thus establishing a baseline conversion rate for the interaction strategy without intervention. The maximum counterfactual conversion rate among the counterfactual conversion rates of each interaction strategy is then determined, and the interaction strategy corresponding to the maximum counterfactual conversion rate is identified as the target interaction strategy for product promotion to the user. Through counterfactual deduction, a target interaction strategy matching the user is determined, increasing the probability of the user accepting the interactive content and avoiding user churn due to inappropriate strategies.

[0008] Furthermore, the user intelligent agent construction process includes: The system acquires user behavior data in applications displaying promoted products, interaction record data in the user's system, and the user's historical transaction data. It then extracts features from the user behavior data, interaction record data, and historical transaction data to construct the user's user characteristics based on the extracted features. The user features are input into a pre-trained state extraction model, and multiple sets of psychological state parameters are extracted from the user features based on the state extraction model; or, multiple sets of historical user features of the user are obtained, and multiple sets of psychological state parameters are obtained by fusing each of the historical user features and the user features, wherein the psychological state parameters are reflected by time sensitivity, resistance threshold, decision preference and time sensitivity; The user intelligent agent is constructed based on the user characteristics and the psychological state parameters of each group.

[0009] Furthermore, the process of constructing the interactive intelligent agent includes: Multiple interaction strategies are determined by combining multiple interaction methods included in each interaction dimension through a Cartesian product. The interaction dimensions include at least financial solutions, communication style, and reach time. The interactive agent is constructed based on multiple interaction strategies.

[0010] Further, the probability of a user accepting each interaction strategy under the psychological state parameters is determined based on the total attractiveness of the interaction strategies adopted by the interactive agent and the psychological state parameters corresponding to the user agent, including: Determine the total attractiveness of the interaction strategy; Substituting the total attractiveness of the interaction strategy and the psychological state parameters corresponding to the user agent into the acceptance probability formula, the user's acceptance probability of the interaction strategy under the psychological state parameters is obtained.

[0011] Further, determining the total attractiveness of the interaction strategy includes: Determine the financial appeal, persuasive appeal, and temporal appeal of the interaction strategy; The total appeal of the interaction strategy is determined based on the financial appeal, the verbal appeal, and the temporal appeal.

[0012] Further, based on the user's acceptance probability for each interaction strategy under each group of psychological state parameters, the counterfactual conversion rate of each interaction strategy is determined, including: For each of the aforementioned interaction strategies, the counterfactual conversion rate of the interaction strategy is obtained by weighted summation and normalization of the user's acceptance probability for the interaction strategy under each of the aforementioned psychological state parameters.

[0013] Further, determining the maximum counterfactual conversion rate among the counterfactual conversion rates of each of the aforementioned interaction strategies includes: Construct a probability surface based on the counterfactual transformations of each of the aforementioned interaction strategies; Gradient calculation is performed on the probability surface to determine the maximum counterfactual conversion rate.

[0014] Secondly, embodiments of the present invention also provide a strategy generation apparatus based on counterfactual deduction, comprising: An interaction module is used to determine the probability of a user accepting each interaction strategy under the psychological state parameters during the virtual interaction between the user agent and the interactive agent, based on the total attractiveness of each interaction strategy adopted by the interactive agent and the psychological state parameters corresponding to the user agent. The user agent is constructed based on the user's user characteristics and psychological state parameters, and the interactive agent is constructed based on multiple interaction strategies. The determining module is used to determine the counterfactual conversion rate of each interaction strategy based on the user's acceptance probability of each interaction strategy under each group of psychological state parameters; An execution module is used to determine the maximum counterfactual conversion rate among the counterfactual conversion rates of each of the interaction strategies, and to determine the interaction strategy corresponding to the maximum counterfactual conversion rate as the target interaction strategy.

[0015] Thirdly, embodiments of the present invention also provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the strategy generation method based on counterfactual deduction as described in any of the first aspects.

[0016] Fourthly, embodiments of the present invention also provide a storage medium containing computer-executable instructions, characterized in that the computer-executable instructions, when executed by a computer processor, are used to perform a strategy generation method based on counterfactual deduction as described in any of the first aspects.

[0017] Fifthly, this application provides a computer program product including computer instructions. When the computer instructions are executed on a computer, the computer causes the computer to perform the counterfactual deduction-based strategy generation method provided in the first aspect.

[0018] Computer instructions may be stored, in whole or in part, on a computer-readable storage medium. This computer-readable storage medium may be packaged together with the processor of the counterfactual policy generation device, or it may be packaged separately from the processor of the counterfactual policy generation device; this application does not impose any limitation on this.

[0019] The descriptions of the second, third, fourth, and fifth aspects in this application can be referred to the detailed description of the first aspect; and the beneficial effects of the descriptions of the second, third, fourth, and fifth aspects can be referred to the analysis of the beneficial effects of the first aspect, which will not be repeated here.

[0020] In this application, the names of the aforementioned counterfactual strategy generation devices do not limit the devices or functional modules themselves. In actual implementation, these devices or functional modules may appear under other names. As long as the functions of each device or functional module are similar to those in this application, they fall within the scope of the claims of this application and their equivalents.

[0021] These or other aspects of this application will become more readily apparent in the following description. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0023] Figure 1 A flowchart illustrating a strategy generation method based on counterfactual deduction provided in an embodiment of the present invention; Figure 2 A flowchart of another strategy generation method based on counterfactual deduction provided in an embodiment of the present invention; Figure 3 A schematic diagram of a strategy generation device based on counterfactual deduction provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0024] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the present invention, and not all of the structures.

[0025] In this article, the term "and / or" is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.

[0026] The terms "first" and "second," etc., used in the specification and drawings of this application are used to distinguish different objects or to distinguish different treatments of the same object, rather than to describe a specific order of objects.

[0027] Furthermore, the terms "comprising" and "having," and any variations thereof, used in the description of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the steps or units listed, but may optionally include other steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus.

