Smart contract generation method based on user portrait

Abstract

The invention relates to the technical field of data processing, in particular to an intelligent contract generation method based on a user portrait, and the method comprises the steps: capturing a plurality of operation behavior features of a user side on a target page, extracting a plurality of semantic tags, and carrying out the tag extraction based on the behavior feature semantic association degree of the semantic tags among the operation behavior features; the method comprises the steps of extracting a main label and a derivative auxiliary label aiming at a user side, constructing a user portrait set of the user side, judging the confidence tendency of the user portrait set according to the distribution characteristics of the main label and the derivative auxiliary label, optimizing the user portrait set aiming at the confidence tendency of the user portrait set, and generating push content based on the optimized user portrait set. And generating a transaction contract when a transaction is concluded at the user side. According to the method and the device, confidence tendency judgment on the user portrait set of the user is realized, the optimization mode of the user portrait set is adaptively adjusted, and the accuracy and the reliability of contract generation based on the user portrait are improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method for generating smart contracts based on user profiles. Background Technology

[0002] In current digital economy applications, smart contracts, as digital protocols that automatically execute contract terms, are increasingly widely used. From decentralized finance to supply chain management and digital copyright transactions, smart contracts can significantly reduce trust costs and improve transaction efficiency. However, the generation of smart contracts depends on the precise and unambiguous definition of their triggering conditions and execution logic. In end-user scenarios, how to automatically generate smart contracts that truly match user intent and interests is a key technical challenge. User profiling can analyze users' historical behavioral data to build a labeled model depicting user interests, preferences, and needs, and then make personalized recommendations based on this model, ultimately guiding transactions and contract generation. However, most existing user profiling methods rely on simple statistics and labeling of user behavior. User behavior data is filled with a lot of noise, such as accidental clicks, temporary interest exploration, or misleading operations. These noisy labels are mixed with users' real and stable core interests, resulting in low confidence in the constructed user profile. This leads to biases in the generation of push content or even smart contracts based on such noisy profiles. Therefore, improving the accuracy and reliability of smart contract generation based on user profiles is an urgent technical problem to be solved.

[0003] For example, Chinese patent application publication number CN113946569A discloses a method for constructing user profiles. The method involves: acquiring a large amount of user behavior data; building a fact label library based on the collected behavior data; training a label model using logistic regression across multiple fact label libraries; constructing a user profile by matching the similarity between the user and the label model library using behavior weights; and continuously revising and adjusting the user profile using a time decay factor. This invention applies a mathematical model based on Newton's law of cooling to predict that the correlation between historical and current user behavior weakens over time. A function related to time decay is established to correct the user's label attributes.

[0004] The following problems still exist in the existing technology: Existing technologies do not consider that user behavior data may contain noisy data, which affects the confidence of user profiles and thus leads to deviations in the content pushed and the generated contracts. Existing technologies cannot determine the confidence tendency of user profile sets, nor can they adaptively adjust the optimization methods of user profile sets, thus affecting the accuracy and reliability of contract generation based on user profiles. Summary of the Invention

[0005] To this end, the application provides a smart contract generation method based on user portrait, to overcome the problem that the prior art cannot determine the confidence tendency of the user portrait set of a user, cannot adaptively adjust the optimization mode of the user portrait set, and affects the accuracy and reliability of contract generation based on user portrait.

[0006] To achieve the above-mentioned object, the application provides a smart contract generation method based on user portrait, comprising: capturing a plurality of operation behavior characteristics of the user terminal on the target page, extracting a plurality of semantic labels corresponding to the operation behavior characteristics within a predetermined period, and performing label extraction based on the behavior characteristic semantic correlation degree of the semantic labels between the operation behavior characteristics, to extract the main label and the derived sub-label for the user terminal; constructing a user portrait set of the user terminal based on the main label and the derived sub-label, and determining the confidence tendency of the user portrait set according to the distribution characteristics of the main label and the derived sub-label; optimizing the user portrait set according to the confidence tendency of the user portrait set, comprising, determining the trigger time corresponding to each main label and each derived sub-label, and determining whether to exclude the derived sub-label based on the time sequence arrangement relationship of the main label and the derived sub-label; or, performing cluster analysis on the operation behavior characteristics corresponding to each derived sub-label, identifying a potential cluster set, performing observation pushing based on the potential cluster set, and determining whether to convert the derived sub-label in the potential cluster set into a main label according to the feedback result of the observation pushing; generating push content based on the optimized user portrait set, and generating a transaction contract when a transaction is achieved at the user terminal; wherein the observation pushing comprises generating push information based on the derived sub-label in the potential cluster set, and obtaining the operation behavior characteristics of the user terminal for the push information.

