Marketing activity user risk warning method, device, storage medium and equipment

Abstract

The present invention relates to a method, device, storage medium and equipment for warning user risks in marketing activities, including a characteristic vector representation corresponding to user risk characteristics corresponding to multiple user risk types existing in the marketing activities, a risk type rule tree is constructed through a chi-square coefficient, and a target rule tree is formed by associating warning conditions. The user data of the marketing activities is monitored, and when the warning conditions are triggered, the virtual business logic flow is guided to avoid the impact on the real business. At the same time, the user account is locked, and a risk warning is generated according to the warning conditions of the target node and the associated node, so as to effectively identify and respond to the user risks in the marketing activities.

Description

Technical Field

[0001] The present disclosure relates to the field of early warning technology, and in particular to a method, device, storage medium and equipment for early warning of user risks in marketing activities. Background Art

[0002] With the rapid development of the Internet and mobile Internet, marketing activities have become one of the important ways for enterprises to attract customers, enhance brand awareness, increase sales, and promote products and services. The scale of marketing activities is getting bigger and bigger, the forms are getting more and more diverse, and the ways to attract user participation are also becoming more and more diverse. The risks of marketing activities are also increasing. For example, maliciously obtaining benefits through fake orders or other illegal and irregular means not only affects the interests of other customers, but also causes actual economic losses and reputational crises to enterprises.

[0003] In relevant scenarios, when conducting marketing activities, traditional audit methods and manual operations by dedicated personnel are used to determine whether there are malicious order-padding or other illegal and irregular means. They lack effective risk control measures, are inefficient and prone to omissions, and are unable to detect potential risks in a timely manner, and cannot guarantee real-time monitoring and early warning of each business event in the marketing activities. Summary of the invention

[0004] In view of the technical problems that there is a lack of effective risk control measures for marketing activities in relevant scenarios, which are inefficient and prone to omissions, and potential risks cannot be discovered in a timely manner, and real-time monitoring and early warning of each business event in the marketing activities cannot be guaranteed, the purpose of the present invention is to provide a marketing activity user risk warning method, device, storage medium and equipment, which aims to effectively prevent and control the risks existing in marketing activities, so as to avoid unnecessary economic losses and reputation crises during the entire marketing activity, thereby maximizing the risks and effects of marketing activities and ensuring the safe conduct of marketing activities.

[0005] In order to achieve the above-mentioned purpose, a first aspect of an embodiment of the present disclosure provides a method for early warning of user risks in a marketing activity, the method comprising:

[0006] According to the user risk characteristics corresponding to the multiple user risk types existing in the marketing activity, the characteristic vector corresponding to each user risk type is represented as the observed frequency, and the characteristic vectors corresponding to the remaining user risk types are represented as the expected frequency to traverse and calculate the chi-square coefficient, and according to the chi-square coefficient, a risk type rule tree for the marketing activity for the user risk type is constructed;

[0007] Obtaining a warning condition corresponding to each node in the risk type rule tree, associating the warning condition with each node in the risk type rule tree in a one-to-one correspondence, and obtaining a target risk type rule tree for the user risk type;

[0008] Determining whether to trigger the warning condition corresponding to any node in the target risk type rule tree according to the acquired user data participating in the marketing activity;

[0009] In the case where the user data triggers the warning condition corresponding to any target node in the target risk type rule tree, the user account corresponding to the user data is guided to the virtual business logic flow corresponding to the target node, wherein the virtual business logic flow is the same as the real business logic flow of the marketing activity, and the user account does not generate real business data for the operation of virtual promotion and discount rules during the operation of the virtual business logic flow;

[0010] The user account corresponding to the user data is locked, and the associated node is determined from the target risk type rule tree according to the target node, and the marketing activity user risk warning is generated and issued according to the user data, the target node and the corresponding warning condition, the associated node and the corresponding warning condition.

[0011] According to a second aspect of the embodiments of the present disclosure, there is provided a marketing activity user risk warning device, the device comprising:

[0012] A calculation and construction module is configured to calculate the chi-square coefficient according to the user risk characteristics corresponding to the multiple user risk types existing in the marketing activity, with the characteristic vector corresponding to each user risk type represented as the observed frequency and the characteristic vectors corresponding to the remaining user risk types represented as the expected frequency, and construct the risk type rule tree of the marketing activity for the user risk type based on the chi-square coefficient;

[0013] an acquisition and association module, configured to acquire the warning condition corresponding to each node in the risk type rule tree, associate the warning condition with each node in the risk type rule tree in a one-to-one correspondence, and obtain a target risk type rule tree for the user risk type;

[0014] a determination module configured to determine whether the warning condition corresponding to any node in the target risk type rule tree is triggered according to the acquired user data participating in the marketing activity;

[0015] a guiding module, configured to guide the user account corresponding to the user data to the virtual business logic flow corresponding to any target node in the target risk type rule tree when the user data triggers the warning condition corresponding to the target node, wherein the virtual business logic flow is the same as the real business logic flow of the marketing activity, and the user account does not generate real business data for the operation of the virtual promotion and discount rules during the operation of the virtual business logic flow;

[0016] The locking and warning module is configured to lock the user account corresponding to the user data, and determine the associated node from the target risk type rule tree according to the target node, and generate and issue the marketing activity user risk warning according to the user data, the target node and the corresponding warning condition, the associated node and the corresponding warning condition.

[0017] According to a third aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the steps of any one of the methods described in the first aspect are implemented.

[0018] According to a fourth aspect of the embodiments of the present disclosure, an electronic device is provided, including:

[0019] a memory having a computer program stored thereon;

[0020] A processor is used to execute the computer program in the memory to implement the steps of any one of the methods in the first aspect.

[0021] The present invention provides a marketing activity user risk warning method, device, storage medium and equipment. Compared with the prior art, it has the following beneficial effects:

[0022] By calculating the chi-square coefficient between the characteristic vectors corresponding to different user risk types, the differences and correlations between risk types can be accurately quantified, and then a risk type rule tree for specific user risk types can be constructed, which improves the accuracy of risk identification and the precision of classification. The warning conditions are associated with the nodes of the risk type rule tree one by one to form a target risk type rule tree. It can automatically monitor the user data participating in marketing activities. Once the preset warning conditions are triggered, it can respond immediately without manual intervention, which greatly improves the efficiency and timeliness of risk warning. When the user data triggers the warning conditions, instead of directly interrupting the user operation or taking compulsory measures, the user account is guided to a virtual business logic flow environment, which not only ensures the continuity of the marketing activities, but also avoids user dissatisfaction or loss caused by direct intervention. At the same time, the user's operation in the virtual environment does not generate real business data, which effectively prevents the further spread of potential risks. After the user account is locked, the user risk is comprehensively evaluated according to the warning conditions of the target node and the associated node, and the corresponding risk warning is generated. It helps enterprises to take targeted risk management measures in a timely manner. Through non-invasive risk handling methods and timely risk warning systems, it can minimize the impact on user experience while protecting its own interests. It helps to enhance users' trust and loyalty to the brand, and effectively prevent and warn of risks in marketing activities, so as to avoid unnecessary economic losses and reputation crises in the entire marketing process, thereby maximizing the risks and effects of marketing activities and ensuring the safe implementation of marketing activities.

[0023] Other features and advantages of the present disclosure will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The accompanying drawings are used to provide a further understanding of the present disclosure and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the present disclosure but do not constitute a limitation of the present disclosure. In the accompanying drawings:

[0025] Figure 1 It is a flow chart of a method for early warning of user risks in marketing activities according to an embodiment of the specification.

[0026] Figure 2 It is a block diagram of a marketing activity user risk warning device shown in an embodiment of the specification.

[0027] Figure 3 It is a block diagram of another marketing activity user risk warning device shown in an embodiment of the specification. DETAILED DESCRIPTION

[0028] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] The specific implementation of the present disclosure is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described herein is only used to illustrate and explain the present disclosure, and is not used to limit the present disclosure.