[0028] Before discussing the exemplary embodiments in more detail, it should be noted that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe the operations (or steps) as sequential processes, many of these operations can be performed in parallel, concurrently, or simultaneously. Furthermore, the order of the operations can be rearranged. The process can be terminated when its operation is completed, but may also have additional steps not included in the figures. The process can correspond to a method, function, procedure, subroutine, subroutine, etc. Moreover, embodiments and features in the embodiments of the present invention can be combined with each other without conflict.

[0029] It should be noted that in the embodiments of this application, the words "exemplary" or "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of the words "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0030] In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0031] Figure 1 The flowchart illustrates a strategy generation method based on counterfactual deduction, as provided in an embodiment of the present invention. This method can be executed by a strategy generation device based on counterfactual deduction, such as... Figure 1 As shown, the specific steps include the following: Step 110: During the virtual interaction between the user agent and the interactive agent, the user's acceptance probability for each interaction strategy under the psychological state parameters is determined based on the total attractiveness of each interaction strategy adopted by the interactive agent and the psychological state parameters corresponding to the user agent.

[0032] User intelligent agents are constructed based on user characteristics and psychological state parameters. Specifically, when a user browses an application displaying a product, their browsing trajectory, page dwell time, and form completion rate constitute user behavior data. Feature extraction is performed on this user behavior data, interaction records within the user's system, and historical transaction data to obtain user characteristics. Based on these user characteristics, psychological state parameters can be further determined. After determining the user characteristics and multiple sets of psychological state parameters, a user intelligent agent can be constructed. This agent can simulate the user's decision-making behavior in a virtual environment.

[0033] An interactive agent is constructed based on multiple interaction strategies. Specifically, a product's interaction strategy typically consists of multiple interaction dimensions, each containing various interaction methods. By combining each interaction method within each dimension, multiple interaction strategies can be generated. These multiple interaction strategies can form an interaction strategy set, and an interactive agent can be constructed based on this set. The interactive agent can simulate real-world interactive behaviors in a virtual environment.

[0034] Specifically, after creating the virtual negotiation environment, the user agent and the interaction agent can be initialized. The user agent can engage in virtual interaction with the interaction agent. The interaction agent selects an interaction strategy from the set of interaction strategies and initiates virtual contact with the user agent. The user agent can determine the probability of the user accepting the interaction strategy under multiple psychological state parameters. The probability of the user accepting the interaction strategy under specific psychological state parameters is determined based on the total attractiveness of the interaction strategy and the current psychological state parameters. Specifically, the probability of the user accepting the interaction strategy under that set of psychological state parameters can be determined by substituting the total attractiveness of the interaction strategy and the resistance threshold included in the psychological state parameters into the acceptance probability formula.

[0035] In practical applications, a corresponding response can be generated based on the acceptance probability. For example, if the acceptance probability is greater than the probability threshold, an acceptance response can be generated; otherwise, a hesitant or reject response can be generated according to the specific value of the acceptance probability.

[0036] In this embodiment of the invention, the user intelligent agent determines the corresponding acceptance probability based on the total attractiveness of the interaction strategy adopted by the interaction intelligent agent and the corresponding psychological state parameters of the user intelligent agent during the interaction process, thereby determining the user's acceptance probability for each interaction strategy under each set of psychological state parameters.

[0037] Step 120: Determine the counterfactual conversion rate of each interaction strategy based on the user's acceptance probability for each interaction strategy under each group of psychological state parameters.

[0038] By statistically analyzing the probability of users accepting interaction strategies under various psychological state parameters, we can determine the counterfactual conversion rate corresponding to the interaction strategy, and then determine the target interaction strategy for the user based on the counterfactual conversion rate.

[0039] Therefore, it is possible to statistically analyze the probability of users accepting interaction strategies under different sets of psychological state parameters. Specifically, the counterfactual conversion rate of the interaction strategy can be obtained by weighted summation and normalization of the probability of user acceptance of the interaction strategy under each set of psychological state parameters.

[0040] In this embodiment of the invention, the counterfactual conversion rate of the interaction strategy is determined by combining the user's acceptance probability of the interaction strategy under each group of psychological state parameters, thereby realizing the determination of the baseline conversion rate of the interaction strategy without intervention.

[0041] Step 130: Determine the maximum counterfactual conversion rate among the counterfactual conversion rates of each of the interaction strategies, and determine the interaction strategy corresponding to the maximum counterfactual conversion rate as the target interaction strategy.

[0042] Specifically, after determining the counterfactual conversion rate of each interaction strategy, a multidimensional surface can be fitted and generated based on these rates. The dimensions of the multidimensional surface correspond to multiple promotions, and the surface height corresponds to the counterfactual conversion rate. Furthermore, the maximum surface height can be determined within the multidimensional surface; the maximum surface height corresponds to the maximum counterfactual conversion rate, and the promotion strategy corresponding to the maximum counterfactual conversion rate is the target interaction strategy. Interacting with the user based on the target interaction strategy maximizes the probability of a positive response.

[0043] In this embodiment of the invention, by determining the maximum counterfactual conversion rate among the counterfactual conversion rates of each interaction strategy, the interaction strategy corresponding to the maximum counterfactual conversion rate is determined as the target interaction strategy for interacting with the user. Through counterfactual deduction, the target interaction strategy matching the user is determined, thereby achieving accurate strategy determination for the user and increasing the probability that the user will accept the response to the interactive content.