[0007] Further, the process of determining the behavior characteristic semantic correlation degree of the semantic labels between the operation behavior characteristics comprises, obtaining a plurality of semantic labels corresponding to each operation behavior characteristic; calculating the semantic correlation degree of each semantic label between a plurality of operation behavior characteristics; determining the average value of the semantic correlation degree as the feature semantic correlation degree.

[0008] Further, the process of marking the first operation behavior characteristic and the second operation behavior characteristic comprises, obtaining the behavior characteristic semantic correlation degree of a plurality of operation behavior characteristics of the user terminal; if the behavior characteristic semantic correlation degree of the operation behavior characteristic exceeds a preset feature semantic correlation degree threshold, the operation behavior characteristic is marked as a first operation behavior characteristic. If the behavior feature semantic correlation degree of the operation behavior feature does not exceed the preset feature semantic correlation degree threshold, the operation behavior feature is marked as a second operation behavior feature.

[0009] Further, the process of extracting the main label and the derived secondary label for the user end includes, clustering the first operation behavior features to obtain an operation behavior clustering set, and marking the semantic labels of all operation behavior features in the operation behavior clustering set as main labels; clustering the second operation behavior features to obtain an operation behavior non-clustering set, and marking the semantic labels of all operation behavior features in the operation behavior non-clustering set as derived secondary labels.

[0010] Further, the process of determining the confidence tendency of the user portrait set according to the distribution characteristics of the main label and the derived secondary label includes, If the distribution characteristics of the main label and the derived secondary label meet the explicit confidence condition, it is determined that the confidence tendency of the user portrait set is an explicit confidence tendency. If the distribution characteristics of the main label and the derived secondary label do not meet the explicit confidence condition, it is determined that the confidence tendency of the user portrait set is a non-explicit confidence tendency. The explicit confidence condition is that the difference between the number of main labels and the number of derived secondary labels exceeds a preset difference threshold.

[0011] Further, the process of determining the optimization mode for optimizing the user portrait set includes, If the confidence tendency of the user portrait set is an explicit confidence tendency, the trigger time corresponding to each main label and each derived secondary label is determined, and it is determined whether to remove the derived secondary label based on the time sequence arrangement relationship of the main label and the derived secondary label. If the confidence tendency of the user portrait set is a non-explicit confidence tendency, clustering analysis is performed on the operation behavior features corresponding to each derived secondary label, a potential clustering set is identified, observation pushing is performed based on the potential clustering set, and it is determined whether to convert the derived secondary label in the potential clustering set to a main label according to the feedback result of the observation pushing.

[0012] Further, the process of determining whether to remove the derived secondary label based on the time sequence arrangement relationship of the main label and the derived secondary label includes, The trigger time corresponding to the main label is determined as the main label time, and the trigger time corresponding to the derived secondary label is determined as the secondary label time. A plurality of main label times and secondary label times are sorted in time sequence. If the secondary label time is within a time period formed by two adjacent main label times, the derived secondary label corresponding to the secondary label time is removed.

[0013] Further, the process of identifying the potential cluster set comprises, obtaining operation behavior features corresponding to each of the derived secondary labels, and calculating semantic correlation degrees of each semantic label between the operation behavior features; determining the average of the semantic correlation degrees as a potential feature semantic correlation degree; if the potential feature semantic correlation degree of the operation behavior feature corresponding to the derived secondary label exceeds a preset potential feature semantic correlation degree threshold, marking the operation behavior feature as a potential operation behavior feature; performing clustering on the potential operation behavior features to obtain the potential cluster set.

[0014] Further, the process of determining whether to convert the derived secondary label in the potential cluster set into a primary label comprises, respectively obtaining operation behavior features of push information generated by the user end for each derived secondary label in the potential cluster set within a predetermined observation period; if the push information generated by the user end for the derived secondary label meets a conversion condition, determining to convert the derived secondary label into a primary label.

[0015] Further, the conversion condition is that the number of operation behavior features of the push information generated by the user end for the derived secondary label exceeds a preset number threshold.

[0016] Compared with the prior art, the present application has the beneficial effects that the present application captures several operation behavior features of the user end on the target page, extracts several semantic labels corresponding to the operation behavior features within a predetermined period, performs label extraction based on the behavior feature semantic correlation degrees of the semantic labels between the operation behavior features, extracts the primary label and the derived secondary label for the user end, constructs the user portrait set of the user end based on the primary label and the derived secondary label, determines the confidence tendency of the user portrait set according to the distribution characteristics of the primary label and the derived secondary label, optimizes the user portrait set according to the confidence tendency of the user portrait set, generates push content based on the optimized user portrait set, generates a transaction contract when a transaction is achieved at the user end, and further, the present application realizes the confidence tendency determination of the user portrait set of the user, adaptively adjusts the optimization mode of the user portrait set, and improves the accuracy and reliability of the contract generation based on the user portrait.