[0030] In order to achieve the above-mentioned purpose, the present disclosure provides a marketing activity user risk warning method, Figure 1 This is a flow chart of a method for early warning of user risk in marketing activities according to an embodiment. The method comprises:

[0031] In step S11, according to the user risk features corresponding to the multiple user risk types existing in the marketing activity, the characteristic vector corresponding to each user risk type is represented as the observed frequency, and the characteristic vectors corresponding to the remaining user risk types are represented as the expected frequency to traverse and calculate the chi-square coefficient, and according to the chi-square coefficient, a risk type rule tree for the marketing activity for the user risk type is constructed;

[0032] Among them, User Risk Type refers to different types of user risks that may be encountered in marketing activities, which may involve fraud, credit risk, abnormal behavior, etc. Each user risk type has its own specific characteristics or behavior patterns.

[0033] User Risk Characteristics: refers to specific information or indicators that can represent or reflect the characteristics of a certain user risk type. These characteristics can be the user's transaction behavior, login mode, historical credit record, registration, login, coupon collection, number of orders, number of payments, number of overdue payments, income level, debt ratio, etc., which are used to identify and distinguish different user risk types.

[0034] Feature Vector: It is used to represent multiple risk features used to represent the risk type of the user.

[0035] Observed Frequency: Observed frequency refers to the number of times a certain event or category is actually observed. In this step, the feature vector corresponding to each user risk type is expressed as the observed frequency, that is, the number of times the risk type feature is actually observed.

[0036] Expected Frequency: Explanation: In statistics, expected frequency is the number of times a certain event or category should appear, calculated based on a certain theory or hypothetical model. In this step, the feature vectors corresponding to the remaining user risk types are expressed as expected frequencies, that is, the number of times these features should appear if all user risk types are evenly distributed or conform to a certain hypothetical model.

[0037] Chi-Square Coefficient: The chi-square coefficient (also known as the chi-square statistic) is a statistic used to measure the degree of difference between the actual observed frequency and the expected frequency. In statistics, it is often used to test whether there is a significant difference between two or more sample distributions. In this step, the chi-square coefficient is calculated to evaluate the differences and correlations between the characteristics of different user risk types.

[0038] Risk Type Rule Tree: A data mining model based on a tree structure, used to represent and infer the hierarchical relationships and rules between different user risk types. It is constructed based on statistics such as the chi-square coefficient. Each node represents a user risk type, and the connections between nodes represent rules or decision paths.

[0039] In the disclosed embodiment, the Chi-Square Test in statistics is used to analyze the feature differences between different user risk types, and then a risk type rule tree is constructed. The Chi-Square Test is a widely used hypothesis testing method, which is used to compare whether the difference between the actual observed frequency and the expected frequency is significant. In this scenario, the feature vector corresponding to each user risk type is regarded as the observed frequency, and the value of the feature vector corresponding to other user risk types under the assumption of uniform distribution or a certain specific distribution is regarded as the expected frequency. By traversing and calculating the chi-square coefficients between all user risk type features, the differences and correlations between them can be quantified, and then a risk type rule tree reflecting these relationships can be constructed.

[0040] For example, suppose that in an e-commerce marketing campaign, there are three types of user risks: fraud risk, credit risk, and normal users. Several key user behavior features are selected as feature vectors, such as transaction amount, transaction frequency, return rate, etc.

[0041] Data collection and preprocessing: First, collect historical data and construct feature vectors for each user risk type. For example, the feature vector of a fraud risk user may be characterized by high transaction amount, low transaction frequency, high return rate, etc.

[0042] Calculate observed frequency and expected frequency: For each user risk type, the value of each dimension of its feature vector is the observed frequency. The expected frequency is estimated based on the value of the feature vector of other risk types under the assumption of uniform or specific distribution.

[0043] Calculate the chi-square coefficient by traversing: For each pair of risk types and their characteristics, calculate the chi-square coefficient between them. For example, calculate the difference between fraud risk and credit risk in transaction amount characteristics, and get the chi-square value by comparing the observed frequency and expected frequency of the two.

[0044] Constructing a risk type rule tree: Construct a risk type rule tree based on the size of the chi-square coefficient. Features with larger chi-square values ​​indicate that the two risk types differ significantly in this feature, so they should be differentiated first in the tree structure. For example, if the chi-square value of fraud risk and credit risk on transaction amount is the largest, the transaction amount may be used as a child node or a late split node.

[0045] In step S12, the warning condition corresponding to each node in the risk type rule tree is obtained, and the warning condition is associated with each node in the risk type rule tree in a one-to-one correspondence to obtain a target risk type rule tree for the user risk type;

[0046] Warning Condition: refers to the pre-set rules or conditions that trigger risk warnings. When user behaviors or features that meet these conditions are detected, the risk warning mechanism will be triggered. In this step, each node of the risk type rule tree corresponds to one or more warning conditions.

[0047] Target Risk Type Rule Tree: It is obtained by further associating and refining the warning conditions based on the risk type rule tree. It is customized and optimized for specific user risk types to improve the accuracy and efficiency of risk warning.

[0048] In the disclosed embodiment, based on the risk type rule tree constructed in step S11, step S12 further refines the rule tree and sets warning conditions for each node. These warning conditions are derived based on business logic and historical data analysis, and are used to determine whether user behavior meets the characteristics of a certain risk type. By associating the warning conditions with the nodes of the rule tree one by one, we obtain a target risk type rule tree for a specific user risk type.

[0049] Continuing with the example of the e-commerce marketing campaign above: Analyze the warning conditions: Based on business experience and historical data, set warning conditions for each node in the risk type rule tree. For example, under the transaction amount node, you can set "transaction amount exceeds XX yuan and return rate is higher than XX%" as a warning condition for fraud risk.

[0050] Associate warning conditions with nodes: Associate each warning condition with the corresponding node. In the target risk type rule tree, each node clearly indicates the warning condition that needs to be met to trigger the node.

[0051] Optimization and verification: Verify the accuracy and effectiveness of the target risk type rule tree through historical data. Adjust and optimize nodes or warning conditions that do not meet expectations to ensure that the rule tree can accurately identify and warn user risks.

[0052] In step S13, determining whether the warning condition corresponding to any node in the target risk type rule tree is triggered according to the acquired user data participating in the marketing activity;

[0053] Among them, user data refers to various information related to users participating in marketing activities, including but not limited to user identity, transaction records, behavior logs, etc. These data are used to analyze and evaluate the risk status of users to determine whether to trigger risk warnings. The system monitors the marketing activities of enterprises in real time, captures and records important business events, so as to adjust and optimize marketing strategies in a timely manner. In the process of monitoring the marketing activities of enterprises, various business events may be involved, such as registration, login, coupon collection, order placement, payment, etc. Specifically, it may include the number of user draws, the total value of prizes issued, and the accumulated points of users for marketing activities, etc. These business events are very important for understanding user behavior, evaluating marketing risks, and handling abnormal situations in a timely manner. Check and verify the scanned business events, compare the business risk control rules, and determine whether the warning conditions are triggered to ensure that they meet the warning threshold range of the previously created business risk control rules.

[0054] In the disclosed embodiment, during the real-time marketing activities, the target risk type rule tree is traversed based on the acquired user data to determine whether the user behavior triggers the warning conditions of any node. If triggered, corresponding risk management measures are taken, such as suspending transactions, further verifying user identities, etc.

[0055] In e-commerce marketing activities, the real-time monitoring system continuously collects data about users participating in the activities. Whenever a new user participates or an existing user makes a transaction, the system inputs this data into the target risk type rule tree for traversal:

[0056] Data input: Input the user's latest data (such as transaction amount, transaction frequency, return rate, number of times the user has drawn a lottery, the total value of prizes issued, and the user's accumulated points for marketing activities, etc.) into the rule tree.