[0044] The counterfactual inference-based strategy generation method provided in this invention includes: during a virtual interaction between a user agent and an interactive agent, determining the user's acceptance probability for each interactive strategy under the psychological state parameters based on the total attractiveness of each interactive strategy adopted by the interactive agent and the corresponding psychological state parameters of the user agent, wherein the user agent is constructed based on the user's user characteristics and psychological state parameters, and the interactive agent is constructed based on multiple interactive strategies; determining the counterfactual conversion rate of each interactive strategy based on the user's acceptance probability for each interactive strategy under each set of psychological state parameters; determining the maximum counterfactual conversion rate among the counterfactual conversion rates of each interactive strategy, and determining the interactive strategy corresponding to the maximum counterfactual conversion rate as the target interactive strategy. The above technical solution, during the virtual interaction between the user agent and the interactive agent, determines the user's acceptance probability for each interaction strategy under the corresponding psychological state parameters based on the total attractiveness of the interaction strategy adopted by the interactive agent and the corresponding psychological state parameters of the user agent. This achieves the determination of the user's acceptance probability for each interaction strategy under each set of psychological state parameters. Based on the user's acceptance probability for each interaction strategy under each set of psychological state parameters, the counterfactual conversion rate of each interaction strategy is determined, thus establishing a baseline conversion rate for the interaction strategy without intervention. The maximum counterfactual conversion rate among the counterfactual conversion rates of each interaction strategy is then determined, and the interaction strategy corresponding to the maximum counterfactual conversion rate is identified as the target interaction strategy for product promotion to the user. Through counterfactual deduction, a target interaction strategy matching the user is determined, increasing the probability of the user accepting the interactive content and avoiding user churn due to inappropriate strategies.

[0045] Figure 2 This is a flowchart of another strategy generation method based on counterfactual deduction provided in an embodiment of the present invention. This embodiment is a specific implementation based on the above embodiments. Figure 2 As shown, in this embodiment, the method may further include: Step 210: During the virtual interaction between the user agent and the interactive agent, the user's acceptance probability for each interaction strategy under the psychological state parameters is determined based on the total attractiveness of each interaction strategy adopted by the interactive agent and the psychological state parameters corresponding to the user agent.

[0046] To increase product sales, it is necessary to engage in appropriate interactions with as many users as possible who have accessed the application. Therefore, it is necessary to identify these users and determine the corresponding interaction strategies for each user among all users who have accessed the application.

[0047] Users deemed low-value by traditional static scoring models can be stored in a discarded user pool. For these users, counterfactual analysis can be performed in a non-standard strategy space to uncover their potential value. Specifically, a user agent can be constructed for each user. In a non-standard interaction strategy space (including rent-to-own, long-tail term loans, flexible repayment, and other non-standard financial solutions), counterfactual probability surface calculations are performed on each non-standard interaction strategy to obtain the counterfactual conversion probability under that strategy. The non-standard interaction strategy corresponding to the highest counterfactual conversion probability is determined as the optimal non-standard interaction strategy for the user, and the corresponding maximum conversion probability is recorded. If the maximum conversion probability exceeds the recall threshold τ (usually set to 5%), the user is considered to have potential value, triggering a recall process. Specifically, the user can be migrated from the discarded user pool to a high-potential user pool. By further screening high-potential users in the discarded user pool, the user screening scope is further improved, and the number of effective users is expanded.

[0048] In practical applications, the abandoned user pool also includes blacklisted users (such as those who have been rejected more than 3 times), users who obviously do not meet the criteria (such as those who are <18 years old or >70 years old), and users with invalid contact information (such as empty numbers or disconnected numbers). Before mining potential value, these users need to be screened out to improve mining efficiency.

[0049] For each user in the high-potential user pool, a corresponding user intelligent agent can also be constructed.

[0050] In one implementation, the user agent construction process includes: The system acquires user behavior data in applications displaying promoted products, interaction records in the user's system, and historical transaction data. It then extracts features from these data to construct user characteristics. These user characteristics are input into a pre-trained state extraction model, which extracts multiple sets of psychological state parameters. Alternatively, the system acquires multiple sets of historical user characteristics and fuses these characteristics with the current user characteristics to obtain multiple sets of psychological state parameters, where these parameters are characterized by time sensitivity, resistance threshold, decision preference, and time sensitivity. Finally, the system constructs a user agent based on these user characteristics and the sets of psychological state parameters.

[0051] Specifically, by setting up tracking points in applications showcasing products, user behavior data within these applications can be obtained. This allows for the acquisition of user interaction records and historical transaction data within the user's system. Historical transaction data may include past loan records and repayment patterns. Based on feature engineering and vectorization techniques, features are extracted from the user behavior data, interaction records, and historical transaction data to obtain the user's characteristics. .

[0052] Furthermore, psychological state parameters can be determined based on user characteristics. On one hand, user characteristics can be input into a pre-trained state extraction model. Based on this model, multiple sets of psychological state parameters can be extracted from the user characteristics. These extracted parameters include time sensitivity, resistance threshold, decision preference, and time sensitivity. Time sensitivity reflects the user's sensitivity to price changes; the resistance threshold reflects the price / condition threshold at which the user begins to develop resistance; decision preference reflects the user's preferred type of financial solution (e.g., preference for low down payment vs. low interest rate); and time sensitivity reflects the user's sensitivity to the timing of the contact. On the other hand, observational data can be distributed based on multiple sets of historical user characteristics determined from historical user behavior data, historical interaction record data, and historical transaction data. Multiple sets of historical user features are randomly obtained from the dataset. A pre-trained mapping function is then used to map these historical user features. With user characteristics By fusing the data, the corresponding psychological state parameters can be obtained. Psychological state parameters Including price sensitivity Resistance threshold Decision preferences and time sensitivity .

[0053] In practical applications, the price sensitivity and decision preference vectors included in the psychological state parameters dynamically evolve with feedback from the external environment and internal interactions. For example, it's possible to monitor competitor price reduction news (using keywords such as "XX brand price reduction" via news APIs), market interest rate fluctuations (using LPR changes via financial data APIs), and policy changes (using regulatory announcement APIs). When these events are detected, an incremental update of price sensitivity can be triggered, meaning the price sensitivity can be updated to... ,in, It represents the learning rate; it can also monitor user dwell time on a page exceeding a threshold, changes in response intensity to certain types of language (e.g., click-through rate jumping from 10% to 30%), and changes in browsing behavior (e.g., switching from browsing the "low down payment" tag to browsing the "zero interest rate" tag). When these events are detected, an incremental update of the decision preference vector can be triggered, that is, the decision preference vector can be updated to... ,in, This represents the learning rate. The updated mental state parameters take effect immediately, ensuring that the next game is based on real-time mental state parameters. By sensing market fluctuations (such as competitor price reductions) and changes in user psychology, and adjusting mental state parameters in real time, the timeliness of the determined interaction strategy is guaranteed.

