Product pushing method, electronic equipment, storage medium and program product

Abstract

The embodiment of the invention discloses a product pushing method, electronic equipment, a storage medium and a program product, and relates to the technical field of financial science and technology, and the method comprises the steps: obtaining behavior data of a user in a target application before the current moment from a distributed publishing and subscribing system through a distributed data flow engine, obtaining real-time behavior data, and transmitting the real-time behavior data to a server; obtaining offline behavior data of the user from the data warehouse to obtain historical behavior data; fusing the real-time behavior data and the historical behavior data to obtain user behavior data; determining a pushing rule satisfied by the user behavior data from a preset pushing rule base to obtain a plurality of candidate rules; obtaining a static feature and a dynamic feature of each candidate rule, and determining a priority score of the corresponding candidate rule based on the static feature and the dynamic feature; and the optimal rule is determined based on the priority score of each candidate rule, and the target product corresponding to the optimal rule is pushed to the user terminal of the user, so that the timeliness and accuracy of product pushing are improved.

Description

Technical Field

[0001] This application relates to the field of financial technology, and in particular to a product push method, electronic device, storage medium, and program product. Background Technology

[0002] With the rapid development of technology and the increasing demand for personalized services, private banking has gradually become one of the hottest areas in the financial industry.

[0003] Currently, the method of determining the financial needs of private banking users by relying on offline batch scheduling data processing and pushing corresponding products to the users in order to provide them with corresponding financial services has data delays and cannot capture users' real-time behavioral data. As a result, it cannot respond to users' dynamic financial needs, leading to low timeliness and accuracy of product push. Summary of the Invention

[0004] This application provides a product push method, electronic device, storage medium, and program product, which realizes the product push function to solve the problems of low timeliness and accuracy of product push in the prior art.

[0005] In a first aspect, embodiments of this application provide a product push method, which includes: using a distributed data stream engine to obtain user behavior data in a target application before the current moment from a distributed publish-subscribe system to obtain real-time behavior data, and obtaining user offline behavior data from a data warehouse to obtain historical behavior data; fusing real-time behavior data and historical behavior data to obtain user behavior data; determining push rules that the user behavior data satisfies from a preset push rule library to obtain multiple candidate rules; obtaining static and dynamic features of each candidate rule, and determining the priority score of the corresponding candidate rule based on the static and dynamic features; determining the optimal rule based on the priority score of each candidate rule, and pushing the target product corresponding to the optimal rule to the user's user terminal.

[0006] Secondly, embodiments of this application provide a product push device, comprising: an acquisition module, configured to acquire user behavior data in a target application up to the current moment from a distributed publish-subscribe system using a distributed data stream engine to obtain real-time behavior data, and acquire user offline behavior data from a data warehouse to obtain historical behavior data; a fusion module, configured to fuse real-time behavior data and historical behavior data to obtain user behavior data; a first determination module, configured to determine push rules satisfied by the user behavior data from a preset push rule library to obtain multiple candidate rules; a second determination module, configured to acquire static and dynamic features of each candidate rule, and determine the priority score of the corresponding candidate rule based on the static and dynamic features; and a push module, configured to determine the optimal rule based on the priority score of each candidate rule, and push the target product corresponding to the optimal rule to the user's user terminal.

[0007] Thirdly, embodiments of this application provide an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to execute the product push method of any embodiment of this application.

[0008] Fourthly, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the product push method as described in any embodiment of this application.

[0009] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the product delivery method as described in any embodiment of this application.

[0010] In this embodiment, a distributed data stream engine can be used to obtain user behavior data in the target application up to the current moment from a distributed publish-subscribe system, resulting in real-time behavior data. This captures real-time user behavior data without data latency issues, thereby improving the timeliness of product push notifications. Secondly, offline user behavior data is obtained from a data warehouse to obtain historical behavior data. By fusing real-time and historical behavior data, user behavior data is obtained. Historical behavior data can supplement missing data fields in real-time behavior data, thus comprehensively combining immediate and long-term user behavior data to create a precise user profile, providing an accurate data foundation for subsequently determining candidate rules. Finally, push rules that match the user behavior data are determined from a preset push rule library, resulting in... Multiple candidate rules are selected, and the static features of each candidate rule are acquired. Simultaneously, the dynamic features of each candidate rule are determined based on user behavior data. A priority score is then assigned to each candidate rule based on both static and dynamic features. Finally, the optimal rule is determined based on the priority score of each candidate rule. This approach comprehensively considers both static and dynamic features of candidate rules, adapting to dynamic changes in the business market, thereby improving the accuracy and efficiency of determining the optimal rule. The target product corresponding to the optimal rule is then pushed to the user's terminal, ensuring that the target product best meets the user's real-time financial needs and satisfies their dynamic financial requirements. This ensures highly personalized and precise product recommendations, improving the accuracy of product recommendations and ultimately enhancing the user experience. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of this application, 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 this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a flowchart illustrating a product push method provided in an embodiment of this application;

[0013] Figure 2 This is another flowchart illustrating the product push method provided in this application embodiment;

[0014] Figure 3 This is an example diagram of the warning page provided in an embodiment of this application;

[0015] Figure 4 This is a schematic diagram of the product push device provided in an embodiment of this application;

[0016] Figure 5This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this application.

[0018] It should be noted that the terms "first," "second," "target," and "original," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in sequences other than those illustrated or described herein. Furthermore, the terms "comprising," "having," and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0019] Figure 1 This is a flowchart illustrating a product push method provided in this application embodiment. This embodiment can be applied to scenarios where financial-related products need to be pushed to users. The product push method provided in this embodiment can be executed by a product push device provided in this application embodiment, which can be implemented through software and / or hardware. In a specific embodiment, the product push device can be integrated into an electronic device, such as a computer. The executing entity of this method can be an electronic device. See also... Figure 1 The product push method in this embodiment includes, but is not limited to, the following steps:

[0020] S110. Use the distributed data stream engine to obtain the user's behavior data in the target application before the current moment from the distributed publish-subscribe system to obtain real-time behavior data, and obtain the user's offline behavior data from the data warehouse to obtain historical behavior data.