[0057] Traversing the rule tree: Starting from the root node, determine whether the warning conditions of the current node are met according to the user data one by one. If not, continue to traverse the child nodes; if they are met, take corresponding risk management measures according to the definition of the rule tree.

[0058] Risk management measures: If user data triggers the warning conditions of a certain node, the system will immediately execute the preset risk management measures. For example, if it is detected that the user's behavior meets the warning conditions of fraud risk, the system may suspend the user's transaction and require identity verification.

[0059] In step S14, when the user data triggers the warning condition corresponding to any target node in the target risk type rule tree, the user account corresponding to the user data is guided to the virtual business logic flow corresponding to the target node, wherein the virtual business logic flow is the same as the real business logic flow of the marketing activity, and the user account does not generate real business data for the operation of the virtual promotion and discount rules during the operation of the virtual business logic flow;

[0060] Among them, virtual business logic flow is a virtual environment that simulates the real business logic process, which is used to test and evaluate user behavior without actually affecting the system status and business data. In marketing activities, when user data triggers risk warnings, the user account can be guided to the virtual business logic flow to further observe and analyze its behavior patterns.

[0061] The same business logic means that the virtual business logic flow is consistent with the real business logic flow in terms of logical design, that is, the two have the same structure and logic in terms of business processes, operation steps, rule judgments, etc. This ensures that the user behavior observed in the virtual environment can truly reflect its behavior pattern in the real environment.

[0062] No real business data is generated. In the virtual business logic flow, all operations of the user account will not affect the real business data of the system. This means that users can try to buy goods, use promotions and discount rules, etc., but these operations will not change the real business data such as inventory quantity, order status, transaction records, etc.

[0063] In the disclosed embodiment, when it is determined that the user data triggers the warning condition corresponding to a target node in the target risk type rule tree, for example, the number of user draws exceeds the upper limit of the activity limit, the total value of the prizes issued exceeds the set activity budget limit; the user accumulates points for the marketing activity too quickly or the total score exceeds the preset score threshold. These can be used to determine the risk that the user has malicious intent for the marketing activity. In order to further assess the user risk without interfering with the business logic of the real marketing activity, the user account is directed to a virtual business logic flow corresponding to the target node. This virtual business logic flow is logically the same as the business logic of the real marketing activity, but all operations are performed in an isolated virtual environment and no real business data is generated. The purpose of this is to allow users to continue their business processes without affecting other users or the system status, while the system can observe and analyze the user's behavior in the virtual environment to further confirm their risk level.

[0064] For example, suppose that in an e-commerce marketing campaign, user A's purchase behavior triggers the warning condition of "high-frequency large-value transactions and frequent address changes" in the target risk type rule tree. At this time, user A's transaction will not be blocked immediately, but will be guided to a virtual shopping process. In this virtual process, user A can see the same products, promotions, and discount rules as in the real marketing campaign, but all operations such as ordering and payment will not actually affect the inventory, order, or financial system.

[0065] In step S15, the user account corresponding to the user data is locked, and the associated node is determined from the target risk type rule tree according to the target node, and the marketing activity user risk warning is generated and issued according to the user data, the target node and the corresponding warning condition, the associated node and the corresponding warning condition.

[0066] Among them, user account locking is used to temporarily limit the access and operation permissions of a user account when it is identified as a potential risk. By locking the user account, potential risk behaviors can be prevented from causing damage to the system or other users. When a business event is detected to have triggered an early warning condition, the system will automatically lock the business event and stop the next business logic flow of the business event. Ensure the security of marketing activities and avoid further asset losses.

[0067] Associated nodes are nodes in the target risk type rule tree that are directly or indirectly associated with the target node. These nodes may represent more detailed risk types or subtypes, or provide additional risk assessment information. By analyzing associated nodes, the judgment of user risk can be further refined.

[0068] Marketing activity user risk warning is used to timely discover and report potential user risks during marketing activities. When the system detects that user data triggers risk warning conditions, it will generate risk warnings based on information such as target nodes, associated nodes, and user data, and notify relevant personnel to take necessary measures to deal with the risks.

[0069] In the disclosed embodiment, after confirming that the user data has triggered the warning conditions in the target risk type rule tree and the user account has been directed to the virtual business logic flow, the user account lock operation will be executed to prevent potential risk behaviors from causing damage to the real system. At the same time, according to the target node, the associated nodes (these associated nodes may represent more detailed risk types or subtypes) are determined from the target risk type rule tree, and the user data, the target node and its warning conditions, the associated nodes and their warning conditions, etc. are integrated to generate and issue a marketing activity user risk warning. This warning will contain detailed information of the user account, the triggered risk type, possible risk behaviors, and recommended countermeasures.

[0070] For example, continuing with the e-commerce marketing campaign example above, when user A's behavior in the virtual shopping process further confirms his risk level (for example, trying to use multiple different payment methods or requesting to change the delivery address to a high-risk area), the system will lock user A's account. Subsequently, the system generates a risk warning based on the target node "high-frequency large-value transactions and frequent address changes" and possible associated nodes (such as "using multiple payment methods" or "delivery address in high-risk areas"). The warning information will be sent to the management team of the marketing campaign or the risk management department, prompting them to pay attention to user A's account and take further investigation or handling measures. At the same time, the system may also send a notification to user A to inform him that his account has been locked and ask him to provide additional authentication information or explain his transaction behavior. Notifications are automatically sent to relevant staff, including business personnel, technical personnel, etc., through various communication forms such as email / SMS / DingTalk robots, so that appropriate measures can be taken in time to deal with the problem or prevent the situation from further deteriorating. The notification content usually includes specific information about the event, the time of occurrence, the scope of impact, and recommended response measures, etc.

[0071] In the above technical solution, by calculating the chi-square coefficient between the characteristic vectors corresponding to different user risk types, the differences and correlations between the risk types can be accurately quantified, and then a risk type rule tree for a specific user risk type can be constructed, which improves the accuracy of risk identification and the precision of classification. The warning conditions are associated with the nodes of the risk type rule tree one by one to form a target risk type rule tree. It can automatically monitor the user data participating in the marketing activities. Once the preset warning conditions are triggered, it can respond immediately without manual intervention, which greatly improves the efficiency and timeliness of risk warning. When the user data triggers the warning conditions, instead of directly interrupting the user operation or taking compulsory measures, the user account is guided to a virtual business logic flow environment, which not only ensures the continuity of the marketing activities, but also avoids user dissatisfaction or loss caused by direct intervention. At the same time, the user's operation in the virtual environment does not generate real business data, which effectively prevents the further spread of potential risks. After the user account is locked, the user risk is comprehensively evaluated according to the warning conditions of the target node and the associated node, and the corresponding risk warning is generated. It helps enterprises to take targeted risk management measures in a timely manner. Through non-invasive risk handling methods and timely risk warning systems, it can protect its own interests while minimizing the impact on user experience. It helps to enhance users' trust and loyalty to the brand, and effectively prevent and control risks in marketing activities, so as to avoid unnecessary economic losses and reputation crises in the entire marketing process, thereby maximizing the risks and effects of marketing activities and ensuring the safety of marketing activities.

[0072] In a possible implementation, in step S11, the user risk features corresponding to the multiple user risk types existing in the marketing activity are expressed as observed frequencies by taking the characteristic vector corresponding to each of the user risk types and expressing the characteristic vectors corresponding to the remaining user risk types as expected frequencies to traverse and calculate the chi-square coefficient, and based on the chi-square coefficient, construct a risk type rule tree for the marketing activity for the user risk type, including:

[0073] In step S111, according to the user risk characteristics corresponding to the multiple user risk types existing in the marketing activity, the characteristic vector corresponding to each user risk type is represented as the observed frequency, and the characteristic vectors corresponding to the remaining user risk types are represented as the expected frequency to traverse and calculate the first chi-square coefficient, and the user risk type corresponding to the observed frequency when the maximum first chi-square coefficient is taken as the root node of the risk type rule tree;

[0074] In the embodiment of the present disclosure, the characteristic vector is constructed: for each user risk type, a characteristic vector is constructed according to its corresponding user risk characteristics. This vector contains a series of numerical values ​​or classification data that can describe the characteristics of the risk type.

