A method, device and apparatus for determining the validity of a rule

Abstract

The embodiments of this specification provide a method, device and equipment for determining the validity of a rule, which relates to the field of big data technology, wherein the method includes: obtaining target rule information sets of multiple business systems; generating a rule tree for the target user based on the attribute information of the rule and the list of users hit by the rule; the target user is a user hit by the rule in the list of users hit by the rule, and the rule tree contains at least one rule that hits the target user; based on the list of users hit by the rule, determining similar rules in the rule tree; using the similar rules in the rule tree and the authenticity of the users hit by the rule, determining the validity of similar rules in the rule tree. In the embodiments of this specification, the validity of rules in multiple business systems can be determined efficiently and accurately.

Description

Technical Field

[0001] The embodiments of this specification relate to the field of big data technology, and in particular to a method, device and apparatus for determining the validity of a rule. Background Art

[0002] Due to the requirements of different rules and regulations such as external national laws and regulations, internal internal control compliance risk management, etc., financial institutions often establish different user monitoring rule systems. These rule systems are formulated by different business or technical departments from different perspectives. The systems are relatively independent of each other, but there may be certain similarities between specific rules. In the prior art, system operation and maintenance personnel usually manually search for similar rules for a single business system on a regular basis, and the same user is often monitored by multiple rules of multiple systems at the same time. After the user is hit by a rule, the system monitoring personnel generally need to verify the situation. Too many rules may cause the monitoring personnel to have too much workload, and it is unreasonable for customers to be hit by too many rules. It can be seen that the technical solutions in the prior art cannot accurately determine the validity of rules in multiple business systems, and then determine redundant rules.

[0003] Currently, no effective solution has been proposed to the above problems. Summary of the invention

[0004] The embodiments of this specification provide a method, apparatus and device for determining the validity of a rule to solve the problem in the prior art that the validity of rules in multiple business systems cannot be accurately determined, and thus redundant rules cannot be determined.

[0005] An embodiment of the present specification provides a method for determining the validity of a rule, including: obtaining a target rule information set of multiple business systems; wherein the target rule information set includes attribute information of the rule, a list of users hit by the rule, and the authenticity of the user hit by the rule, and the authenticity is used to indicate whether the user hit by the rule truly complies with the rule; generating a rule tree for the target user according to the attribute information of the rule and the list of users hit by the rule; wherein the target user is a user hit by the rule in the list of users hit by the rule, and the rule tree includes at least one rule that hits the target user; based on the list of users hit by the rule, determining similar rules in the rule tree; and determining the validity of similar rules in the rule tree using similar rules in the rule tree and the authenticity of the user hit by the rule.

[0006] The embodiments of the present specification also provide a device for determining the validity of a rule, including: an acquisition module, used to acquire a target rule information set of multiple business systems; wherein the target rule information set includes attribute information of the rule, a list of users hit by the rule, and the authenticity of the user hit by the rule, and the authenticity is used to indicate whether the user hit by the rule truly complies with the rule; a generation module, used to generate a rule tree for the target user according to the attribute information of the rule and the list of users hit by the rule; wherein the target user is a user hit by the rule in the list of users hit by the rule, and the rule tree includes at least one rule that hits the target user; a first determination module, used to determine similar rules in the rule tree based on the list of users hit by the rule; and a second determination module, used to determine the validity of similar rules in the rule tree using similar rules in the rule tree and the authenticity of the user hit by the rule.

[0007] The embodiments of the present specification also provide a method and device for determining the validity of a rule, including a processor and a memory for storing instructions executable by the processor, wherein when the processor executes the instructions, the steps of any one of the method embodiments in the embodiments of the present specification are implemented.

[0008] The embodiments of the present specification also provide a computer-readable storage medium on which computer instructions are stored. When the instructions are executed, the steps of any method embodiment in the embodiments of the present specification are implemented.

[0009] The embodiment of this specification provides a method for determining the validity of a rule, which can obtain a target rule information set of multiple business systems; wherein the target rule information set includes the attribute information of the rule, the list of users hit by the rule, and the authenticity of the user hit by the rule, and the authenticity is used to indicate whether the user hit by the rule truly complies with the rule. A rule tree of the target user can be generated according to the attribute information of the rule and the list of users hit by the rule; wherein the target user is a user hit by the rule in the list of users hit by the rule, and the rule tree contains at least one rule that hits the target user. Further, similar rules in the rule tree can be determined based on the list of users hit by the rule. Since there may be a rule in the similar rules that can be replaced, or the similar rules themselves cannot be accurately hit and can be deleted, etc., the similar rules in the rule tree and the authenticity of the users hit by the rule can be used to determine the validity of similar rules in the rule tree. Thus, the validity of rules in multiple business systems can be determined efficiently and accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The drawings described herein are used to provide a further understanding of the embodiments of this specification, constitute a part of the embodiments of this specification, and do not constitute a limitation of the embodiments of this specification. In the drawings:

[0011] Figure 1 It is a schematic diagram of the steps of a method for determining the validity of a rule provided in an embodiment of this specification;

[0012] Figure 2 is a schematic diagram of a rule tree provided according to an embodiment of this specification;

[0013] Figure 3 It is a structural schematic diagram of a device for determining the validity of a rule provided in an embodiment of this specification;

[0014] Figure 4 It is a structural diagram of a device for determining the validity of a rule provided in an embodiment of this specification. DETAILED DESCRIPTION

[0015] The principles and spirit of the embodiments of this specification will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and implement the embodiments of this specification, and are not intended to limit the scope of the embodiments of this specification in any way. On the contrary, these embodiments are provided to make the disclosure of the embodiments of this specification more thorough and complete, and to fully convey the scope of the disclosure to those skilled in the art.