[0021] The distributed data stream engine is a computing framework capable of real-time and efficient processing of continuously flowing data streams in a distributed cluster environment. The distributed publish-subscribe system is a distributed messaging system based on the "publish-subscribe" model, used for real-time data acquisition and asynchronous data transmission between different components or services. The target application is the source of the data, i.e., the specific application operated by the user's behavior. Behavioral data consists of records of various operations, interactions, or state changes generated by the user in the target application. For example, the data fields of behavioral data may include user identifier, transaction type, total asset amount, and time.

[0022] Real-time behavioral data refers to user behavior data that is collected and processed immediately after it occurs within the target application. A data warehouse stores batch-processed behavioral data, i.e., it stores offline behavioral data. Offline behavioral data refers to behavioral data that is not collected in real-time but is stored in the data warehouse after batch processing. Historical behavioral data refers to offline behavioral data stored in the data warehouse, i.e., all behavioral data generated by the user within a past period, such as yesterday's user behavior data within the target application.

[0023] Specifically, when a user browses within a target application, the distributed publish-subscribe system can capture user behavior that meets certain conditions in real time, such as browsing a high-net-worth product page for more than 30 seconds. It then collects behavioral data generated by these user actions, such as user identifiers, transaction types, total asset amounts, and timestamps. This behavioral data is then packaged according to a preset format and stored in the distributed publish-subscribe system. The preset format is a pre-defined data format, which can be a key-value pair data format.

[0024] Next, the distributed data stream engine can be used to retrieve user behavior data in the target application up to the current moment from the distributed publish-subscribe system, obtaining real-time behavior data. This real-time behavior data is the behavior data within a preset time period before the current moment, where the preset time period can be 1 millisecond, 1 second, or 1 minute, depending on the processing capacity of the distributed publish-subscribe system. Then, based on the user identifier, offline behavior data of the same user is retrieved from the data warehouse to obtain historical behavior data. In other words, the real-time behavior data and historical behavior data belong to the same user.

[0025] S120: By integrating real-time and historical behavioral data, user behavior data is obtained.

[0026] Among them, user behavior data is the data obtained by fusing real-time behavior data and historical behavior data, which can represent a hybrid user profile that takes into account both immediacy and long-term perspectives.

[0027] Specifically, after obtaining real-time and historical behavior data of the same user, the real-time and historical behavior data can be merged. That is, the behavior data in the historical behavior data that belongs to the same data field as the real-time behavior data can be removed to obtain the user behavior data.

[0028] For example, if real-time behavior data is {"User ID": 001, "Transaction Type": "Type 1", "Time": "Time 1"}, and historical behavior data is {"User ID": 001, "Transaction Type": "Type 1", "Total Asset Amount": "XXXXX", "Time": "Time 2"}, then user behavior data would be {"User ID": 001, "Transaction Type": "Type 1", "Total Asset Amount": "XXXXX", "Time": "Time 1"}.

[0029] S130. Determine the push rules that the user behavior data meets from the preset push rule base, and obtain multiple candidate rules.

[0030] The preset push rule library is a pre-configured set of push rules, including multiple push rules. Push rules are the logical criteria for determining whether to trigger a push notification for a specific product. They are pre-set by business personnel based on actual scenarios, and one push rule corresponds to one product; for example, the product could be a financial product. Candidate rules are push rules that meet the criteria set by user behavior data.

[0031] Specifically, after obtaining user behavior data, it can be determined whether the user behavior data meets the push rules in the preset push rule library. That is, when the user behavior data matches the push triggering condition in the push rule, the user behavior data meets the push rule. Then, the push rules that the user behavior data meets are determined as candidate rules. At this time, there are multiple candidate rules.

[0032] S140. Obtain the static and dynamic features of each candidate rule, and determine the priority score of the corresponding candidate rule based on the static and dynamic features.

[0033] Static features are the characteristics of relatively fixed attributes or conditions in push rules that do not change with changes in real-time user behavior; that is, the fixed attribute features of push rules. Optionally, static features may include static rule weights; static rule weights are relatively fixed weight values ​​assigned to each push rule in advance when setting push rules, which do not change with changes in real-time user behavior and external dynamics, and are used to measure the relative importance of push rules in the entire push rule system.

[0034] Dynamic features are attributes or conditions in push notification rules that change with variations in real-time user behavior, external environment, and other factors; these are the dynamic attribute features of push notification rules. Optionally, dynamic features may include real-time user value scores and behavior trigger intensity. Real-time user value scores quantify the potential value of user behavior data for the product corresponding to the push notification rule; behavior trigger intensity quantifies the urgency or criticality of a specific user behavior when triggering a push notification rule, reflecting the "strength" of that behavior as a triggering condition.

[0035] Priority scoring is a quantitative indicator used to measure the relative importance of each candidate rule. In other words, the higher the priority score, the more the corresponding candidate rule's product meets the user's real-time financial needs.