[0075] Observed frequency and expected frequency: The feature vector corresponding to each user risk type is regarded as the observed frequency, which indicates the performance of the risk type in the actual data. At the same time, the feature vectors corresponding to other user risk types are regarded as expected frequencies, which indicate the number of times these risk types should appear without specific assumptions or conditions (but in fact, this is not entirely the case because they are used for comparison).

[0076] Chi-square coefficient calculation: Traverse all user risk types and calculate the chi-square coefficient between the characteristic vector (observed frequency) of each risk type and the characteristic vector (expected frequency) of other risk types. The chi-square coefficient is used to measure the difference between the observed frequency and the expected frequency. The greater the difference, the more unique the risk type is in the data.

[0077] By traversing all user risk types, the chi-square coefficient between the feature vector of each risk type and the feature vectors of other risk types is calculated, and the risk type with the largest chi-square coefficient is selected as the root node.

[0078] If there are multiple types of risk, such as fraud risk, credit risk, low-value user risk, overdue payment risk, return risk, refund without return risk, etc. For each type, calculate the chi-square coefficient between its feature vector and the feature vectors of other types. Select the fraud risk with the largest chi-square coefficient as the root node of the risk type rule tree.

[0079] In step S112, according to the characteristic vector representation corresponding to the root node and the characteristic vector representation corresponding to the user risk characteristics corresponding to the multiple user risk types existing in the marketing activity, a child node user risk type that is a child node of the root node in the risk type rule tree is determined;

[0080] In the disclosed embodiment, based on the root node (such as fraud risk) that has been determined, the user risk type of the child node in the risk type rule tree is further determined. This usually involves further analysis and comparison of the remaining risk types to determine which risk types should be used as child nodes of the root node. This process may involve recursively applying chi-square coefficient calculations or other statistical methods.

[0081] Specifically, feature vector matching and differentiation: Analyze the feature vector of the root node (fraud risk) to understand its performance pattern in the data. Then, compare the feature vectors of other user risk types (such as credit risk, low-value user risk, etc.) with the feature vector of the root node to find out the differences and similarities between them. These differences and similarities will determine which risk types should be used as child nodes.

[0082] Business logic and rule design: Based on the comparison of feature vectors, business logic and expert knowledge are combined to design rules. This can accurately reflect the inherent connections and differences between different risk types and help with subsequent risk assessment and early warning. For example, it may be known that fraud risk is closely related to abnormal transaction patterns, while credit risk is more related to the user's repayment history and credit score, so different risk types are organized into rule trees.

[0083] Recursive construction or hierarchical design: Recursively apply the above process to build a multi-level rule tree. That is, after determining the child nodes of the root node, the characteristic vectors of these child nodes can be further analyzed and their child nodes can be determined. Through this recursive construction method, a complete, multi-level rule tree can be formed. However, in actual applications, the appropriate hierarchy is selected according to business needs and data characteristics.

[0084] In step S113, the following steps are repeated: the user risk type that has been determined to correspond to the node is eliminated, and after the elimination, the characteristic vectors of the user risk characteristics corresponding to the child nodes are respectively represented as the observed frequency, and the characteristic vectors corresponding to the remaining user risk types are represented as the expected frequency to traverse and calculate the second chi-square coefficient;

[0085] In the embodiment of the present disclosure, the classification of user risk types is refined through an iterative process until all risk types are properly assigned to each node of the risk type rule tree. The specific process involves:

[0086] Eliminate identified risk types: At each iteration step, the user risk type that has been identified as corresponding to a specific node (such as a parent node or a processed child node) will be eliminated from the list of candidates currently under consideration. This is done to avoid double counting and misclassification.

[0087] Setting of observed frequency and expected frequency: For each remaining user risk type, according to its feature vector (such as quantitative features such as user behavior and transaction mode), compare them with the user risk features corresponding to the sub-node currently considered. Among them, the feature vector of the user risk feature corresponding to the sub-node is set as the observed frequency, while the feature vectors corresponding to other user risk types are used as the expected frequency.

[0088] Calculate the second chi-square coefficient: Use the chi-square test formula to calculate the chi-square coefficient (also called the second chi-square coefficient) between the remaining user risk type and the risk characteristics of the current child node. This step aims to quantify the degree of match between the remaining risk type and the characteristics of the current child node in order to make further classification decisions.

[0089] In step S114, according to the second chi-square coefficient, the sub-child node user risk types of each sub-node in the risk type rule tree are determined from the eliminated user risk types, until all the user risk types are in the risk type rule tree.

[0090] In the disclosed embodiment, the user risk type with the highest (or lowest, depending on how the chi-square coefficient is used) chi-square coefficient is selected as the sub-child node of the current child node. This selection is based on the degree of match or difference represented by the chi-square coefficient, that is, the higher the chi-square coefficient, the less the risk type matches the characteristics of the current child node (or the more it matches, depending on the specific usage scenario and interpretation).

[0091] Assume that it is found through calculation that risk type C has the highest chi-square coefficient with the characteristics of the current child node (assuming that a high chi-square coefficient indicates a higher degree of match). Therefore, when constructing the risk type rule tree, risk type C is determined as the second child node of the current child node. Then, repeat steps S113 and S114, and for risk type C (now as a new child node), eliminate the determined risk types (including A and C itself, if other risk types are subsequently added to C's child nodes), and continue to calculate the chi-square coefficients of the remaining risk types with the characteristics of C's child nodes in the same way, so as to further construct the risk type rule tree. Continue until all user risk types are assigned to the appropriate positions in the risk type rule tree.

[0092] In a possible implementation, in step S112, determining the child node user risk type of a child node of the root node in the risk type rule tree according to the characteristic vector representation corresponding to the root node and the characteristic vector representation corresponding to the user risk characteristics corresponding to the multiple user risk types existing in the marketing activity includes:

[0093] In step S1121, the user risk feature corresponding to the root node is eliminated;

[0094] In the disclosed embodiment, the user risk features that have been determined as the root node are removed from the risk type set. When constructing the next level of the risk type rule tree, focus on the risk types that have not been classified as direct subclasses of the root node. By removing the user risk features corresponding to the root node, it can be ensured that subsequent calculations and analyses are performed on the remaining risk types, thereby avoiding duplication and confusion.

[0095] Assume that in the process of constructing the risk type rule tree, "fraud risk" has been determined as the root node. User risk features directly related to "fraud risk" will be removed from the set of all user risk types. This means that if there is a specific user risk type (such as "high frequency trading fraud") that is clearly classified as a subclass of "fraud risk", then in this step it will no longer be considered as a candidate for subsequent child node selection.

[0096] In step S1122, after elimination, the characteristic vector corresponding to each of the user risk types is represented as the observed frequency, and the characteristic vectors corresponding to the remaining user risk types are represented as the expected frequency to traverse and calculate the chi-square coefficient of the child node;

[0097] In the disclosed embodiment, the remaining user risk types after removing the risk features corresponding to the root node are traversed to calculate their chi-square coefficients as potential child nodes. The purpose of this step is to quantify the difference between each remaining risk type and other risk types outside the root node, so as to determine which risk types are more suitable as direct subclasses of the root node.

[0098] Specifically, the feature vector corresponding to each remaining user risk type is taken as the observed frequency, and the average value (or some form of summary) of the feature vectors corresponding to all other remaining risk types is taken as the expected frequency. Then, the chi-square test formula is used to calculate the chi-square coefficient of each risk type. The size of the chi-square coefficient reflects the degree of difference between the observed frequency and the expected frequency. The greater the difference, the higher the degree of distinction between the risk type and other risk types, and the more likely it is to become a direct subclass of the root node.