[0017] In particular, this invention determines the confidence tendency of a user profile set based on the distribution characteristics of the main tags and derived sub-tags. It can be understood that by quantifying the quantitative differences in the distribution of main tags and derived sub-tags, the overall quality and reliability of the user profile are macroscopically diagnosed. The concentration and consistency of user behavior are used as the criteria for confidence. When the number of main tags is significantly greater than that of derived sub-tags, it indicates that the user's interests are clear and their behavioral patterns are stable, and the profile has a high degree of certainty and representativeness, i.e., an explicit confidence tendency. When the number of both is roughly the same, it indicates that the user's behavior is scattered, their interests are vague, or they are still in the formation stage, and the profile is full of uncertainty, i.e., a non-explicit confidence tendency. By clarifying the confidence level of user profiles through differences in tag distribution, this provides a quality basis for subsequent applications, avoids directly using vague and unreliable profiles for recommendations or contract generation, reduces decision-making errors, provides precise direction for profile optimization, improves the accuracy and reliability of downstream applications, reduces resource waste, and optimizes system efficiency. By introducing confidence tendency determination, the system automatically selects the optimal processing path based on the confidence status of the profile, improving the accuracy and security of subsequent content push and smart contract generation, avoiding waste of computing resources. Ultimately, this enables confidence tendency determination of user profile sets, improving the accuracy and reliability of contract generation based on user profiles.

[0018] In particular, this invention optimizes user profiles by selecting an optimization method based on the user profile set's confidence tendency. This means that for explicit confidence tendencies, by analyzing the temporal relationships of behaviors, unrepresentative derivative sub-labels mixed within the core interest time periods are removed. For non-explicit confidence tendencies, scattered weak signals are clustered, and tests are conducted by observing push notifications. Potential interests with positive user feedback, i.e., derivative sub-labels, are transformed into primary labels. When user interests are clear, the system can quickly purify the profile, concentrating all computing power and push resources on high-value primary labels, thereby improving push efficiency and the reliability of contract generation. When user interests are ambiguous, resources are redirected to valuable exploration, uncovering potential user needs through low-cost, high-return exploratory pushes, laying the foundation for precision, rather than blindly pushing or making arbitrary decisions. This improves the success rate of personalized recommendations and user satisfaction, ensuring that smart contracts generated based on user profiles are built on a relatively credible user intent, reducing business risks. Therefore, this invention achieves an optimization method that selects an optimization method based on the user profile set's confidence tendency, improving the accuracy and reliability of contract generation based on user profiles.

[0019] Especially, when the confidence tendency of the user portrait set is an explicit confidence tendency, whether the derived secondary label is removed is determined based on the time sequence arrangement relationship of the main label and the derived secondary label. It can be understood that noise is removed through the time sequence relationship, incidental interference embedded in the core interest context of the user is identified and removed, the purity of the portrait is improved without damaging the real interest portrait of the user, so that the subsequent personalized push content is more pure and focused, avoiding resource waste and recommendation inaccuracy caused by incidental interest, at the same time, reducing the risk of mistakenly deleting valuable interest labels, laying a solid foundation for generating high-reliability smart contracts, ensuring that the contract content is highly consistent with the real and stable intention of the user, improving the automation level and security of the entire transaction process. When the user is in a highly focused and stable interest period, the core behavior will form a relatively continuous interest field. The secondary or incidental behavior, i.e. the derived secondary label, that occurs in this period is most likely to be noise or temporary attention drift. The user's behavior is arranged on the timeline, and when it is found that the triggering moment of a derived secondary label is wrapped between two consecutive main label moments, it is determined that the behavior is incidental interference generated by the user in the process of browsing the core interest. The context of the noise is identified, rather than isolated, and then the optimization mode of selecting the optimization of the user portrait set according to the confidence tendency of the user portrait set is realized, and the accuracy and reliability of the contract generation based on the user portrait are improved.