[0036] Specifically, after obtaining multiple candidate rules, for the current candidate rule among the multiple candidate rules, the static and dynamic features of the current candidate rule can be obtained. That is, the static weights of the rules set in advance for the current candidate rule can be obtained to obtain the static features. Secondly, the real-time value score of the user is determined based on user behavior data and preset scoring standards. For example, the asset score is determined based on the total asset amount in the user behavior data and the asset of the product corresponding to the current candidate rule, as well as the preset asset scoring standard. The transaction score is determined based on the transaction frequency of the product corresponding to the current candidate rule in the user behavior data and the preset transaction scoring standard. The asset score and the transaction score are merged to obtain the real-time value score of the user. The fusion strategy can be to calculate the average value, or to perform weighted fusion based on the weights preset for the asset score and the transaction score respectively. Next, the behavior trigger intensity is determined based on user behavior data and preset intensity standards. For example, the behavior trigger intensity is determined based on the click rate of the product corresponding to the current candidate rule in the user behavior data and the preset click intensity standard, or the behavior trigger intensity is determined based on the number of visits to the product corresponding to the current candidate rule in the user behavior data and the preset visit intensity standard. Among them, the preset asset scoring standard is used to quantify the asset score corresponding to the asset, and the asset score is used to measure the potential value of the user's assets to the product; the preset transaction scoring standard is used to quantify the transaction score corresponding to the transaction frequency, and the transaction score is used to measure the potential value of the user's transaction frequency to the product; the preset click intensity standard is used to quantify the behavior trigger intensity corresponding to the click rate; and the preset access intensity standard is used to quantify the behavior trigger intensity corresponding to the access volume.

[0037] Then, the priority score of the current candidate rule can be determined based on static and dynamic features. For example, the average of the rule's static weight, the user's real-time value score, and the behavior trigger intensity can be calculated to obtain the priority score corresponding to the current candidate rule.

[0038] S150. Determine the optimal rule based on the priority score of each candidate rule, and push the target product corresponding to the optimal rule to the user's terminal.

[0039] The optimal rule is the push notification rule that best matches the user's financial needs among multiple candidate rules. The target product is the product corresponding to the optimal rule.

[0040] Specifically, after obtaining the priority scores of multiple candidate rules, the optimal rule can be determined based on the priority score of each candidate rule. For example, the candidate rule corresponding to the maximum priority score can be determined as the optimal rule. Then, the product corresponding to the optimal rule is determined, the target product is obtained, and the target product is pushed to the user's terminal. That is, the target product can be pushed to the user's terminal through push methods such as SMS, target application push, or business personnel pop-up.

[0041] The technical solution of this application embodiment can utilize a distributed data stream engine to obtain user behavior data in the target application up to the current moment from a distributed publish-subscribe system, thus obtaining real-time behavior data. This captures real-time user behavior data without data latency issues, thereby improving the timeliness of product push notifications. Secondly, it obtains offline user behavior data from a data warehouse to obtain historical behavior data. By fusing real-time and historical behavior data, it obtains user behavior data. Historical behavior data can supplement missing data fields in real-time behavior data, thus comprehensively combining immediate and long-term user behavior data to create a precise user profile, providing an accurate data foundation for subsequent candidate rule determination. Finally, it determines the push rules that the user behavior data satisfies from a preset push rule library. Multiple candidate rules are obtained, and the static features of each candidate rule are acquired. Simultaneously, the dynamic features of each candidate rule are determined based on user behavior data. A priority score is then determined for each candidate rule based on both its static and dynamic features. Finally, the optimal rule is determined based on the priority score of each candidate rule. This approach comprehensively considers both the static and dynamic features of candidate rules, adapting to dynamic changes in the business market, thereby improving the accuracy and efficiency of determining the optimal rule. The target product corresponding to the optimal rule is then pushed to the user's terminal, ensuring that the target product best meets the user's real-time financial needs and satisfies their dynamic financial requirements. This ensures highly personalized and precise product recommendations, improving the accuracy of product recommendations and ultimately enhancing the user experience.

[0042] The following further describes a product push method provided by an embodiment of this application. Figure 2 This is another flowchart illustrating the product push method provided in this application. This application's embodiments are optimizations based on the above embodiments. See also... Figure 2The method in this embodiment includes, but is not limited to, the following steps:

[0043] S210. Use the distributed data stream engine to obtain the user's behavior data in the target application before the current moment from the distributed publish-subscribe system to obtain real-time behavior data, and obtain the user's offline behavior data from the data warehouse to obtain historical behavior data.

[0044] S220: By integrating real-time and historical behavioral data, user behavior data is obtained.

[0045] S230. Determine the push rules that the user behavior data meets from the preset push rule library, and obtain multiple candidate rules.

[0046] Specifically, push rules that match user behavior data are determined from a pre-defined push rule base, resulting in multiple candidate rules, including Sa1-Sa4:

[0047] Sa1: Determine the push rules that the user behavior data meets from the preset push rule base, and obtain multiple first intermediate rules.

[0048] The first intermediate rule is the push rule that the user behavior data satisfies.

[0049] Specifically, the push rules that user behavior data satisfies can be determined as the first intermediate rules, and there are multiple first intermediate rules.

[0050] Sa2: Remove the first intermediate rules that do not meet user compliance requirements from the multiple first intermediate rules to obtain multiple second intermediate rules.

[0051] The second intermediate rule is the rule that satisfies user compliance among multiple first intermediate rules.

[0052] Specifically, a pre-defined non-compliance relationship table can be queried based on the user identifier to obtain the set of non-compliance rules corresponding to the user identifier. The pre-defined non-compliance relationship table is used to store the correspondence between the user identifier and the set of non-compliance rules corresponding to the user identifier. For example, the set of non-compliance rules corresponding to risk-sensitive customers includes push rules corresponding to derivative products, etc. Then, the first intermediate rules and the set of non-compliance rules are compared, and the first intermediate rules that are within the set of non-compliance rules are removed from multiple first intermediate rules to obtain multiple second intermediate rules.

[0053] Sa3. Determine similar rule pairs and dissimilar rules from multiple second intermediate rules.

[0054] Among them, the similarity rule pair includes two similar second intermediate rules.