[0054] Furthermore, a user intelligent agent can be constructed based on user characteristics and psychological state parameters of each group.

[0055] In one implementation, the interactive agent construction process includes: Multiple interaction strategies are determined by combining the Cartesian product of multiple interaction methods included in each interaction dimension, wherein the interaction dimension includes at least financial scheme, communication style and reach time; the interactive intelligent agent is constructed based on the multiple interaction strategies.

[0056] Specifically, a product's interaction strategy typically consists of multiple interaction dimensions, such as financial dimensions, communication language dimensions, and time of reach dimensions. Each interaction dimension includes multiple interaction methods. For example, the financial dimension may include down payment ratios (e.g., 10%, 20%, 30%, 40%, 50%), interest rate ranges (e.g., 3.5%, 4.0%, 4.5%, 5.0%, 5.5%), and repayment ranges (e.g., 12, 24, 36, 48, 60). The communication language dimension may include rational persuasion (emphasizing data comparison and cost analysis), emotional resonance (emphasizing improved quality of life and family happiness), and urgency-driven (emphasizing limited-time offers and limited availability). The time of reach dimension may include date divisions (weekdays vs. weekends) and time period divisions (morning (9:00-12:00) vs. afternoon (14:00-18:00) vs. evening (19:00-21:00)). By combining the Cartesian products of each interaction method in each interaction dimension, multiple interaction strategies can be determined, and these multiple interaction strategies can form an interaction strategy set. , of which each It represents a complete interaction strategy, and an interactive intelligent agent can be constructed based on the set of interaction strategies.

[0057] The interactive agent is trained using the Soft Actor-Critic (SAC) algorithm based on the maximum entropy framework, and the reward function used for training is... ,in, Let represent the probability of a user with mental state parameter s converting to interaction strategy a. This indicates the cost of action for interaction strategy a (such as the complexity of the message or the discount offered). and These are the weighting coefficients.

[0058] In real-world business scenarios, the weight of action costs relative to conversion probability weights Typically smaller (i.e.) Under these constraints, the reward function strictly satisfies quasi-convexity in the local policy space. This mathematical property theoretically guarantees that the gradient ascent algorithm can converge to a local optimum, avoiding the risk of getting stuck in a saddle point. When the action cost weight... Exceeding the safety threshold (Threshold here) When searching for a target interaction strategy, an adaptive population evolution strategy can be used to determine the target interaction strategy, ensuring the robustness of the search target interaction strategy.

[0059] The Soft Actor-Critic (SAC) algorithm within the maximum entropy framework employs a maximum entropy objective function that maximizes the transformation probability while introducing a policy entropy regularization term. ,in, Represents the entropy term. This represents the temperature coefficient. This improvement allows the interactive agent to maintain diversity in interaction strategies during the exploration phase, avoiding premature convergence to local optima (i.e., avoiding always recommending the same solution), thus covering more potential interaction strategies. Furthermore, to address the overestimation bias problem of traditional DQN, two parallel Critic networks are constructed. When calculating the target value, the minimum of the two values ​​is taken: This greatly improves the numerical stability in complex financial game scenarios.

[0060] In one implementation, determining the user's acceptance probability for each interaction strategy under the given psychological state parameters, based on the total attractiveness of the interaction strategies employed by the interactive agent and the corresponding psychological state parameters of the user agent, includes: Determine the total attractiveness of the interaction strategy; substitute the total attractiveness of the interaction strategy and the psychological state parameter corresponding to the user agent into the acceptance probability formula to obtain the user's acceptance probability of the interaction strategy under the psychological state parameter.

[0061] Further, determining the total attractiveness of the interaction strategy includes: Determine the financial appeal, verbal appeal, and temporal appeal of the interaction strategy; and determine the total appeal of the interaction strategy based on the financial appeal, verbal appeal, and temporal appeal.

[0062] By using the user's psychological state parameters as the query and the financial, verbal, and time-of-reach dimensions of the interaction strategy as the key and value, attention weighting is used to capture the non-linear interaction between the psychological state parameters and the promotion parameters. This mechanism transforms the attraction function A from a simple linear weighted sum into a high-order feature representation extracted by a deep neural network.

[0063] Specifically, interaction strategy For a given psychological state parameter Total user appeal It can be determined using Formula 1.

[0064] Formula 1 in, Indicating financial attractiveness, The exponential function ensures that financial attractiveness monotonically decreases as the cost difference of the interaction strategy increases, accurately simulating the changing trend of users' decision-making intentions. Indicate the attractiveness of the sales pitch. It indicates the attractiveness of time.

[0065] After determining the total attractiveness of the interaction strategy, the total attractiveness and the resistance threshold included in the psychological state parameters can be substituted into the acceptance probability formula, i.e., into Formula 2, to determine the user's acceptance probability of the interaction strategy under the psychological state parameters. .

[0066] Formula 2 Where k represents the sensitivity parameter, and the value of k is selected through rigorous sensitivity analysis to ensure numerical stability. For example, k=5. Indicates a correction term.

[0067] It should be noted that a batch parallel processing architecture can be adopted to conduct concurrent sandbox simulations for multiple user agents, exploring in parallel the acceptance probability of each user for each interaction strategy under various psychological state parameters. Specifically, users in the high-potential user pool can be divided into pre-set batch sizes, and users within each batch can be processed in parallel. For each lead user, a corresponding user agent can be constructed. In a virtual sandbox environment, an interactive agent and the user agent engage in multiple rounds of game theory to determine the user's acceptance probability for each interaction strategy under various psychological state parameters. By leveraging the parallel computing capabilities of the GPU cluster, multiple batches of users can be processed simultaneously, enabling concurrent simulations for a large number of user agents, achieving industrial-grade real-time processing capabilities, and controlling the simulation time of each user agent to within 100 milliseconds.