[0020] Especially, when the confidence tendency of the user portrait set is an explicit confidence tendency, whether the derived secondary label is removed is determined based on the time sequence arrangement relationship of the main label and the derived secondary label. It can be understood that noise is removed through the time sequence relationship, incidental interference embedded in the core interest context of the user is identified and removed, the purity of the portrait is improved without damaging the real interest portrait of the user, so that the subsequent personalized push content is more pure and focused, avoiding resource waste and recommendation inaccuracy caused by incidental interest, at the same time, reducing the risk of mistakenly deleting valuable interest labels, laying a solid foundation for generating high-reliability smart contracts, ensuring that the contract content is highly consistent with the real and stable intention of the user, improving the automation level and security of the entire transaction process. When the user is in a highly focused and stable interest period, the core behavior will form a relatively continuous interest field. The secondary or incidental behavior, i.e. the derived secondary label, that occurs in this period is most likely to be noise or temporary attention drift. The user's behavior is arranged on the timeline, and when it is found that the triggering moment of a derived secondary label is wrapped between two consecutive main label moments, it is determined that the behavior is incidental interference generated by the user in the process of browsing the core interest. The context of the noise is identified, rather than isolated, and then the optimization mode of selecting the optimization of the user portrait set according to the confidence tendency of the user portrait set is realized, and the accuracy and reliability of the contract generation based on the user portrait are improved. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 A step diagram of the smart contract generation method based on the user portrait according to an embodiment of the present application is shown in FIG. 1. The smart contract generation method based on the user portrait according to an embodiment of the present application comprises the following steps. Figure 2 A logic flow diagram of the marking of the first operation behavior feature and the second operation behavior feature according to an embodiment of the present application is shown in FIG. 2. The marking of the first operation behavior feature and the second operation behavior feature according to an embodiment of the present application comprises the following steps. Figure 3 A logic flow diagram of the determination of the confidence tendency of the user portrait set according to an embodiment of the present application is shown in FIG. 3. The determination of the confidence tendency of the user portrait set according to an embodiment of the present application comprises the following steps. Figure 4 A logic flow diagram of the determination of the optimization mode for optimizing the user portrait set according to an embodiment of the present application is shown in FIG. 4. The determination of the optimization mode for optimizing the user portrait set according to an embodiment of the present application comprises the following steps. DETAILED DESCRIPTION

[0022] In order to make the objects and advantages of the present application clearer, the present application will be further described below with reference to the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and should not be used to limit the present application.

[0023] The preferred embodiments of the present application will be described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present application and are not used to limit the protection scope of the present application.

[0024] It should be noted that in the description of the present application, the terms indicating the direction or positional relationship such as "upper", "lower", "inner", "outer" and the like are based on the direction or positional relationship shown in the drawings, which is only for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present application.

[0025] In addition, it should also be noted that in the description of the present application, unless otherwise explicitly specified and limited, the terms "mounting", "connecting" should be understood broadly, for example, it can be fixed connection, or detachable connection, or integral connection; it can be mechanical connection, or electrical connection; it can be direct connection, or indirect connection through an intermediate medium, or internal communication of two elements. Those skilled in the art can understand the specific meaning of the above terms in the present application according to the specific circumstances.

[0026] Please refer to Figure 1 A step diagram of the smart contract generation method based on the user portrait according to an embodiment of the present application is shown in FIG. 1. The smart contract generation method based on the user portrait according to an embodiment of the present application comprises the following steps. Step S100, capturing a plurality of operation behavior features of the user terminal on the target page, extracting a plurality of semantic tags corresponding to the operation behavior features within a predetermined period, and performing tag extraction based on the behavior feature semantic correlation degree of the semantic tags between the operation behavior features to extract the main tag and the derived secondary tag for the user terminal; Specifically, the operation behavior feature can be a click behavior of the user, the predetermined period can be set by a person skilled in the art according to the accuracy requirement of the smart contract generated based on the user portrait, the higher the accuracy requirement, the shorter the setting, and the value range can be [12, 24] with an interval unit of h, and preferably, 15 h.

[0027] Step S200, constructing a user portrait set of the user end based on the main label and the derived secondary label, determining the confidence tendency of the user portrait set according to the distribution characteristics of the main label and the derived secondary label; Step S300, optimizing the user portrait set according to the confidence tendency of the user portrait set, including, determining the trigger time corresponding to each main label and each derived secondary label, and determining whether to remove the derived secondary label based on the time sequence arrangement relationship of the main label and the derived secondary label; Or, clustering analysis is performed on the operation behavior features corresponding to each of the derived secondary labels, a potential cluster set is identified, observation pushing is performed based on the potential cluster set, and it is determined whether to convert the derived secondary label in the potential cluster set into a main label according to the feedback result of the observation pushing; Step S400, generating pushing content based on the optimized user portrait set, and generating a transaction contract when a transaction is achieved at the user end; Wherein, the observation pushing includes generating pushing information based on the derived secondary label in the potential cluster set, and obtaining the operation behavior feature of the user end for the pushing information.