[0055] Specifically, a similarity calculation method can be used to calculate the semantic similarity between multiple second intermediate rules. If the similarity between two second intermediate rules is greater than or equal to a preset similarity threshold, then these two second intermediate rules are identified as a similar rule pair. If the similarity between a second intermediate rule and all other second intermediate rules is less than the preset similarity threshold, then the second intermediate rule is identified as a dissimilar rule. This process is used to determine similar rule pairs and dissimilar rules. The preset similarity threshold is a pre-set similarity value used to characterize the similarity between two push rules.

[0056] Sa4. The similar rule corresponding to the maximum value of the static weight of the dissimilar rule and the similar rule is determined as the candidate rule.

[0057] In this embodiment, firstly, the first intermediate rules that do not meet user compliance are removed from multiple first intermediate rules. This removes the recommendation rules corresponding to products that users do not like, thereby reducing the computational load and improving the accuracy of product recommendations. Then, among multiple second intermediate rules, the similar rules corresponding to the maximum static weight of dissimilar rules and similar rules are determined as candidate rules. This removes rules with smaller static weights in similar rules and retains only the rule with the largest static weight in similar rules, thereby reducing the computational load and saving computational resources.

[0058] S240. Obtain the static and dynamic features of each candidate rule.

[0059] Among them, static features also include rule types; rule types are categories formed by pre-classifying push rules according to different classification standards. For example, rule types may include compliance, benefit, and experience categories.

[0060] S250. Using a conversion rate prediction model, predict the conversion rate of each candidate rule based on user behavior data to obtain the expected conversion rate of the corresponding candidate rule.

[0061] The conversion rate prediction model is a pre-trained model used to predict the conversion rate of candidate rules in the current scenario (i.e., user behavior data). The conversion rate measures the impact of pushing a product to a user on the user's click behavior, reflecting the increased user attention brought about by the push behavior. The expected conversion rate is the conversion rate predicted by the conversion rate prediction model based on user behavior data.

[0062] Specifically, for the current candidate rule among multiple candidate rules, the behavior data related to the product corresponding to the current candidate rule can be obtained from user behavior data, and the behavior data related to the product corresponding to the current candidate rule can be input into the conversion rate prediction model. At this time, the conversion rate prediction model uses the pre-learned model parameters to analyze the behavior data related to the product corresponding to the current candidate rule, obtains the output result, and determines the expected conversion rate of the current candidate rule.

[0063] S260. Determine the priority score of the corresponding candidate rules based on the expected conversion rate, static features, and dynamic features.

[0064] Specifically, in one implementation, for the current candidate rule among multiple candidate rules, the average of the expected conversion rate, static weight of the rule, real-time value score of the user, and the intensity of the behavior trigger can be calculated to obtain the priority score of the current candidate rule.

[0065] Optionally, in another implementation, the first weight corresponding to the static weight of the rule, the second weight corresponding to the real-time value score of the user, the third weight corresponding to the behavior trigger intensity, and the fourth weight corresponding to the expected conversion rate are determined based on the rule type. That is, different weight allocation schemes can be set in advance for each rule type. In this case, for the current candidate rule among multiple candidate rules, the weight allocation scheme set in advance for the rule type can be obtained based on the rule type of the current candidate rule, resulting in the first weight corresponding to the static weight, the second weight corresponding to the real-time value score of the user, the third weight corresponding to the behavior trigger intensity, and the fourth weight corresponding to the expected conversion rate. Then, based on the first weight, the second weight, the third weight, and the fourth weight, the static weight of the rule, the real-time value score of the user, the behavior trigger intensity, and the expected conversion rate are weighted and fused to obtain the priority score of the corresponding candidate rule. That is, based on the determined first weight, the second weight, the third weight, and the fourth weight, the weighted sum of the static weight of the rule, the real-time value score of the user, the behavior trigger intensity, and the expected conversion rate of the current candidate rule can be calculated to obtain the priority score of the current candidate rule. It can improve computational efficiency, reduce implementation complexity, and determine the key features to focus on based on actual scenarios and needs, and assign greater weights to these key features, thereby improving the accuracy and efficiency of priority scoring and providing an accurate data foundation for determining the optimal rules in the future.

[0066] The first weight is used to characterize the importance of the static weight of the rule when determining the priority score; the second weight is used to characterize the importance of the real-time value score of the user when determining the priority score; the third weight is used to characterize the importance of the intensity of the behavior trigger when determining the priority score; and the fourth weight is used to characterize the importance of the expected conversion rate when determining the priority score.

[0067] S270. Identify the rule types of multiple candidate rules, obtain multiple candidate rule types, and determine the optimal rule type from the multiple candidate rule types based on the preset type priority.

[0068] Among them, the candidate rule type is the rule type corresponding to multiple candidate rules; the preset type priority is the priority set in advance for different rule types, which can be adjusted and set according to the actual scenario and actual needs. For example, the priority of compliance type is greater than the priority of benefit type, and the priority of benefit type is greater than the priority of experience type.

[0069] Specifically, the rule type of each candidate rule can be identified, and the rule type of each candidate rule can be determined as a candidate rule type, thereby obtaining multiple candidate rule types; then, the preset type priority of the multiple candidate rule types is compared, and the candidate rule type corresponding to the maximum preset type priority is determined as the optimal rule type.

[0070] S280. Obtain the candidate rule corresponding to the maximum priority score from the candidate rules belonging to the optimal rule type, and obtain the optimal rule.

[0071] Specifically, candidate rules belonging to the optimal rule type can be obtained from multiple candidate rules, and the candidate rule corresponding to the maximum priority score among the candidate rules belonging to the optimal rule type is determined as the optimal rule.