[0068] During the game, a trajectory anchoring algorithm can be executed based on the complete game trajectory to determine a sequence of decision snapshots containing key nodes and correlate them with the deductive conversation. Specifically, each time point in the game trajectory can be traversed, and the change in user intention value between adjacent time points Δintention=I(ti)-I(ti-1) can be calculated. If the change in user intention value exceeds a preset trigger threshold, it is determined to be a trigger node. The information recorded at the trigger node includes: game round, trigger action (strategy parameters), intention value transition magnitude, timestamp, user psychological state vector, and key phrase ID. User state transitions can also be detected. If the user's state changes from firm rejection to hesitant consideration or positive consideration at adjacent time points, it is determined to be a turning point. The information recorded at the turning point node includes: game round, trigger action, state transition pair (previous state, subsequent state), timestamp, and change in user sentiment value. It can also detect the time point when a user finally agrees to the conversion in the game trajectory. This time point is the decision node. The information recorded at the decision node includes: the game round, the strategy parameters for final agreement, the final conversion probability, the timestamp, and the complete decision conditions (financial solution parameters, communication style, and timing of contact). A decision snapshot is constructed for each key node. The decision snapshot includes: node type (trigger / turning point / decision), timestamp, financial dimension, psychological state parameters, key communication ID, state transition information, etc. All decision snapshots are serialized and stored in chronological order to form a complete evidence chain of the game trajectory. Of course, a unique simulation session ID can also be generated for each simulation and associated with the decision snapshot sequence. Through trajectory anchoring algorithms, it is ensured that the target interaction strategy has an interpretable evidence chain, effectively reducing the churn rate of high-potential users.

[0069] During the virtual interaction between the user agent and the interactive agent, the real user characteristics of the user corresponding to the user agent can be collected in real time. The Euclidean distance between the real user characteristics and the user characteristics used to construct the user agent can be calculated. When the Euclidean distance is greater than a distance threshold, the interaction is determined to have failed. At this point, aggressive information push strategies (such as high-frequency follow-up questions) can be stopped, and a reassuring fallback strategy or a transfer to a human agent can be automatically implemented. At the same time, the user agent can be corrected based on the real user characteristics.

[0070] In this embodiment of the invention, the user intelligent agent determines the corresponding acceptance probability based on the total attractiveness of the interaction strategy adopted by the interaction intelligent agent and the corresponding psychological state parameters of the user intelligent agent during the interaction process, thereby determining the user's acceptance probability for each interaction strategy under each set of psychological state parameters.

[0071] Step 220: For each of the interaction strategies, the counterfactual conversion rate of the interaction strategy is obtained by weighted summation and normalization of the user's acceptance probability of the interaction strategy under each of the psychological state parameters.

[0072] Specifically, for each interaction strategy Based on Formula 3, the counterfactual conversion rate, the probability of a user accepting an interaction strategy under specific psychological state parameters, can be determined.

[0073] Formula 3 In practical applications, continuous integrals of high-dimensional features are difficult to solve analytically directly in engineering systems. Therefore, the Monte Carlo sampling method is used in actual calculations to discretize and numerically approximate the integral equation of the counterfactual conversion rate. The specific algorithm logic steps are as follows: First, starting from the observation data distribution pre-constructed based on historical user characteristics... In the process, a set of historical user feature samples were randomly selected. Sample size The accuracy is set according to the requirements (usually on the order of 10,000). Then, a pre-trained mapping function is used. Historical user feature samples With user characteristics The data is then integrated and converted into corresponding psychological state parameters. Interaction strategy With psychological state parameters Substituting into Formula 2, we obtain the user's psychological state parameters. The following is about interaction strategies Acceptance probability The probability of user acceptance of the interaction strategy under each set of psychological state parameters is weighted, summed, and normalized to obtain the interaction strategy. Counterfactual conversion rate .

[0074] In this embodiment of the invention, the counterfactual conversion rate of the interaction strategy is determined by combining the user's acceptance probability of the interaction strategy under each group of psychological state parameters, thereby realizing the determination of the baseline conversion rate of the interaction strategy without intervention.

[0075] Step 230: Construct a probability surface based on the counterfactual transformation of each interaction strategy; perform gradient calculation on the probability surface to determine the maximum counterfactual transformation rate.

[0076] Step 240: Determine the interaction strategy corresponding to the maximum counterfactual conversion rate as the target interaction strategy.

[0077] Specifically, firstly, a probability surface can be constructed based on the counterfactual transformations of each interaction strategy. Secondly, an iterative optimization algorithm can be used to locate the peak point on the probability surface. The specific process is as follows: First, an initial interaction strategy is selected from the set of interaction strategies. The initial interaction strategy can be randomly selected or based on prior knowledge, choosing an interaction strategy corresponding to a higher probability counterfactual conversion rate. Then, in the current interaction strategy... At this point, the numerical gradient of the probability surface is calculated using the finite difference method or automatic differentiation technique. The gradient direction points in the direction of the fastest probability growth. Then, along the gradient direction, a preset learning rate is applied. Update strategy parameters: Update the strategy parameters Project the parameters onto the effective range of the interaction strategy, ensuring that each parameter meets the constraints of each interaction dimension (e.g., down payment ratio between 10% and 50%, interest rate between 3.5% and 5.5%). Calculate the Euclidean distance between the strategy parameters before and after the update. If the distance is less than the preset convergence threshold If the algorithm converges, the current promotion strategy is output as the target interaction strategy. Otherwise, As the new current interaction strategy, return to the execution of the current interaction strategy. At each point, the numerical gradient of the probability surface is calculated using the finite difference method or automatic differentiation technique until the maximum number of iterations is reached or the convergence condition is met.