[0028] Specifically, in the implementation, the semantic correlation degrees of each semantic label between the operation behavior features are calculated respectively, and the average of the semantic correlation degrees is taken as the behavior feature semantic correlation degree. The operation behavior feature corresponds to multiple semantic labels, and the semantic label is essentially a keyword. The cosine similarity measurement method can be used, for example, the cosine similarity is calculated after the keyword is vectorized, and the cosine similarity is determined as the semantic correlation degree between the keywords. Of course, other methods can also be used, which will not be described here.

[0029] Specifically, when generating the pushing content based on the user portrait set, the associated content of the main label is preferentially pushed.

[0030] Specifically, the process of determining the behavior feature semantic correlation degree of the semantic label between the operation behavior features includes, obtaining a plurality of semantic labels corresponding to each operation behavior feature; calculating the semantic correlation degrees of each semantic label between a plurality of operation behavior features; determining the average of the semantic correlation degrees as the feature semantic correlation degree.

[0031] Please refer to Figure 2As shown, it is a logic flow chart for marking the first operation behavior feature and the second operation behavior feature according to the embodiment of the application, and the process of marking the first operation behavior feature and the second operation behavior feature comprises, obtaining the behavior feature semantic correlation degree corresponding to the operation behavior feature of the user terminal; if the behavior feature semantic correlation degree of the operation behavior feature exceeds the preset feature semantic correlation degree threshold, marking the operation behavior feature as the first operation behavior feature; if the behavior feature semantic correlation degree of the operation behavior feature does not exceed the preset feature semantic correlation degree threshold, marking the operation behavior feature as the second operation behavior feature.

[0032] Specifically, the preset feature semantic correlation degree threshold is the product of the feature semantic correlation degree reference value and the feature semantic factor, the feature semantic correlation degree reference value is the mean value of the feature semantic correlation degree in the historical data, the feature semantic factor can be set by the person skilled in the art according to the accuracy requirement of the smart contract based on the user portrait, the higher the accuracy requirement, the larger the setting, and the value range can be [1.1, 1.25], preferably, it can be 1.2.

[0033] Specifically, the process of extracting the main label and the derived sub-label for the user terminal comprises, clustering the first operation behavior feature to obtain an operation behavior clustering set, and marking the semantic label of all operation behavior features in the operation behavior clustering set as the main label; clustering the second operation behavior feature to obtain an operation behavior non-clustering set, and marking the semantic label of all operation behavior features in the operation behavior non-clustering set as the derived sub-label.

[0034] Please refer to Figure 3 As shown, it is a logic flow chart for determining the confidence tendency of the user portrait set according to the embodiment of the application, and the process of determining the confidence tendency of the user portrait set according to the distribution characteristics of the main label and the derived sub-label comprises, if the distribution characteristics of the main label and the derived sub-label meet the explicit confidence condition, determining that the confidence tendency of the user portrait set is the explicit confidence tendency; if the distribution characteristics of the main label and the derived sub-label do not meet the explicit confidence condition, determining that the confidence tendency of the user portrait set is the non-explicit confidence tendency; The explicit confidence condition is that the difference between the number of the main label and the number of the derived sub-label exceeds the preset difference threshold.

[0035] Specifically, the preset difference threshold is a product of a difference reference value and an explicit factor, the difference reference value is a mean value of the differences in the historical data, the explicit factor can be set by a person skilled in the art according to the accuracy requirement of the smart contract generated based on the user portrait, the higher the accuracy requirement, the larger the set value, and the value range can be [1.1, 1.2], preferably, it can be 1.15.

[0036] Specifically, the embodiment of the present application determines the confidence tendency of the user portrait set according to the distribution characteristics of the main label and the derived sub-label. It can be understood that by quantifying the distribution difference of the main label and the derived sub-label in quantity, the overall quality and reliability of the user portrait is macro-diagnosed, and the concentration and consistency of user behavior are taken as the judgment standard of confidence. When the number of main labels is significantly more than that of derived sub-labels, it indicates that the user's interest is clear and the behavior mode is stable, and the portrait has high certainty and representativeness, i.e. explicit confidence tendency. When the number of the two is similar, it indicates that the user's behavior is scattered, the interest is vague or is still in the formation period, and the portrait is full of uncertainty, i.e. non-explicit confidence tendency. The confidence level of the portrait is determined by the label distribution difference, which provides a quality basis for subsequent application, avoids using ambiguous and unreliable portraits directly for recommendation or contract generation, reduces decision-making errors, provides a precise direction for portrait optimization, improves the accuracy and reliability of downstream applications, reduces resource waste, optimizes system efficiency, and automatically selects the optimal processing path according to the confidence state of the portrait by introducing the confidence tendency determination, which improves the accuracy and reliability of the contract generation based on the user portrait.