[0072] Furthermore, the candidate rule corresponding to the maximum priority score from the candidate rules belonging to the optimal rule type is obtained as the third intermediate rule. There may be one or more third intermediate rules. If there is one third intermediate rule, and the difference in priority score between the third intermediate rule and the remaining candidate rules belonging to the optimal rule type is greater than or equal to a preset difference threshold, it indicates that the difference in priority score between the third intermediate rule and the remaining candidate rules belonging to the optimal rule type is significant. In this case, the third intermediate rule can be directly determined as the optimal rule. The preset difference threshold is a pre-set value used to characterize whether the difference between priority scores is small.

[0073] If there are at least two third intermediate rules, or if the priority score difference between the third intermediate rule and the remaining candidate rules belonging to the optimal rule type is less than a preset difference threshold, it indicates that the priority scores of the multiple candidate rules belonging to the optimal rule type are relatively similar and the differences are small. In this case, an optimal rule determination request can be generated based on the candidate rules belonging to the optimal rule type. That is, the optimal rule determination request can include multiple candidate rules belonging to the optimal rule type, and the optimal rule determination request can also include user behavior data. Then, the optimal rule determination request is displayed on the screen of the electronic device. That is, the multiple candidate rules and user behavior data included in the optimal rule determination request can be displayed, and each candidate rule corresponds to a selection control. Then, the business personnel can determine the candidate rule that best matches the displayed user behavior data, which is recorded as the optimal candidate rule, and click the selection control corresponding to the optimal candidate rule. After that, after the electronic device detects the trigger operation of the business personnel on the selection control, it can obtain the selection information of the business personnel on the optimal rule determination request and determine the candidate rule corresponding to the selection information as the optimal rule. The optimal rule determination request is used to solicit opinions from business personnel to clarify the final optimal rule; the selection information is the feedback content made by users in response to the optimal rule determination request. When there are at least two third intermediate rules, or when the priority score difference between the third intermediate rule and the remaining candidate rules belonging to the optimal rule type is less than a preset difference threshold, the final optimal rule can be determined by manual selection. This can improve the accuracy of the optimal rule determination and thus provide accurate data basis for subsequent push of target products.

[0074] Optionally, if there are at least two third intermediate rules, a second optimal rule determination request can be generated based on at least two third intermediate rules, and the second optimal rule determination request can be displayed. Then, the second selection information of the business personnel in response to the second optimal rule determination request can be obtained, and the candidate rule corresponding to the second selection information can be determined as the optimal rule. Only at least two third intermediate rules need to be displayed, which reduces the difficulty of manual judgment and the amount of data, thereby improving the efficiency of determining the optimal rule.

[0075] S290: Push the target product corresponding to the optimal rule to the user's user terminal.

[0076] Optionally, user behavior data may include the user's first click-through rate (CTR) for the target product, i.e., the CTR before the target product is pushed to the user; after the target product corresponding to the optimal rule is pushed to the user's device, it may also include Sb1-Sb5:

[0077] Sb1. Obtain user behavior data related to the target product after it is pushed to the user, and obtain target behavior data.

[0078] The target behavior data refers to user behavior data related to the target product after it is pushed to the user. The target behavior data includes real-time behavior data related to the target product after it is pushed to the user and historical behavior data related to the target product before it is pushed to the user. The target behavior data may also include the user's second click-through rate for the target product. The second click-through rate is the user's click-through rate for the target product after it is pushed to the user.

[0079] Specifically, user behavior data after the target product is pushed to the user can be obtained. The steps for obtaining the data are the same as those in S210 and S220, and can be referred to the above description, which will not be repeated here. Then, user behavior data related to the target product can be obtained from the user behavior data after the target product is pushed to the user to obtain the target behavior data.

[0080] Sb2. Determine the actual impact of the optimal rule based on the first click-through rate and the second click-through rate, and determine the learning rate based on the online duration of the optimal rule.

[0081] Among them, the actual impact is used to quantify the actual influence on user behavior after pushing the target product to the user, and is used to characterize the actual influence of the optimal rule on user behavior. The online duration is the time from when the push rule is first launched to the current moment. The learning rate is used to measure the extent to which the static weights of the push rule are adjusted and optimized as the push rule's online time progresses.

[0082] Specifically, the difference between the second click-through rate (CTR) and the first CTR can be calculated, and the ratio of this difference to the first CTR can be calculated to obtain the actual impact of the optimal rule. Then, the online duration of the optimal rule is obtained, and the preset duration learning rate relationship is queried based on the online duration of the optimal rule to obtain the learning rate corresponding to the online duration. The preset duration learning rate relationship is a pre-set mapping relationship, which includes the relationship between the online duration and the learning rate corresponding to the online duration. The shorter the online duration, the larger the corresponding learning rate. This allows for dynamic adjustment of the static weight of the corresponding rule according to the life cycle of the push rule, so that the adjustment range of the static weight of the push rule with a shorter online duration is larger, thereby improving the accuracy of the static weight of the rule.

[0083] Sb3. Using the rule-based influence prediction model, the influence of the optimal rule is predicted based on the target behavior data to obtain the expected influence degree.

[0084] Among them, the rule impact prediction model is a pre-trained model used to predict the expected impact on user behavior after pushing target products to users, and is used to characterize the expected impact of the optimal rule on user behavior.

[0085] Specifically, target behavior data can be input into a pre-trained rule-based influence prediction model. The rule-based influence prediction model can then analyze the target behavior data based on the pre-learned model parameters, obtain the output result, and determine the expected influence degree.

[0086] Sb4. Calculate the difference between the actual impact and the expected impact, calculate the product of the difference and the learning rate, and calculate the sum of the product and the static weights of the optimal rule to obtain the static weights of the new rule.

[0087] Among them, the new rule static weights are the new rule static weights obtained by adjusting the rule static weights based on the actual impact, expected impact, and learning rate.

[0088] Sb5. Update the static weights of the optimal rule to the static weights of the new rule.