[0099] Continuing with the previous example, assume that after removing "fraud risk" and its direct subclasses, the following user risk types remain: credit risk (C1), low-value user risk (C2), and payment abnormality risk (C3). Take the feature vector of each risk type as the observed frequency, calculate the difference between them and the average value of the feature vector of the remaining risk types, and obtain their respective chi-square coefficients. For example, for credit risk (C1), its chi-square coefficient will reflect the degree of difference between it and low-value user risk (C2) and payment abnormality risk (C3) in terms of feature vectors.

[0100] In step S1123, a preset number of the user risk types with the largest chi-square coefficients of the child nodes are selected as the child-node user risk types of the child nodes of the root node in the risk type rule tree.

[0101] In the disclosed embodiment, a preset number of user risk types with the largest chi-square coefficients are selected as child nodes of the root node. This step is based on an assumption that the risk type with a larger chi-square coefficient has a higher degree of distinction from other risk types outside the root node, and is therefore more suitable as a direct child of the root node.

[0102] The choice of preset number depends on the specific application scenario and business needs. For example, if you want to build a simpler rule tree, you may choose a smaller preset number; if you want to divide risk types more finely, you may choose a larger preset number.

[0103] Continuing with the previous example, assuming that the preset number of child nodes is 2, the two user risk types with the largest chi-square coefficients are selected as child nodes of the "fraud risk" root node. For example, if the chi-square coefficients of credit risk (C1) and payment abnormality risk (C3) rank first and second respectively, they will be selected as direct subclasses of "fraud risk" and as nodes of the next level in the risk type rule tree.

[0104] In a possible implementation, in step S112, determining the child node user risk type of a child node of the root node in the risk type rule tree according to the characteristic vector representation corresponding to the root node and the characteristic vector representation corresponding to the user risk characteristics corresponding to the multiple user risk types existing in the marketing activity includes:

[0105] In step S11201, a deviation value between the characteristic vector representation corresponding to the root node and the characteristic vector representation corresponding to the user risk features corresponding to the plurality of user risk types existing in the marketing activity is calculated;

[0106] In the disclosed embodiment, the difference between the feature vector representation corresponding to the root node and the feature vector representation corresponding to the multiple user risk types existing in the marketing activity is quantified. This difference is measured by calculating the deviation value, and the size of the deviation value reflects the distance or inconsistency between a certain user risk type and the root node risk type in the feature space. The deviation value can be calculated by a variety of methods, such as Euclidean distance, Manhattan distance, inverse of cosine similarity, etc. The specific choice depends on the nature of the feature vector and the requirements of the application scenario.

[0107] Assuming the root node is "fraud risk", its corresponding feature vector is expressed as V_root = [f1_root, f2_root, ..., fn_root], where fi_root represents the quantitative value of "fraud risk" on a certain feature. At the same time, there are multiple user risk types in marketing activities, such as "high-risk users" (feature vector is V_high = [f1_high, f2_high,..., fn_high]), "abnormal transaction behavior" (feature vector is V_abnormal = [f1_abnormal, f2_abnormal, ..., fn_abnormal]), etc.

[0108] You can choose Euclidean distance as the calculation method of the deviation value. For each user risk type Vi (i represents a different risk type), its deviation value D_i from the root node. You can get the deviation value between each user risk type and the root node, and these deviation values ​​will be used for child node selection in subsequent steps.

[0109] In step S11202, the user risk type represented by a preset number of characteristic vectors and having the largest deviation value from the characteristic vector corresponding to the root node user risk type is selected as the child node user risk type of the child node of the root node in the risk type rule tree.

[0110] In the disclosed embodiment, according to the calculated deviation value, a preset number of user risk types with the largest deviation values ​​are selected as the child node user risk types of the root node. The user risk type with a larger deviation value is farther away from the root node in the feature space, and its characteristics are more significantly different from the root node risk type, so it is more suitable as a direct subclass of the root node.

[0111] Continuing with the previous example, assuming that the preset number of child nodes is 2, the two user risk types with the largest deviation values ​​will be selected as child nodes of the "fraud risk" root node based on the calculated deviation value. Assume that through calculation, the deviation value of "high-risk users" is D_high, the deviation value of "abnormal transaction behavior" is D_abnormal, and D_high>D_abnormal (that is, the deviation between "high-risk users" and "fraud risk" is larger). If there are other user risk types, continue to compare their deviation values. Finally, select the two user risk types with the largest deviation values ​​(in this case, "high-risk users" and another risk type with the second largest deviation value. If "abnormal transaction behavior" is the second largest, both are selected at the same time) as the direct subclasses of "fraud risk" and construct the nodes of the next level in the risk type rule tree.

[0112] In a possible implementation, in step S114, determining, from the eliminated user risk types according to the second chi-square coefficient, the sub-child node user risk type of each sub-node in the risk type rule tree includes:

[0113] In step S1141, when each of the sub-nodes is determined as the observation frequency, a preset number of user risk types with the largest second chi-square coefficients are selected as candidate user risk types;

[0114] In the disclosed embodiment, for each child node, the user risk type that is most likely to become its child node is screened out from its potential user risk types, that is, the candidate user risk type. This is achieved by calculating the second chi-square coefficient between each user risk type as the observed frequency of the current child node and the remaining user risk types (as the expected frequency). The second chi-square coefficient measures the difference between the observed frequency and the expected frequency. The larger the difference (in some cases, depending on the interpretation and application of the chi-square coefficient), the more the user risk type does not match the characteristics of the current child node (or in some applications, the more it matches, depending on how the chi-square coefficient is used), and thus is more likely to become its child node. By selecting a preset number of user risk types with the largest second chi-square coefficients, a set of candidate user risk types can be obtained.

[0115] Assume that there is a subnode "abnormal trading behavior" and consider multiple user risk types related to it, such as "large amount and fast transfer", "frequent trading at night", etc. For each user risk type, calculate its second chi-square coefficient with the remaining user risk types when it is the observation frequency of the subnode of "abnormal trading behavior". Assuming the preset number is 2, select the two user risk types with the largest second chi-square coefficient as the candidate user risk types of "abnormal trading behavior".

[0116] In step S1142, the priority of the child node is determined according to the deviation value between the characteristic vector representation corresponding to the child node and the characteristic vector representation corresponding to the root node;

[0117] In the disclosed embodiment, the priority of the subnode is determined based on a certain relationship between the subnode and the root node (such as feature difference, business importance, etc.). This is achieved by calculating the deviation value between the feature vector representation corresponding to the subnode and the feature vector representation corresponding to the root node. The larger the deviation value, the greater the difference in features between the subnode and the root node, which may represent a more important risk classification dimension, and therefore has a higher priority.

[0118] Continuing with the previous example, the deviation value between each child node (such as "abnormal transaction behavior") and the root node ("fraud risk") is calculated. The deviation value can be calculated by a variety of methods, such as Euclidean distance, Manhattan distance, etc. Based on the calculated deviation value, a priority can be assigned to each child node, and the child node with a larger deviation value has a higher priority.

[0119] In step S1143, according to the priority, it is determined whether the candidate user risk type corresponding to the child node is used as a candidate user risk type for other child nodes with higher priority;

[0120] In the disclosed embodiment, the final attribution of the candidate user risk type is determined according to the priority of the child node. Specifically, if a candidate user risk type is simultaneously regarded as a potential child node by multiple child nodes, the priority of the child node is used to determine which child node the user risk type finally belongs to. This is to avoid repeated allocation of user risk types in the risk type rule tree and ensure that each user risk type is uniquely classified under a child node.