[0037] Please refer to Figure 4 The embodiment of the present application determines the optimization mode of optimizing the user portrait set, and the process of determining the optimization mode of optimizing the user portrait set includes, If the confidence tendency of the user portrait set is explicit confidence tendency, the trigger time corresponding to each main label and each derived sub-label is determined, and whether the derived sub-label is removed is determined based on the time sequence arrangement relationship of the main label and the derived sub-label. If the confidence tendency of the user portrait set is non-explicit confidence tendency, the operation behavior characteristics corresponding to each derived sub-label are clustered and analyzed, the potential cluster set is identified, the observation push is performed based on the potential cluster set, and whether the derived sub-label in the potential cluster set is converted into a main label is determined according to the feedback result of the observation push.

[0038] Specifically, the embodiment of the present application can understand that, for explicit confidence tendency, by analyzing the behavior time sequence relationship, removing the derivative sub-labels mixed in the core interest time period and not representative, for non-explicit confidence tendency, clustering the scattered weak signals, and testing through observation pushing, converting the potential interest of the user with positive feedback, i.e. derivative sub-labels, into main labels, when the user's interest is clear, the system can quickly purify the portrait, and concentrate all computing power and pushing resources on high-value main labels, thereby improving the pushing efficiency and the reliability of contract generation, when the user's interest is ambiguous, diverting resources to valuable exploration, and through exploratory pushing with low cost and high return, mining the potential demand of the user, laying the foundation for precision, rather than blind pushing or arbitrary decision-making, improving the success rate of personalized recommendation and user satisfaction, ensuring that the intelligent contract generated based on the portrait is established on the basis of relatively reliable user intention, reducing business risk, and further, realizing the optimization mode of optimizing the user portrait set according to the confidence tendency of the user portrait set, improving the accuracy and reliability of the contract generation based on the user portrait.

[0039] Specifically, the process of determining whether to remove the derivative sub-label based on the time sequence arrangement relationship between the main label and the derivative sub-label includes, determining the trigger time corresponding to the main label as the main label time, and determining the trigger time corresponding to the derivative sub-label as the sub-label time; sequencing the main label time and the sub-label time in time sequence; if the sub-label time is within the period formed by the adjacent two main label times, the derivative sub-label corresponding to the sub-label time is removed; if the sub-label time is not within the period formed by the adjacent two main label times, the derivative sub-label corresponding to the sub-label time is not removed.

[0040] Specifically, when the confidence tendency of the user portrait set is an explicit confidence tendency, the embodiment of the present application determines whether to remove the derived secondary label based on the time sequence arrangement relationship of the main label and the derived secondary label. It can be understood that by removing noise through the time sequence relationship, accidental interference embedded in the core interest context of the user is identified and removed, the purity of the portrait is improved without damaging the real interest portrait of the user, so that the subsequent personalized push content is more pure and focused, avoiding resource waste and recommendation inaccuracy caused by accidental interest, at the same time, reducing the risk of mistakenly deleting valuable interest labels, laying a solid foundation for generating high-reliability smart contracts, ensuring that the contract content is highly consistent with the user's real and stable intention, improving the automation level and security of the entire transaction process. When the user is in a highly focused and stable interest period, the core behavior will form a relatively continuous interest field. The secondary or accidental behavior, i.e. the derived secondary label, that occurs in this period is most likely to be noise or temporary attention drift. Arranging the user's behavior on the timeline, when it is found that the triggering moment of a derived secondary label is wrapped between two consecutive main label moments, it can be determined that the behavior is accidental interference generated by the user in the process of browsing the core interest. Contextual recognition of noise rather than isolated judgment, and then, the optimization mode of selecting the user portrait set for optimization according to the confidence tendency of the user portrait set is realized, improving the accuracy and reliability of the contract generation based on the user portrait.

[0041] Specifically, the process of identifying the potential clustering set includes, Obtaining operation behavior characteristics corresponding to each of the derived secondary labels, calculating semantic correlation degrees of each semantic label between a plurality of operation behavior characteristics; Determining the average of the semantic correlation degrees as a potential feature semantic correlation degree; If the potential feature semantic correlation degree of the operation behavior characteristics corresponding to the derived secondary label exceeds a preset potential feature semantic correlation degree threshold, the operation behavior characteristics are marked as potential operation behavior characteristics; If the potential feature semantic correlation degree of the operation behavior characteristics corresponding to the derived secondary label does not exceed the preset potential feature semantic correlation degree threshold, the operation behavior characteristics are not marked; Clustering the potential operation behavior characteristics to obtain the potential clustering set.