[0078] After determining the target interaction strategy, if its execution requires high negotiation skills, it indicates that the promoter's ability score needs to be greater than a threshold. If the current promoter's ability score is not greater than the threshold, the promoter can be automatically switched. Alternatively, if the target strategy is deemed suitable for automatic execution, AI-powered promotion can be switched. The promoter's ability score is the output of a pre-trained ability scoring model after inputting the promoter's ability information.

[0079] When the product is a vehicle, specific information from the vehicle finance sector can be collected, including browsing behavior data, interaction records, and historical transaction data of potential car buyers. A user agent is constructed based on this specific information. Multiple rounds of virtual interaction are conducted between the vehicle's interactive agent and the user agent. During the interaction, the user agent calculates the counterfactual conversion rate of the interaction strategy, and the interaction strategy corresponding to the highest counterfactual conversion rate is determined as the target interaction strategy. The triggering nodes, turning points, and decision points of the user agent in the game process are identified, generating a chain of evidence for the game trajectory.

[0080] When the product is a training course, specific information related to the training course domain can be collected, including browsing behavior data, interaction log data, and historical transaction data of potential learners. A user agent is constructed based on this specific information. The interaction log data can include educational background and coupon selection records. The interaction dimensions of the training course can include class type (live class, pre-recorded class, face-to-face class), payment method (full payment, 3-month installment, 6-month installment), communication style (career prospect driven, price reduction promotion driven, case study incentive driven), and value-added services (job referral, resume optimization, mock interview). Multiple rounds of virtual interaction are conducted between the training course's interactive agent and the user agent. During the interaction, the user agent calculates the counterfactual conversion rate of the interaction strategy, and the interaction strategy corresponding to the maximum counterfactual conversion rate is determined as the target interaction strategy. The trigger nodes, turning points, and decision nodes of the user agent in the game process are identified, generating a game trajectory evidence chain.

[0081] In this embodiment of the invention, by determining the maximum counterfactual conversion rate among the counterfactual conversion rates of each interaction strategy, the interaction strategy corresponding to the maximum counterfactual conversion rate is determined as the target interaction strategy for promoting the product to the user. Through counterfactual deduction, a target interaction strategy matching the user is determined, thereby increasing the probability that the user will accept the interactive content.

[0082] The counterfactual inference-based strategy generation method provided in this invention includes: during a virtual interaction between a user agent and an interactive agent, determining the user's acceptance probability for each interaction strategy under the psychological state parameters based on the total attractiveness of each interaction strategy adopted by the interactive agent and the psychological state parameters corresponding to the user agent; for each interaction strategy, obtaining the counterfactual conversion rate of the interaction strategy by performing weighted summation and normalization on the user's acceptance probability for the interaction strategy under each psychological state parameter; constructing a probability surface based on the counterfactual conversion of each interaction strategy; performing gradient calculation on the probability surface to determine the maximum counterfactual conversion rate; and determining the interaction strategy corresponding to the maximum counterfactual conversion rate as the target interaction strategy. The above technical solution, during the virtual interaction between the user agent and the interactive agent, determines the user's acceptance probability for each interaction strategy under the corresponding psychological state parameters based on the total attractiveness of the interaction strategy adopted by the interactive agent and the corresponding psychological state parameters of the user agent. This achieves the determination of the user's acceptance probability for each interaction strategy under each set of psychological state parameters. Based on the user's acceptance probability for each interaction strategy under each set of psychological state parameters, the counterfactual conversion rate of each interaction strategy is determined, thus establishing a baseline conversion rate for the interaction strategy without intervention. The maximum counterfactual conversion rate among the counterfactual conversion rates of each interaction strategy is then determined, and the interaction strategy corresponding to the maximum counterfactual conversion rate is identified as the target interaction strategy for product promotion to the user. Through counterfactual deduction, a target interaction strategy matching the user is determined, increasing the probability of the user accepting the interactive content and avoiding user churn due to inappropriate strategies.

[0083] Figure 3 This is a schematic diagram of a strategy generation device based on counterfactual reasoning, provided as an embodiment of the present invention. This device can be implemented through software and / or hardware and is generally integrated into electronic devices, such as computer devices.

[0084] like Figure 3 As shown, the device includes: The interaction module 310 is used to determine the user's acceptance probability of each interaction strategy under the psychological state parameters during the virtual interaction between the user intelligent agent and the interactive intelligent agent, based on the total attractiveness of each interaction strategy adopted by the interactive intelligent agent and the psychological state parameters corresponding to the user intelligent agent. The user intelligent agent is constructed based on the user's user characteristics and psychological state parameters, and the interactive intelligent agent is constructed based on multiple interaction strategies. The determining module 320 is used to determine the counterfactual conversion rate of each interaction strategy based on the user's acceptance probability of each interaction strategy under each group of psychological state parameters; The execution module 330 is used to determine the maximum counterfactual conversion rate among the counterfactual conversion rates of each of the interaction strategies, and to determine the interaction strategy corresponding to the maximum counterfactual conversion rate as the target interaction strategy.

[0085] The counterfactual inference-based strategy generation device provided in this embodiment determines the user's acceptance probability for each interaction strategy under the psychological state parameters during virtual interaction between the user agent and the interactive agent, based on the total attractiveness of each interaction strategy adopted by the interactive agent and the corresponding psychological state parameters of the user agent. The user agent is constructed based on the user's user characteristics and psychological state parameters, and the interactive agent is constructed based on multiple interaction strategies. Based on the user's acceptance probability for each interaction strategy under each set of psychological state parameters, the counterfactual conversion rate of each interaction strategy is determined. The device then determines the maximum counterfactual conversion rate among the counterfactual conversion rates of each interaction strategy and identifies the interaction strategy corresponding to the maximum counterfactual conversion rate as the target interaction strategy. The above technical solution, during the virtual interaction between the user agent and the interactive agent, determines the user's acceptance probability for each interaction strategy under the corresponding psychological state parameters based on the total attractiveness of the interaction strategy adopted by the interactive agent and the corresponding psychological state parameters of the user agent. This achieves the determination of the user's acceptance probability for each interaction strategy under each set of psychological state parameters. Based on the user's acceptance probability for each interaction strategy under each set of psychological state parameters, the counterfactual conversion rate of each interaction strategy is determined, thus establishing a baseline conversion rate for the interaction strategy without intervention. The maximum counterfactual conversion rate among the counterfactual conversion rates of each interaction strategy is then determined, and the interaction strategy corresponding to the maximum counterfactual conversion rate is identified as the target interaction strategy for product promotion to the user. Through counterfactual deduction, a target interaction strategy matching the user is determined, increasing the probability of the user accepting the interactive content and avoiding user churn due to inappropriate strategies.