[0121] Assume that "Large amount and fast transfer" is considered as an alternative user risk type by both "Abnormal transaction behavior" and another child node (assuming "High-risk user behavior"). According to the priority determined in step S1142, if "Abnormal transaction behavior" has a higher priority than "High-risk user behavior", "Large amount and fast transfer" will be assigned to "Abnormal transaction behavior" as its child node.

[0122] In step S1144, when the alternative user risk type corresponding to the child node is used as an alternative user risk type for other child nodes with higher priority, the user risk type is removed from the alternative user risk types of the child node to obtain the secondary child node user risk type of the child node in the risk type rule tree.

[0123] In the disclosed embodiment, user risk types that have been determined as child nodes by other child nodes with higher priority are removed from the candidate user risk types of the child nodes. This is to ensure that each user risk type is uniquely classified under a child node in the risk type rule tree to avoid duplication and confusion.

[0124] Continuing with the previous example, if "Large amount and fast transfer" has been determined as a child node of "Abnormal transaction behavior" in step S1143, then when constructing the risk type rule tree, "Large amount and fast transfer" is removed from the candidate user risk types of "High-risk user behavior" to ensure its uniqueness in the risk type rule tree. Ultimately, "Abnormal transaction behavior" will contain "Large amount and fast transfer" as one of its child nodes, while "High-risk user behavior" needs to find other candidate user risk types to construct its child nodes.

[0125] In a possible implementation, in step S15, determining the associated node from the target risk type rule tree according to the target node includes:

[0126] In step S151, it is determined whether the target node has a parent node. If the target node has a parent node, the parent node of the target node and the node having the same parent node as the target node are determined as the associated nodes;

[0127] In the disclosed embodiment, nodes associated with the target node are determined, in particular nodes identified by parent node relationships. First, the system checks whether the target node exists in the risk type rule tree and whether it has a parent node. If the target node does have a parent node, then the parent node and other nodes that share the same parent node with the target node are considered to be nodes associated with the target node (i.e., associated nodes). This association is established based on their hierarchical structure and direct parent-child relationship in the risk type rule tree.

[0128] Assume that in the risk type rule tree, there is a target node "Nighttime High Frequency Trading", which is located under the parent node "Abnormal Trading Behavior". At the same time, "Large Amount and Fast Transfer" is also a child node of "Abnormal Trading Behavior", but is at the same level as "Nighttime High Frequency Trading". When "Nighttime High Frequency Trading" is determined as the target node, check whether it has a parent node (i.e. "Abnormal Trading Behavior"), and then further identify all nodes that have the same parent node as "Nighttime High Frequency Trading" (i.e. "Large Amount and Fast Transfer" and other possible child nodes). Therefore, in this example, "Abnormal Trading Behavior" (parent node) and "Large Amount and Fast Transfer" (same-level node) are both considered as associated nodes of "Nighttime High Frequency Trading".

[0129] In step S152, it is determined whether the target node has a child node. If the target node has a child node, the child node of the target node is determined as the associated node.

[0130] In the disclosed embodiment, the determination of the associated nodes is based on whether the target node has child nodes. If the target node does have child nodes, then these child nodes are directly regarded as nodes associated with the target node. This association is also established based on the hierarchical structure and direct parent-child relationship in the risk type rule tree, but this time the focus is on the child nodes at the lower level when the target node is the parent node.

[0131] Continuing with the previous example, assume that the target node becomes "abnormal trading behavior". Check whether "abnormal trading behavior" has child nodes. Since "high-frequency trading at night" and "fast transfer of large amounts" are both child nodes of "abnormal trading behavior", during the execution of step S152, these two child nodes will be identified as associated nodes of "abnormal trading behavior". This means that when constructing risk type association relationships or conducting risk analysis, the impact of "high-frequency trading at night" and "fast transfer of large amounts" on "abnormal trading behavior" or the interaction between them can be considered.

[0132] In one possible implementation, the method further includes:

[0133] In step S16, in response to an unlocking operation of the locked user account by the enterprise account having unlocking authority, the corresponding locked user account is unlocked;

[0134] In the disclosed embodiment, the unlocking operation of a locked user account by a corporate account with unlocking authority can be identified and executed. This process usually involves three main steps: identity authentication, authority verification, and status update.

[0135] Authentication: First, the identity of the enterprise account initiating the unlock request needs to be verified to ensure that it is a legitimate account with corresponding permissions. This is usually achieved by checking the login credentials of the enterprise account (such as tokens, session IDs, etc.) and their matching with the backend database.

[0136] Permission verification: After confirming the identity of the enterprise account, the system needs to further verify whether the account has the permission to unlock the locked user account. This can be done by querying the permission database or permission model to confirm whether the role or permission set of the enterprise account includes the operation of "unlocking the user account".

[0137] Status update: Once the permission check is passed, the system will update the status of the locked user account from "locked" to "normal". At the same time, it may be necessary to record the log of this unlocking operation, including time, executor, unlocked account and other information for subsequent audit or tracking.

[0138] Assume that an e-commerce platform has a corporate account named "Admin Corp", which is responsible for managing multiple user accounts. When a user account "User123" is found to be automatically locked by the system due to abnormal operation, the administrator of the Admin Corp account decides to unlock the user account.

[0139] Authentication: When an administrator logs in to the Admin Corp account, the system verifies the administrator's identity by checking the login credentials (such as username and password) entered by the administrator.

[0140] Permission verification: The system queries the permission settings of the Admin Corp account to confirm that it has the permission to "unlock user accounts."

[0141] Status update: The administrator selects the "User123" account in the backend management interface and performs the unlock operation. After receiving the instruction, the system updates the status of the User123 account to "normal" and records the log of this unlock operation, including timestamp, operator (Admin Corp), unlocked account (User123) and other information.

[0142] In step S17, when the locked user account is unlocked, the operations of the user account on the virtual promotion and preferential rules during the operation of the virtual business logic flow are used as the operations of the real promotion and preferential rules in the real business logic flow of the marketing activity, so that the operations of the virtual promotion and preferential rules become the operations of the real promotion and preferential rules to generate real business data.

[0143] In the disclosed embodiment, the promotion and preferential rule operations performed by the locked user account in the virtual business logic flow are regarded as part of the real business logic flow, and real business data is generated accordingly. This requires the system to be able to track the operation records of the user account in the virtual environment, and after the account is unlocked, these operation records are seamlessly converted into operation records in the real business system.

[0144] Operation record tracking: During the flow of virtual business logic, the system needs to record all operations of the user account in real time, especially those related to promotion and discount rules. These records should include key information such as operation time, operation type, operation object, discount details, etc.

[0145] Verification after unlocking: Once the user account is unlocked, the system should automatically check the account's operation records in the virtual environment, especially those promotion and discount rule operation records that have not yet been converted into real business data.

[0146] Data conversion: For operation records that meet the conditions, the system should execute data conversion logic to convert the virtual operations in these records into corresponding operations in the real business system. This may involve calling the API of the real business system, updating database records, or triggering other business processes.

[0147] Business data generation: After the conversion is completed, the system should generate real business data, such as orders, transaction records, discount records, etc., and ensure that these data are consistent and accurate with the original virtual operation records.

[0148] Continuing with the above example, before unlocking, User123 account participated in the virtual promotion activity of the e-commerce platform and received a coupon worth 20 off for purchases over 100. However, because the account was locked, the use of the coupon was not reflected in the real business system. Operation record tracking: The system records the operation of User123 to receive the coupon in the virtual environment, including the time of receipt, coupon number, and usage conditions. Post-unlock verification: After the User123 account was unlocked, it was detected that the account had unprocessed virtual promotion and discount rule operation records. Data conversion: Execute the data conversion logic to convert the coupon information received by User123 in the virtual environment into the coupon record in the real business system, and ensure that the coupon can be used in subsequent shopping processes. Business data generation: When User123 makes the first purchase after unlocking and meets the coupon usage conditions, the system automatically applies the coupon and generates the corresponding order and transaction records, which include the usage details of the coupon and the discount amount. In this way, the operations originally performed in the virtual environment are successfully converted into business data in the real business system.