[0042] Specifically, the preset potential feature semantic correlation degree threshold is not more than the preset feature semantic correlation degree threshold, the preset potential feature semantic correlation degree threshold is a product of a potential feature semantic correlation degree reference value and a potential feature factor, the potential feature semantic correlation degree reference value is a mean value of the potential feature semantic correlation degrees in the historical data, the potential feature factor can be set by a person skilled in the art according to an accuracy requirement of the smart contract based on the user portrait, the higher the accuracy requirement, the larger the potential feature factor, and the value range can be [1.1, 1.2], preferably, the potential feature factor can be 1.15.

[0043] Specifically, the process of determining whether to convert the derived sub-label in the potential cluster set into a main label includes, respectively acquiring operation behavior features of push information generated by the user terminal for each derived sub-label in the potential cluster set within a predetermined observation period; If the push information generated by the user terminal for the derived sub-label meets the conversion condition, it is determined that the derived sub-label is converted into a main label. If the push information generated by the user terminal for the derived sub-label does not meet the conversion condition, it is determined that the derived sub-label is not converted.

[0044] Specifically, the predetermined observation period can be set by a person skilled in the art according to an accuracy requirement of the smart contract based on the user portrait, the higher the accuracy requirement, the longer the predetermined observation period, and the value range can be [2, 6], the interval unit being h, preferably, the predetermined observation period can be 4h.

[0045] Specifically, the conversion condition is that the number of operation behavior features of the push information generated by the user terminal for the derived sub-label exceeds a preset number threshold.

[0046] Specifically, the preset number threshold is a product of a number reference value and a conversion factor, the number reference value is a mean value of the number in the historical data, and the conversion factor can be set by a person skilled in the art according to an accuracy requirement of the smart contract based on the user portrait, the higher the accuracy requirement, the larger the conversion factor, and the value range can be [1.1, 1.25], preferably, the conversion factor can be 1.2.

[0047] Specifically, when the confidence tendency of the user portrait set is the non-explicit confidence tendency, the embodiment of the present application performs clustering analysis on the operation behavior characteristics corresponding to each derived secondary label, identifies a potential clustering set, performs observation pushing based on the potential clustering set, and determines whether to convert the derived secondary label in the potential clustering set into a primary label. It can be understood that when the confidence tendency is non-explicit confidence, the user's scattered behavior, i.e., the derived secondary label, may hide a potential interest pattern that has not yet taken shape. Through clustering analysis, the internal relationship is found in these weak signals, and the sporadic behaviors with similar semantics are aggregated into a more meaningful potential clustering set. The observation pushing mechanism is started, an active and data-driven hypothesis testing process is performed, and the hypothesis is verified through the real-time feedback of the user, i.e., the number of operation behavior characteristics. If the user feedback is positive, it proves that the potential interest exists in reality, and it is converted from the derived secondary label to the primary label. Thus, the resource allocation is optimized, and blind pushing is not required for all scattered behaviors. Instead, the high-potential interest clustered is probed in a targeted, low-cost and efficient manner, the portrait quality is improved, empirical and high-credibility data basis is provided for the generation of intelligent contracts, commercial risks caused by misjudgment of potential demand are reduced, and thus, the optimization mode of selecting the user portrait set for optimization according to the confidence tendency of the user portrait set is realized, and the accuracy and reliability of the contract generation based on the user portrait are improved.

[0048] So far, the technical solutions of the present application have been described in combination with the preferred embodiments shown in the drawings, but those skilled in the art can easily understand that the protection scope of the present application is obviously not limited to these specific embodiments. Those skilled in the art can make equivalent changes or replacements to related technical features without departing from the principles of the present application, and the technical solutions after the changes or replacements will fall within the protection scope of the present application.

[0049] The above description is only the preferred embodiments of the present application and is not intended to limit the present application; for those skilled in the art, the present application can have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for generating smart contracts based on user profiles, characterized in that, include: Capture several operational behavior features of the user on the target page, extract several semantic tags corresponding to the operational behavior features within a predetermined period, and extract tags based on the semantic correlation of the semantic tags between operational behavior features to extract the main tags and derived sub-tags for the user. A user profile set for the user terminal is constructed based on the main label and the derived sub-label. The confidence tendency of the user profile set is determined based on the distribution characteristics of the main label and the derived sub-label. Optimizing the user profile set based on the confidence level of the user profile set includes, Determine the trigger time corresponding to each main tag and each derived sub-tag, and determine whether to remove the derived sub-tag based on the temporal arrangement relationship between the main tags and the derived sub-tags; Alternatively, cluster analysis can be performed on the operational behavior characteristics corresponding to each of the derived sub-tags to identify potential cluster sets. Based on the potential cluster sets, observation and push can be performed, and the feedback results of the observation and push can be used to determine whether to convert the derived sub-tags in the potential cluster sets into main tags. Push content is generated based on the optimized user profile set, and a transaction contract is generated when a transaction is completed on the user's end. The observation push includes generating push information based on derived sub-labels in the potential cluster set and obtaining user terminal operation behavior characteristics in response to the push information.