[0089] In this embodiment, the static weight of the optimal rule can be dynamically updated based on the feedback results after pushing the target product to the user. This allows the static weight of the rule to quickly respond to changes in the business market, thereby achieving automation and optimization of accurate push notifications. It can also improve computational efficiency, reduce implementation complexity, and thus improve the accuracy of the static weight of the rule.

[0090] Optionally, real-time behavioral data obtained from the distributed publish-subscribe system using a distributed data stream engine can be stored in a database, and then the behavioral data in the database and data warehouse can be displayed in real time to help business personnel optimize push rules.

[0091] Specifically, datasets can be created on the dataset creation page. This means connecting to a database based on its configuration information (such as database address and password) to associate the database and its tables, and connecting to a data warehouse based on its configuration information to associate the data warehouse and its tables. Next, users can write Structured Query Language (SCL) statements according to their private banking needs, copy these SCL statements to the SCL statement area on the dataset creation page, click the "Refresh Data" control, and then click the "Save" control to complete dataset creation. The retrieved data is then displayed in the "Data Details" area on the dataset creation page for business personnel to view.

[0092] Then, you can create and display the report on the report creation page. Specifically, you can click the "Create Report" control on the report creation page to create a new report, then click the icon components on the report creation page and select the desired display components by dragging and dropping, then click the "Dataset" control on the report creation page, select the dataset to be bound, and complete the report creation. The successfully created report will be displayed in the "Report Details" area on the report creation page. At this time, the report includes key indicator data such as click-through rate, asset changes, and asset distribution for business personnel to view.

[0093] Afterwards, business personnel can view the key indicator data in the report to determine whether the push rules need to be optimized. If the push rules need to be optimized, they can update the corresponding push rules on the push rule settings page.

[0094] Optionally, alert rules can be set on the alert page. This means that information related to the alert rules, such as alert name, alert details, alert conditions, alert execution cycle, and alert method, can be set on the alert page. When user behavior data meets the alert rules, an alert will be sent to the user based on the corresponding alert method. For example, such as... Figure 3 The image shown is an example of an alert page provided in this application embodiment. Users can enter the alert name in the "Name" input box, the alert details in the "Details" input box, the alert conditions in the "Alert Conditions" input box, and the alert execution period in the "Alert Execution Period" selection control. Then, users can select the alert method according to actual needs, i.e., select the checkboxes for "Simultaneously push to email" and / or "Push to SMS". Finally, click the "OK" control to create an alert rule.

[0095] For example, business personnel can pre-set push rules related to financial products, then use S210 and S220 to obtain user behavior data in private banking related applications, and use S230 to S280 to determine the optimal rule that best matches the user behavior data, and then push the financial product corresponding to the optimal rule to the user's terminal, thereby achieving precise push.

[0096] The technical solution of this application embodiment can utilize a distributed data stream engine to obtain user behavior data in the target application before the current moment from a distributed publish-subscribe system, obtaining real-time behavior data, and obtain offline user behavior data from a data warehouse, obtaining historical behavior data. Then, it merges the real-time behavior data and historical behavior data to obtain user behavior data, and determines the push rules satisfied by the user behavior data from a preset push rule library, obtaining multiple candidate rules. Next, it obtains the static and dynamic features of each candidate rule, and uses a conversion rate prediction model to predict the conversion rate of each candidate rule based on the user behavior data, obtaining the expected conversion rate of the corresponding candidate rule. This can improve the accuracy of the expected conversion rate prediction. Furthermore, it determines the priority score of the corresponding candidate rule based on the expected conversion rate, static features, and dynamic features. This comprehensively considers the expected conversion rate, static features, and dynamic features of the candidate rules, quickly adapting to changes in the business market. Improving computational efficiency and reducing implementation complexity enhances the accuracy and efficiency of priority scoring. Next, the rule types of multiple candidate rules are identified, resulting in multiple candidate rule types. Based on preset type priorities, the optimal rule type is determined from these candidate rule types. Then, the candidate rule with the highest priority score among the candidate rules belonging to the optimal rule type is obtained, thus yielding the optimal rule. Preset type priorities based on actual business scenarios can be considered to make the optimal rule selection strategy more adaptable to real-world business situations, improving computational efficiency and reducing implementation complexity, thereby enhancing the efficiency and accuracy of optimal rule determination. Finally, pushing the target product corresponding to the optimal rule to the user's terminal ensures that the target product best meets the user's real-time financial needs, satisfying their dynamic financial requirements. This ensures highly personalized and precise product recommendations, improving the accuracy of product recommendations and ultimately enhancing the user experience.

[0097] Figure 4 This is a schematic diagram of a product push device provided in an embodiment of this application, referring to... Figure 4 The product delivery device may include:

[0098] The acquisition module 410 is used to obtain the user's behavior data in the target application before the current moment from the distributed publish-subscribe system using the distributed data stream engine to obtain real-time behavior data, and to obtain the user's offline behavior data from the data warehouse to obtain historical behavior data.

[0099] The fusion module 420 is used to fuse real-time behavior data and historical behavior data to obtain user behavior data.

[0100] The first determining module 430 is used to determine the push rules that the user behavior data meets from the preset push rule library, and obtain multiple candidate rules;

[0101] The second determining module 440 is used to obtain the static and dynamic features of each candidate rule, and determine the priority score of the corresponding candidate rule based on the static and dynamic features.

[0102] The push module 450 is used to determine the optimal rule based on the priority score of each candidate rule, and push the target product corresponding to the optimal rule to the user's terminal.

[0103] In one embodiment, the second determining module 440 determines the priority score of the corresponding candidate rule based on static features and dynamic features, including: using a conversion rate prediction model to predict the conversion rate of each candidate rule under user behavior data based on user behavior data, and obtaining the expected conversion rate of the corresponding candidate rule; and determining the priority score of the corresponding candidate rule based on the expected conversion rate, static features and dynamic features.