[0086] Based on the above embodiments, the device further includes: The first construction module is used to acquire user behavior data of the user in applications displaying promoted products, interaction record data of the user's system, and the user's historical transaction data. It then performs feature extraction on the user behavior data, interaction record data, and historical transaction data to construct the user's features based on the extracted features. The user features are input into a pre-trained state extraction model, and multiple sets of psychological state parameters are extracted from the user features based on the state extraction model. Alternatively, multiple sets of historical user features are acquired, and multiple sets of psychological state parameters are obtained by fusing these historical user features with the current user features. These psychological state parameters are reflected by time sensitivity, resistance threshold, decision preference, and time sensitivity. Finally, the user agent is constructed based on the user features and the sets of psychological state parameters. The second construction module is used to determine multiple interaction strategies by combining multiple interaction methods included in each interaction dimension through Cartesian product, wherein the interaction dimension includes at least financial scheme, speech style and reach time; and to construct the interactive intelligent agent according to the multiple interaction strategies.

[0087] Based on the above embodiments, the interaction module 310 is specifically used for: During the virtual interaction between the user agent and the interactive agent, the total attractiveness of the interaction strategy is determined; the total attractiveness of the interaction strategy and the psychological state parameter corresponding to the user agent are substituted into the acceptance probability formula to obtain the user's acceptance probability of the interaction strategy under the psychological state parameter.

[0088] In one implementation, determining the total attractiveness of the interaction strategy includes: Determine the financial appeal, verbal appeal, and temporal appeal of the interaction strategy; and determine the total appeal of the interaction strategy based on the financial appeal, verbal appeal, and temporal appeal.

[0089] Based on the above embodiments, module 320 is specifically used for: For each of the aforementioned interaction strategies, the counterfactual conversion rate of the interaction strategy is obtained by weighted summation and normalization of the user's acceptance probability for the interaction strategy under each of the aforementioned psychological state parameters.

[0090] Based on the above embodiments, the execution module 330 is specifically used for: Construct a probability surface based on the counterfactual transformations of each of the aforementioned interaction strategies; perform gradient calculations on the probability surface to determine the maximum counterfactual transformation rate; and determine the interaction strategy corresponding to the maximum counterfactual transformation rate as the target interaction strategy.

[0091] The counterfactual deduction-based strategy generation apparatus provided in this embodiment of the invention can execute the counterfactual deduction-based strategy generation method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects for executing the counterfactual deduction-based strategy generation method.

[0092] It is worth noting that in the above embodiments of the strategy generation device based on counterfactual deduction, the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy differentiation and are not used to limit the scope of protection of the present invention.

[0093] Figure 4 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. Figure 4 A block diagram of an exemplary computer device 4 suitable for implementing embodiments of the present invention is shown. Figure 4 The computer device 4 shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of the present invention.

[0094] like Figure 4 As shown, the computer device 4 is represented in the form of a general-purpose computing electronic device. The components of the computer device 4 may include, but are not limited to: one or more processors or processing units 16, system memory 28, and a bus 18 connecting different system components (including system memory 28 and processing unit 16).

[0095] Bus 18 represents one or more of several bus architectures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus architectures. For example, these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.

[0096] Computer device 4 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by computer device 4, including volatile and non-volatile media, removable and non-removable media.

[0097] System memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. Computer device 4 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 may be used to read and write non-removable, non-volatile magnetic media (…). Figure 4 Not shown; usually referred to as a "hard drive"). Although Figure 4 As not shown, disk drives for reading and writing to removable non-volatile disks (e.g., "floppy disks") and optical disc drives for reading and writing to removable non-volatile optical discs (e.g., CD-ROMs, DVD-ROMs, or other optical media) may be provided. In these cases, each drive may be connected to bus 18 via one or more data media interfaces. System memory 28 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of the present invention.

[0098] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in system memory 28. Such program modules 42 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. Program modules 42 typically perform the functions and / or methods described in the embodiments of the present invention.

[0099] Computer device 4 can also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, etc.), and with one or more devices that enable a user to interact with computer device 4, and / or with any device that enables computer device 4 to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed through input / output (I / O) interface 22. Furthermore, computer device 4 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) through network adapter 20. Figure 4 As shown, network adapter 20 communicates with other modules of computer device 4 via bus 18. It should be understood that, although... Figure 4 Not shown, it can be used in conjunction with computer device 4 with other hardware and / or software modules, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0100] Processing unit 16 executes various functional applications and page displays by running programs stored in system memory 28, such as implementing the counterfactual inference-based strategy generation method provided in this embodiment of the invention, which includes: During the virtual interaction between the user agent and the interactive agent, the user's acceptance probability for each interactive strategy under the psychological state parameters is determined based on the total attractiveness of each interactive strategy adopted by the interactive agent and the psychological state parameters corresponding to the user agent. The user agent is constructed based on the user's user characteristics and psychological state parameters, and the interactive agent is constructed based on multiple interactive strategies. The counterfactual conversion rate of each interaction strategy is determined based on the user's acceptance probability for each interaction strategy under each set of psychological state parameters. Determine the maximum counterfactual conversion rate among the counterfactual conversion rates of each of the interaction strategies, and determine the interaction strategy corresponding to the maximum counterfactual conversion rate as the target interaction strategy.