2. The smart contract generation method based on user profiles according to claim 1, characterized in that, The process of determining the semantic correlation between semantic labels of operational behavioral features includes, Obtain several semantic labels corresponding to the characteristics of each operational behavior; Calculate the semantic correlation degree of each semantic label among several operational behavior features; The mean of semantic relevance is determined as the semantic relevance of the feature.

3. The smart contract generation method based on user profiles according to claim 2, characterized in that, The process of marking the first operational behavior feature and the second operational behavior feature includes, Obtain the semantic correlation degree of several operational behavior features of the user terminal; If the semantic correlation degree of the operation behavior feature exceeds the preset semantic correlation degree threshold, then the operation behavior feature is marked as the first operation behavior feature. If the semantic correlation of the operational behavior feature does not exceed the preset semantic correlation threshold, then the operational behavior feature is marked as the second operational behavior feature.

4. The smart contract generation method based on user profiles according to claim 3, characterized in that, The process of extracting the main tags and derived sub-tags for the user includes, Cluster the first operational behavior features to obtain an operational behavior cluster set, and mark the semantic labels of all operational behavior features in the operational behavior cluster set as the main labels; Cluster the second operational behavior features to obtain an operational behavior non-cluster set, and mark the semantic labels of all operational behavior features in the operational behavior non-cluster set as derived sub-labels.

5. The smart contract generation method based on user profiles according to claim 4, characterized in that, The process of determining the confidence tendency of the user profile set based on the distribution characteristics of the main label and derived sub-labels includes: If the distribution characteristics of the main label and the derived sub-labels meet the explicit confidence condition, then the confidence tendency of the user profile set is determined to be an explicit confidence tendency. If the distribution characteristics of the main label and the derived sub-label do not meet the explicit confidence condition, then the confidence tendency of the user profile set is determined to be a non-explicit confidence tendency. The explicit confidence condition is that the difference between the number of primary labels and the number of derived secondary labels exceeds a preset difference threshold.

6. The smart contract generation method based on user profiles according to claim 5, characterized in that, The process of determining optimization methods for user profile sets includes, If the confidence tendency of the user profile set is an explicit confidence tendency, then determine the trigger time corresponding to each main tag and each derived sub-tag, and determine whether to remove the derived sub-tag based on the temporal arrangement relationship of the main tag and the derived sub-tag; If the confidence tendency of the user profile set is non-explicit confidence tendency, then cluster analysis is performed on the operational behavior features corresponding to each of the derived sub-labels to identify potential cluster sets, observation push is performed based on the potential cluster sets, and the feedback results of the observation push determine whether to convert the derived sub-labels in the potential cluster sets into main labels.

7. The smart contract generation method based on user profiles according to claim 6, characterized in that, The process of determining whether to remove the derived sub-tags based on the temporal arrangement relationship between the main tag and the derived sub-tags includes, The trigger time corresponding to the main tag is determined as the main tag time, and the trigger time corresponding to the derived sub-tag is determined as the sub-tag time. Sort the main tag moments and sub-tag moments in chronological order; If the time of the sub-label falls within the time period formed by two adjacent time periods of the main label, then the derived sub-label corresponding to the time of the sub-label will be removed.

8. The smart contract generation method based on user profiles according to claim 7, characterized in that, The process of identifying potential cluster sets includes, Obtain the operational behavior features corresponding to each of the derived sub-labels, and calculate the semantic correlation degree between each semantic label among several operational behavior features; The mean of semantic relevance is determined as the semantic relevance of latent features; If the potential feature semantic correlation degree of the operation behavior feature corresponding to the derived sub-label exceeds the preset potential feature semantic correlation degree threshold, then the operation behavior feature is marked as a potential operation behavior feature. Cluster the potential operational behavior features to obtain the potential cluster set.

9. The smart contract generation method based on user profiles according to claim 8, characterized in that, The process of determining whether to convert derived sublabels in a potential cluster set into primary labels includes, The user's operational behavior characteristics for push information generated for each derived sub-label in the potential cluster set within a predetermined observation period are obtained respectively. If the push notification generated by the user client for the derived sub-tag meets the conversion conditions, it is determined that the derived sub-tag will be converted into the main tag.

10. The smart contract generation method based on user profiles according to claim 9, characterized in that, The conversion condition is that the number of user-side action behavior features in the push information generated by the derived sub-tag exceeds a preset threshold.

Images