[0104] In one embodiment, the static features include rule static weights and rule types, and the dynamic features include user real-time value scores and behavior trigger intensity. The second determining module 440 determines the priority score of the corresponding candidate rule based on the expected conversion rate, static features, and dynamic features, including: determining a first weight corresponding to the rule static weight, a second weight corresponding to the user real-time value score, a third weight corresponding to the behavior trigger intensity, and a fourth weight corresponding to the expected conversion rate based on the rule type; and performing a weighted fusion of the rule static weight, user real-time value score, behavior trigger intensity, and expected conversion rate based on the first weight, second weight, third weight, and fourth weight to obtain the priority score of the corresponding candidate rule.

[0105] In one embodiment, the first determining module 430 is specifically used to: determine the push rules that the user behavior data satisfies from the preset push rule library, and obtain multiple first intermediate rules; remove the first intermediate rules that do not meet the user compliance from the multiple first intermediate rules, and obtain multiple second intermediate rules; determine similar rule pairs and dissimilar rules from the multiple second intermediate rules; and determine the similar rules corresponding to the maximum value of the static weight of the dissimilar rules and the similar rule pairs as candidate rules.

[0106] In one embodiment, the push module 450 determines the optimal rule based on the priority score of each candidate rule, including: identifying the rule types of multiple candidate rules, obtaining multiple candidate rule types, and determining the optimal rule type from the multiple candidate rule types based on a preset type priority; obtaining the candidate rule corresponding to the maximum priority score from the candidate rules belonging to the optimal rule type, and obtaining the optimal rule.

[0107] In one embodiment, the push module 450 obtains the candidate rule corresponding to the maximum priority score from the candidate rules belonging to the optimal rule type to obtain the optimal rule, including: obtaining the candidate rule corresponding to the maximum priority score from the candidate rules belonging to the optimal rule type to obtain the third intermediate rule; if there are at least two third intermediate rules, or if the difference in priority score between the third intermediate rule and the remaining candidate rules belonging to the optimal rule type is less than a preset difference threshold, then an optimal rule determination request is generated based on the candidate rules belonging to the optimal rule type, and the optimal rule determination request is displayed; the selection information of the business personnel for the optimal rule determination request is obtained, and the candidate rule corresponding to the selection information is determined as the optimal rule.

[0108] In one embodiment, the user behavior data includes the user's first click-through rate (CTR) for the target product. The product push device further includes an update module, which is specifically used for: after pushing the target product corresponding to the optimal rule to the user's terminal, acquiring user behavior data related to the target product after pushing the target product to the user, obtaining target behavior data, which includes the user's second CTR for the target product; determining the actual impact of the optimal rule based on the first CTR and the second CTR, and determining the learning rate based on the online duration of the optimal rule; using a rule impact prediction model, predicting the rule impact of the optimal rule based on the target behavior data, and obtaining the expected impact; calculating the difference between the actual impact and the expected impact, calculating the product of the difference and the learning rate, and calculating the sum of the product and the static weight of the optimal rule, obtaining the new static weight of the rule; and updating the static weight of the optimal rule to the new static weight of the rule.

[0109] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional modules is merely an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working process of the functional modules described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0110] The product push device provided in this embodiment can be applied to the product push method provided in any of the above embodiments, and has corresponding functions and beneficial effects.

[0111] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 5 A block diagram is shown of an exemplary electronic device 11 suitable for implementing embodiments of the present application. Figure 5 The electronic device 11 shown is merely an example and should not impose any limitations on the functionality and scope of use of this embodiment.

[0112] like Figure 5 As shown, the electronic device 11 is represented in the form of a general-purpose computing electronic device. The components of the electronic device 11 may include, but are not limited to: one or more processors or processing units 16, system memory 28, and bus 18 connecting different system components (including system memory 28 and processing unit 16).

[0113] 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. Examples of these architectures include, but are not limited to, industry-standard architecture buses, microchannel architecture buses, enhanced industry-standard architecture buses, Video Electronics Standards Association (VESA) local buses, and peripheral component interconnect buses.

[0114] Electronic device 11 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by electronic device 11, including volatile and non-volatile media, removable and non-removable media.

[0115] System memory 28 may include computer system readable media in the form of volatile memory, such as random access memory 30 and / or cache memory 32. Electronic device 11 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 5 Not shown; usually referred to as a "hard drive"). Although Figure 5 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., portable compact disk read-only memory, digital multifunction optical disc read-only memory, 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 this application.

[0116] 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 this application.

[0117] Electronic device 11 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 electronic device 11, and / or with any device that enables electronic device 11 to communicate with one or more other computing devices (e.g., network card and modem, etc.). Such communication can be performed through input / output interface 22. Furthermore, electronic device 11 can also communicate with one or more networks (e.g., local area network, wide area network, and / or public network) through network adapter 20.

[0118] like Figure 5 As shown, network adapter 20 communicates with other modules of electronic device 11 via bus 18. It should be understood that, although... Figure 5 As not shown, other hardware and / or software modules may be used in conjunction with electronic device 11, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, independent disk redundant array systems, tape drives, and data backup storage systems.

[0119] The processing unit 16 executes various functional applications and page displays by running programs stored in the system memory 28, such as implementing a product push method provided in any embodiment of this application.

[0120] This application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements a product push method, such as that provided in any embodiment of this application.

[0121] The computer storage medium of this embodiment can be any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. For example, a computer-readable storage medium can be, but is 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, a random access memory, a read-only memory, an erasable programmable read-only memory, flash memory, an optical fiber, a portable compact disk read-only memory, an optical storage device, a 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.

[0122] 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.

[0123] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, radio frequency, etc., or any suitable combination thereof.