[0149] The present disclosure also provides a marketing activity user risk warning device, see Figure 2 As shown, the device comprises:

[0150] The calculation and construction module 210 is configured to calculate the chi-square coefficient according to the user risk characteristics corresponding to the multiple user risk types existing in the marketing activity, with the characteristic vector corresponding to each user risk type represented as the observed frequency, and the characteristic vectors corresponding to the remaining user risk types represented as the expected frequency, and construct the risk type rule tree of the marketing activity for the user risk type based on the chi-square coefficient;

[0151] The acquisition and association module 220 is configured to acquire the warning condition corresponding to each node in the risk type rule tree, associate the warning condition with each node in the risk type rule tree in a one-to-one correspondence, and obtain a target risk type rule tree for the user risk type;

[0152] The determination module 230 is configured to determine whether the warning condition corresponding to any node in the target risk type rule tree is triggered according to the acquired user data participating in the marketing activity;

[0153] The guidance module 240 is configured to guide the user account corresponding to the user data to the virtual business logic flow corresponding to the target node when the user data triggers the warning condition corresponding to any target node in the target risk type rule tree, wherein the virtual business logic flow is the same as the real business logic flow of the marketing activity, and the user account does not generate real business data for the operation of the virtual promotion and discount rules during the operation of the virtual business logic flow;

[0154] The locking and warning module 250 is configured to lock the user account corresponding to the user data, and determine the associated node from the target risk type rule tree according to the target node, and generate and issue the marketing activity user risk warning according to the user data, the target node and the corresponding warning condition, the associated node and the corresponding warning condition.

[0155] In a possible implementation, the calculation and construction module 210 is configured as follows:

[0156] According to the user risk characteristics corresponding to the multiple user risk types existing in the marketing activity, the characteristic vector corresponding to each user risk type is represented as the observed frequency, and the characteristic vectors corresponding to the other user risk types are represented as the expected frequency to traverse and calculate the first chi-square coefficient, and the user risk type corresponding to the observed frequency when the maximum first chi-square coefficient is obtained is used as the root node of the risk type rule tree;

[0157] Determine, according to the characteristic vector representation corresponding to the root node and the characteristic vector representation corresponding to the user risk features corresponding to the plurality of user risk types existing in the marketing activity, a child node user risk type that is a child node of the root node in the risk type rule tree;

[0158] Repeated execution: Eliminate the user risk type that has been determined to correspond to the node, and after elimination, respectively use the characteristic vectors of the user risk characteristics corresponding to the child nodes as the observed frequency, and the characteristic vectors corresponding to the remaining user risk types as the expected frequency to traverse and calculate the second chi-square coefficient;

[0159] According to the second chi-square coefficient, the user risk types as sub-child nodes of each of the child nodes in the risk type rule tree are determined from the eliminated user risk types until all of the user risk types are in the risk type rule tree.

[0160] In a possible implementation, the calculation and construction module 210 is configured as follows:

[0161] Eliminate the user risk feature corresponding to the root node;

[0162] After elimination, the characteristic vector corresponding to each of the user risk types is expressed as the observed frequency, and the characteristic vectors corresponding to the remaining user risk types are expressed as the expected frequency to traverse and calculate the chi-square coefficient of the sub-nodes;

[0163] The user risk types with the largest chi-square coefficients of a preset number of the child nodes are selected as the child-node user risk types of the child nodes of the root node in the risk type rule tree.

[0164] In a possible implementation, the calculation and construction module 210 is configured as follows:

[0165] Calculating a deviation value between a characteristic vector representation corresponding to the root node and a characteristic vector representation corresponding to user risk features corresponding to a plurality of user risk types existing in the marketing activity;

[0166] The user risk type represented by a preset number of characteristic vectors and having the largest deviation value from the characteristic vector corresponding to the root node user risk type is selected as the child node user risk type of the child node of the root node in the risk type rule tree.

[0167] In a possible implementation, determining, from the eliminated user risk types according to the second chi-square coefficient, the sub-child node user risk type of each child node in the risk type rule tree includes:

[0168] When each of the sub-nodes is determined as the observation frequency, a preset number of user risk types with the largest second chi-square coefficients are selected as candidate user risk types;

[0169] Determining the priority of the child node according to a deviation value between the characteristic vector representation corresponding to the child node and the characteristic vector representation corresponding to the root node;

[0170] According to the priority, determining whether the candidate user risk type corresponding to the subnode is used as a candidate user risk type for other subnodes with higher priorities;

[0171] In the case where the candidate user risk type corresponding to the child node is used as the candidate user risk type of other child nodes with higher priority, the user risk type is removed from the candidate user risk types of the child node to obtain the secondary child node user risk type of the child node in the risk type rule tree.

[0172] In a possible implementation, the locking and warning module 250 is configured as follows:

[0173] Determine whether the target node has a parent node, and if the target node has a parent node, determine the parent node of the target node and the node having the same parent node as the target node as the associated node; and / or

[0174] It is determined whether the target node has a child node, and if the target node has a child node, the child node of the target node is determined as the associated node.

[0175] In a possible implementation, the device further includes: an unlocking module configured to:

[0176] In response to an unlocking operation of the locked user account by an enterprise account having unlocking authority, unlocking the corresponding locked user account;

[0177] When the locked user account is unlocked, the operations of the user account on the virtual promotion and preferential rules during the operation of the virtual business logic flow will be used as the operations of the real promotion and preferential rules in the real business logic flow of the marketing activity, so that the operations of the virtual promotion and preferential rules become the operations of the real promotion and preferential rules to generate real business data.

[0178] The present disclosure also provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of the method described in any one of the aforementioned embodiments are implemented.

[0179] The present disclosure also provides an electronic device, including:

[0180] a memory having a computer program stored thereon;

[0181] A processor is used to execute the computer program in the memory to implement the steps of the method in any one of the aforementioned embodiments.

[0182] Figure 3 The marketing activity user risk warning device 100 shown includes: a processor 1001 and a memory 1003. The processor 1001 and the memory 1003 are connected, such as through a bus 1002. Optionally, the marketing activity user risk warning device 100 may also include a communication component, which can be used for data interaction between the device 100 and other devices, such as data transmission and / or data reception. It should be noted that the communication component is not limited to one in actual scheduling, and the structure of the marketing activity user risk warning device 100 does not constitute a limitation on the embodiments of the present application.

[0183] Processor 1001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. Processor 1001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0184] The bus 1002 may include a path to transmit information between the above components. The bus 1002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus 1002 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0185] The memory 1003 may be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc-Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), magnetic disk storage medium, other magnetic storage devices, or any other medium that can be used to carry or store program code and can be read by a computer, without limitation herein.

[0186] The memory 1003 is used to store program codes for executing the embodiments of the present disclosure, and the execution is controlled by the processor 1001. The processor 1001 is used to execute the program codes stored in the memory 1003 to implement the steps shown in the embodiment of the above-mentioned marketing activity user risk warning method.

[0187] The embodiment of the present disclosure also provides a computer-readable storage medium having program code stored thereon. When the program code is executed by a processor, the steps and corresponding contents of the aforementioned embodiment of the marketing activity user risk warning method can be implemented.

[0188] The preferred embodiments of the present disclosure are described in detail above in conjunction with the accompanying drawings; however, the present disclosure is not limited to the specific details in the above embodiments; within the technical concept of the present disclosure, various changes, modifications, substitutions and variations may be made to these embodiments, and these changes, modifications, substitutions and variations all fall within the protection scope of the present disclosure.