[0101] Of course, those skilled in the art will understand that the processor can also implement the technical solution of the strategy generation method based on counterfactual deduction provided in any embodiment of the present invention.

[0102] This invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements, for example, the counterfactual deduction-based strategy generation method provided in this invention, the method comprising: During the virtual interaction between the user agent and the interactive agent, the user's acceptance probability for each interactive strategy under the psychological state parameters is determined based on the total attractiveness of each interactive strategy adopted by the interactive agent and the psychological state parameters corresponding to the user agent. The user agent is constructed based on the user's user characteristics and psychological state parameters, and the interactive agent is constructed based on multiple interactive strategies. The counterfactual conversion rate of each interaction strategy is determined based on the user's acceptance probability for each interaction strategy under each set of psychological state parameters. Determine the maximum counterfactual conversion rate among the counterfactual conversion rates of each of the interaction strategies, and determine the interaction strategy corresponding to the maximum counterfactual conversion rate as the target interaction strategy.

[0103] The computer storage medium of this invention can be any combination of one or more computer-readable media. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0104] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0105] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0106] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0107] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computing device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0108] Furthermore, the acquisition, storage, use, and processing of data in the technical solution of this invention all comply with relevant laws and regulations.

[0109] Note that the above description is merely a preferred embodiment of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of the present invention, the scope of which is determined by the scope of the appended claims.

Claims

1. A strategy generation method based on counterfactual deduction, characterized in that, include: During the virtual interaction between the user agent and the interactive agent, the user's acceptance probability for each interactive strategy under the psychological state parameters is determined based on the total attractiveness of each interactive strategy adopted by the interactive agent and the psychological state parameters corresponding to the user agent. The user agent is constructed based on the user's user characteristics and psychological state parameters, and the interactive agent is constructed based on multiple interactive strategies. The counterfactual conversion rate of each interaction strategy is determined based on the user's acceptance probability for each interaction strategy under each set of psychological state parameters. Determine the maximum counterfactual conversion rate among the counterfactual conversion rates of each of the interaction strategies, and determine the interaction strategy corresponding to the maximum counterfactual conversion rate as the target interaction strategy.

2. The strategy generation method based on counterfactual deduction according to claim 1, characterized in that, The user agent construction process includes: The system acquires user behavior data in applications displaying promoted products, interaction record data in the user's system, and the user's historical transaction data. It then extracts features from the user behavior data, interaction record data, and historical transaction data to construct the user's user characteristics based on the extracted features. The user features are input into a pre-trained state extraction model, and multiple sets of psychological state parameters are extracted from the user features based on the state extraction model; or, multiple sets of historical user features of the user are obtained, and multiple sets of psychological state parameters are obtained by fusing each of the historical user features and the user features, wherein the psychological state parameters are reflected by time sensitivity, resistance threshold, decision preference and time sensitivity; The user intelligent agent is constructed based on the user characteristics and the psychological state parameters of each group.

3. The strategy generation method based on counterfactual deduction according to claim 1, characterized in that, The process of constructing the interactive intelligent agent includes: Multiple interaction strategies are determined by combining multiple interaction methods included in each interaction dimension through a Cartesian product. The interaction dimensions include at least financial solutions, communication style, and reach time. The interactive agent is constructed based on multiple interaction strategies.

4. The strategy generation method based on counterfactual deduction according to claim 1, characterized in that, The probability of a user accepting each interaction strategy under the given psychological state parameters is determined based on the total attractiveness of the interaction strategies employed by the interactive agent and the corresponding psychological state parameters of the user agent, including: Determine the total attractiveness of the interaction strategy; Substituting the total attractiveness of the interaction strategy and the psychological state parameters corresponding to the user agent into the acceptance probability formula, the user's acceptance probability of the interaction strategy under the psychological state parameters is obtained.

5. The strategy generation method based on counterfactual deduction according to claim 4, characterized in that, Determining the total attractiveness of the interaction strategy includes: Determine the financial appeal, persuasive appeal, and temporal appeal of the interaction strategy; The total appeal of the interaction strategy is determined based on the financial appeal, the verbal appeal, and the temporal appeal.

6. The strategy generation method based on counterfactual deduction according to claim 1, characterized in that, Based on the user's acceptance probability for each interaction strategy under each set of psychological state parameters, the counterfactual conversion rate of each interaction strategy is determined, including: For each of the aforementioned interaction strategies, the counterfactual conversion rate of the interaction strategy is obtained by weighted summation and normalization of the user's acceptance probability for the interaction strategy under each of the aforementioned psychological state parameters.

7. The strategy generation method based on counterfactual deduction according to claim 1, characterized in that, Determining the maximum counterfactual conversion rate among the counterfactual conversion rates of each of the aforementioned interaction strategies includes: Construct a probability surface based on the counterfactual transformations of each of the aforementioned interaction strategies; Gradient calculation is performed on the probability surface to determine the maximum counterfactual conversion rate.

8. A strategy generation device based on counterfactual deduction, characterized in that, include: An interaction module is used to determine the probability of a user accepting each interaction strategy under the psychological state parameters during the virtual interaction between the user agent and the interactive agent, based on the total attractiveness of each interaction strategy adopted by the interactive agent and the psychological state parameters corresponding to the user agent. The user agent is constructed based on the user's user characteristics and psychological state parameters, and the interactive agent is constructed based on multiple interaction strategies. The determining module is used to determine the counterfactual conversion rate of each interaction strategy based on the user's acceptance probability of each interaction strategy under each group of psychological state parameters; An execution module is used to determine the maximum counterfactual conversion rate among the counterfactual conversion rates of each of the interaction strategies, and to determine the interaction strategy corresponding to the maximum counterfactual conversion rate as the target interaction strategy.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the strategy generation method based on counterfactual deduction as described in any one of claims 1-7.

10. A storage medium containing computer-executable instructions, characterized in that, The computer-executable instructions, when executed by a computer processor, are used to perform the counterfactual inference-based strategy generation method as described in any one of claims 1-7.

Images