[0124] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages ​​as well as conventional procedural 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).

[0125] This application also provides a computer program product, including a computer program that, when executed by a processor, implements a product push method as provided in any embodiment of this application.

[0126] In the implementation of a computer program product, computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages ​​as well as conventional procedural 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 local area networks (LANs) or wide area networks (WANs), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0127] Those skilled in the art will understand that the modules or steps described above in this application 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, which can then be stored in a storage device for execution by a computing device. Alternatively, they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, this application is not limited to any particular combination of hardware and software.

[0128] It should be noted that in the technical solutions of this application embodiment, the information collected is information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of related data all comply with the relevant laws, regulations and standards of relevant countries and regions, necessary confidentiality measures have been taken, and they do not violate public order and good morals. Corresponding operation entry points are provided for users to choose to authorize or refuse. In addition, corresponding operation entry points are provided for users to choose to agree to or refuse the automated decision results. If the user chooses to refuse, the process enters the expert decision-making process.

[0129] Note that the above are merely preferred embodiments and the technical principles employed in this application. Those skilled in the art will understand that this application 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 this application. Therefore, although this application has been described in detail through the above embodiments, this application is not limited to the above embodiments. Many other equivalent embodiments may be included without departing from the inventive concept of this application, and the scope of this application is determined by the scope of the appended claims.

Claims

1. A product push method, characterized in that, The method includes: The distributed data stream engine is used to obtain real-time behavior data of the user in the target application before the current moment from the distributed publish-subscribe system, and the offline behavior data of the user is obtained from the data warehouse to obtain historical behavior data. By integrating the real-time behavior data and the historical behavior data, user behavior data is obtained; The user behavior data is determined to satisfy the push rules from a preset push rule base, resulting in multiple candidate rules; Obtain the static and dynamic features of each candidate rule, and determine the priority score of the corresponding candidate rule based on the static and dynamic features; The optimal rule is determined based on the priority score of each candidate rule, and the target product corresponding to the optimal rule is pushed to the user's terminal.

2. The product push method according to claim 1, characterized in that, The step of determining the priority score of the corresponding candidate rule based on the static features and the dynamic features includes: Using a conversion rate prediction model, the conversion rate of each candidate rule is predicted based on the user behavior data to obtain the expected conversion rate of the corresponding candidate rule. The priority score of the corresponding candidate rule is determined based on the expected conversion rate, the static features, and the dynamic features.

3. The product push method according to claim 2, characterized in that, The static features include rule static weights and rule types, and the dynamic features include user real-time value scores and behavior trigger strength. Determining the priority score of the corresponding candidate rule based on the expected conversion rate, the static features, and the dynamic features includes: Based on the rule type, determine the first weight corresponding to the static weight of the rule, the second weight corresponding to the real-time value score of the user, the third weight corresponding to the intensity of the behavior triggering, and the fourth weight corresponding to the expected conversion rate; Based on the first weight, the second weight, the third weight, and the fourth weight, the static weight of the rule, the real-time value score of the user, the intensity of the behavior trigger, and the expected conversion rate are weighted and fused to obtain the priority score of the corresponding candidate rule.

4. The product push method according to claim 1, characterized in that, The step of determining the push rules that the user behavior data satisfies from the preset push rule base yields multiple candidate rules, including: The user behavior data is determined to satisfy the push rules from the preset push rule base, and multiple first intermediate rules are obtained. Remove the first intermediate rules that do not meet user compliance requirements from the multiple first intermediate rules to obtain multiple second intermediate rules; From the plurality of second intermediate rules, determine similar rule pairs and dissimilar rules; The similar rule corresponding to the maximum value of the static weight of the dissimilar rule and the similar rule is determined as the candidate rule.

5. The product push method according to claim 1, characterized in that, The process of determining the optimal rule based on the priority score of each candidate rule includes: The rule types of the multiple candidate rules are identified to obtain multiple candidate rule types, and the optimal rule type is determined from the multiple candidate rule types based on a preset type priority. The optimal rule is obtained by selecting the candidate rule with the highest priority score from the candidate rules belonging to the optimal rule type.

6. The product push method according to claim 5, characterized in that, The optimal rule is obtained by selecting the candidate rule corresponding to the maximum priority score from the candidate rules belonging to the optimal rule type, including: The third intermediate rule is obtained by selecting the candidate rule with the highest priority score from the candidate rules belonging to the optimal rule type. If there are at least two third intermediate rules, or if the priority score difference between the third intermediate rule and the remaining candidate rules belonging to the optimal rule type is less than a preset difference threshold, then an optimal rule determination request is generated based on the candidate rules belonging to the optimal rule type, and the optimal rule determination request is displayed. Obtain the selection information of the business personnel regarding the optimal rule determination request, and determine the candidate rule corresponding to the selection information as the optimal rule.

7. The product push method according to claim 1, characterized in that, The user behavior data includes the user's first click-through rate for the target product, and after pushing the target product corresponding to the optimal rule to the user's terminal, it also includes: After the target product is pushed to the user, user behavior data related to the target product is obtained to obtain target behavior data, which includes the user's second click rate for the target product. The actual impact of the optimal rule is determined based on the first click-through rate and the second click-through rate, and the learning rate is determined based on the online duration of the optimal rule. Using a rule-based influence prediction model, the influence of the optimal rule is predicted based on the target behavior data to obtain the expected influence degree. Calculate the difference between the actual impact and the expected impact, calculate the product of the difference and the learning rate, and calculate the sum of the product and the static weights of the optimal rule to obtain the new rule static weights; Update the static weights of the optimal rule to the new static weights of the rule.

8. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor to enable the at least one processor to perform the product delivery method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the product push method as described in any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the product push method as described in any one of claims 1 to 7.

Images