[0189] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any appropriate manner without contradiction, and they should also be regarded as the contents disclosed in this disclosure. In order to avoid unnecessary repetition, this disclosure will not further describe various possible combinations. The technical scope of this application is not limited to the contents in the specification, and its technical scope must be determined according to the scope of the claims.

Claims

1. A marketing activity user risk warning method, characterized in that: The method comprises: According to the user risk characteristics corresponding to the multiple user risk types existing in the marketing activity, the characteristic vector corresponding to each user risk type is represented as the observed frequency, and the characteristic vectors corresponding to the remaining user risk types are represented as the expected frequency to traverse and calculate the chi-square coefficient, and according to the chi-square coefficient, a risk type rule tree for the marketing activity for the user risk type is constructed; Obtaining a warning condition corresponding to each node in the risk type rule tree, associating the warning condition with each node in the risk type rule tree in a one-to-one correspondence, and obtaining a target risk type rule tree for the user risk type; Determining whether to trigger the warning condition corresponding to any node in the target risk type rule tree according to the acquired user data participating in the marketing activity; In the case where the user data triggers the warning condition corresponding to any target node in the target risk type rule tree, the user account corresponding to the user data is guided to the virtual business logic flow corresponding to the target node, wherein the virtual business logic flow is the same as the real business logic flow business logic of the marketing activity, and the user account does not generate real business data for the operation of virtual promotion and preferential rules during the operation of the virtual business logic flow. The virtual business logic flow is a virtual environment that simulates the real business logic flow, and is used to test and evaluate user behavior without actually affecting the system status and business data; The user account corresponding to the user data is locked, and the associated node is determined from the target risk type rule tree according to the target node, and the marketing activity user risk warning is generated and issued according to the user data, the target node and the corresponding warning condition, the associated node and the corresponding warning condition.

2. The marketing activity user risk warning method according to claim 1, characterized in that: The user risk features corresponding to the multiple user risk types existing in the marketing activity are represented by the characteristic vector corresponding to each user risk type as the observed frequency, and the characteristic vectors corresponding to the remaining user risk types are represented as the expected frequency to traverse and calculate the chi-square coefficient, and according to the chi-square coefficient, a risk type rule tree for the marketing activity for the user risk type is constructed, including: According to the user risk characteristics corresponding to the multiple user risk types existing in the marketing activity, the characteristic vector corresponding to each user risk type is represented as the observed frequency, and the characteristic vectors corresponding to the remaining user risk types are represented as the expected frequency to traverse and calculate the first chi-square coefficient, and the user risk type corresponding to the observed frequency corresponding to the maximum first chi-square coefficient is used as the root node of the risk type rule tree; Determine, according to the characteristic vector representation corresponding to the root node and the characteristic vector representation corresponding to the user risk features corresponding to the plurality of user risk types existing in the marketing activity, a child node user risk type that is a child node of the root node in the risk type rule tree; Repeated execution: Eliminate the user risk type that has been determined to correspond to the node, and after elimination, respectively use the characteristic vectors of the user risk characteristics corresponding to the child nodes as the observed frequency, and the characteristic vectors corresponding to the remaining user risk types as the expected frequency to traverse and calculate the second chi-square coefficient; According to the second chi-square coefficient, the user risk types as sub-child nodes of each of the child nodes in the risk type rule tree are determined from the eliminated user risk types until all of the user risk types are in the risk type rule tree.

3. The method according to claim 2, characterized in that The determining, according to the characteristic vector representation corresponding to the root node and the characteristic vector representation corresponding to the user risk features corresponding to the plurality of user risk types existing in the marketing activity, the child node user risk type that is a child node of the root node in the risk type rule tree comprises: Eliminate the user risk feature corresponding to the root node; After elimination, the characteristic vector corresponding to each of the user risk types is expressed as the observed frequency, and the characteristic vectors corresponding to the remaining user risk types are expressed as the expected frequency to traverse and calculate the chi-square coefficient of the sub-nodes; The user risk types with the largest chi-square coefficients of a preset number of the child nodes are selected as the child-node user risk types of the child nodes of the root node in the risk type rule tree.

4. The method according to claim 2, characterized in that: The determining, according to the characteristic vector representation corresponding to the root node and the characteristic vector representation corresponding to the user risk features corresponding to the plurality of user risk types existing in the marketing activity, the child node user risk type that is a child node of the root node in the risk type rule tree comprises: Calculating a deviation value between a characteristic vector representation corresponding to the root node and a characteristic vector representation corresponding to user risk features corresponding to a plurality of user risk types existing in the marketing activity; The user risk type represented by a preset number of characteristic vectors and having the largest deviation value from the characteristic vector corresponding to the root node user risk type is selected as the child node user risk type of the child node of the root node in the risk type rule tree.

5. The method according to claim 2, characterized in that: The step of determining, from the eliminated user risk types according to the second chi-square coefficient, a sub-child node user risk type as each of the child nodes in the risk type rule tree includes: When each of the sub-nodes is determined as the observation frequency, a preset number of user risk types with the largest second chi-square coefficients are selected as candidate user risk types; Determining the priority of the child node according to a deviation value between the characteristic vector representation corresponding to the child node and the characteristic vector representation corresponding to the root node; According to the priority, determining whether the candidate user risk type corresponding to the subnode is used as a candidate user risk type for other subnodes with higher priorities; In the case where the candidate user risk type corresponding to the child node is used as the candidate user risk type of other child nodes with higher priority, the user risk type is removed from the candidate user risk types of the child node to obtain the secondary child node user risk type of the child node in the risk type rule tree.

6. The method according to claim 1, characterized in that The determining, according to the target node, an associated node from the target risk type rule tree comprises: Determine whether the target node has a parent node, and if the target node has a parent node, determine the parent node of the target node and the node having the same parent node as the target node as the associated node; and / or It is determined whether the target node has a child node, and if the target node has a child node, the child node of the target node is determined as the associated node.

7. The method according to any one of claims 1 to 6, characterized in that The method further comprises: In response to an unlocking operation of the locked user account by an enterprise account having unlocking authority, unlocking the corresponding locked user account; When the locked user account is unlocked, the operations of the user account on the virtual promotion and preferential rules during the operation of the virtual business logic flow will be used as the operations of the real promotion and preferential rules in the real business logic flow of the marketing activity, so that the operations of the virtual promotion and preferential rules become the operations of the real promotion and preferential rules to generate real business data.

8. A marketing activity user risk warning device, characterized in that: The device comprises: A calculation and construction module is configured to calculate the chi-square coefficient according to the user risk characteristics corresponding to the multiple user risk types existing in the marketing activity, with the characteristic vector corresponding to each user risk type represented as the observed frequency and the characteristic vectors corresponding to the remaining user risk types represented as the expected frequency, and to construct the risk type rule tree of the marketing activity for the user risk type according to the chi-square coefficient; an acquisition and association module, configured to acquire the warning condition corresponding to each node in the risk type rule tree, associate the warning condition with each node in the risk type rule tree in a one-to-one correspondence, and obtain a target risk type rule tree for the user risk type; a determination module configured to determine whether the warning condition corresponding to any node in the target risk type rule tree is triggered according to the acquired user data participating in the marketing activity; a guiding module configured to guide the user account corresponding to the user data to the virtual business logic flow corresponding to the target node when the user data triggers the warning condition corresponding to any target node in the target risk type rule tree, wherein the virtual business logic flow is identical to the business logic of the real business logic flow of the marketing activity, and the user account does not generate real business data for the operation of the virtual promotion and preferential rules during the operation of the virtual business logic flow, and the virtual business logic flow is a virtual environment simulating the real business logic flow, and is used to test and evaluate user behavior without actually affecting the system status and business data; The locking and warning module is configured to lock the user account corresponding to the user data, and determine the associated node from the target risk type rule tree according to the target node, and generate and issue the marketing activity user risk warning according to the user data, the target node and the corresponding warning condition, the associated node and the corresponding warning condition.

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

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

Images