[0016] Those skilled in the art know that the implementation of the embodiments of this specification can be implemented as a system, device, method or computer program product. Therefore, the embodiments disclosed in this specification can be specifically implemented in the following forms, namely: complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0017] Although the processes described below include multiple operations that occur in a particular order, it should be clear that these processes may include more or fewer operations, which may be performed sequentially or in parallel (eg, using parallel processors or a multi-threaded environment).

[0018] See also Figure 1 This embodiment can provide a method for determining the validity of a rule. The method for determining the validity of a rule can be used to efficiently and accurately determine the validity of rules in multiple business systems. The above method for determining the validity of a rule can include the following steps.

[0019] S101: Obtain target rule information sets of multiple business systems; wherein the target rule information set includes attribute information of the rule, a list of users hit by the rule, and the authenticity of the users hit by the rule, wherein the authenticity is used to indicate whether the users hit by the rule actually comply with the rule.

[0020] In this embodiment, in order to determine the validity of each rule deployed in multiple business systems, target rule information sets of multiple business systems may be obtained. The target rule information set may include attribute information of the rule, a list of users hit by the rule, and the authenticity of the users hit by the rule, where the authenticity indicates whether the users hit by the rule actually comply with the rule.

[0021] In this embodiment, the user list hit by the rule may include each user who has been hit by the rule in the past, and the user list hit by the rule may include: the hit users and the corresponding adopted rules, wherein the relationship between the rules and the hit users can be one-to-one, one-to-many or many-to-many, and the specific relationship can be determined based on actual conditions, and this specification embodiment does not limit this.

[0022] In this embodiment, the attribute information of the above rules can be used to characterize the rules and can be used to distinguish different rules. The attribute information of the above rules can include: system rules of rules (for example, the division of rules into large, medium and small categories), rule descriptions, nature of rules (for example: prompt rules, verification rules, rigid control rules, high-risk rules, etc.), etc. Of course, it can be understood that the attribute information of the above rules can also include other parameters, such as the business system described in the rules, the time when the rules were created, etc., which can be determined according to actual conditions, and this specification embodiment does not limit this.

[0023] In this embodiment, the authenticity of the user hit by the rule can be used to indicate whether the user hit by the rule actually complies with the rule. Because the user hit by the rule does not necessarily mean that the behavior or event described in the rule will actually occur, the authenticity of the user hit by the rule can be marked. The authenticity of the user hit by the rule can be represented by characters, for example: 0 represents that the user hit by the rule has not performed the behavior or event described in the rule, and 1 represents that the user hit by the rule has actually performed the behavior or event described in the rule. Of course, the way of marking the authenticity of the user hit by the rule is not limited to the above examples. The technical personnel in the relevant field may make other changes under the inspiration of the technical essence of the embodiments of this specification, but as long as the functions and effects achieved are the same or similar to the embodiments of this specification, they should be covered within the protection scope of the embodiments of this specification.

[0024] In this implementation, the target rule information set may also include other parameters, such as the time when the user is hit by the rule and the corresponding rule, etc. The specific parameters may be determined based on actual conditions, and this specification does not limit this.

[0025] S102: Generate a rule tree of a target user according to attribute information of the rule and a list of users hit by the rule; wherein the target user is a user hit by the rule in the list of users hit by the rule, and the rule tree contains at least one rule that hits the target user.

[0026] In this embodiment, a rule tree of a target user may be generated according to attribute information of the rule and a list of users hit by the rule; wherein the target user is a user hit by the rule in the list of users hit by the rule, and the rule tree contains at least one rule that hits the target user.

[0027] In this embodiment, the target user may be multiple, and the target user may be any user hit by the rule in the list of users hit by the rule. In some embodiments, a rule tree may be generated for each user hit by the rule, so that the validity of the rules in various business systems can be fully determined later. Alternatively, corresponding rule trees may be generated only for a preset number of users in the list of users hit by the rule to perform the following steps, which may be determined specifically according to actual conditions, and the embodiments of this specification do not limit this.

[0028] In this embodiment, for the target user, the various rules that hit the target user can be arranged according to different businesses and the order of time of hitting, so as to generate a rule tree. The rule tree can include a timeline and various rules that hit the target user. The timeline is used to mark the time when each rule hits the target user. The various rules that hit the target user and belong to the same business system can have a connection line, which can be used to represent the association relationship between the rules.

[0029] In this embodiment, a Rule-Tree metafile of a target user is generated according to the attribute information of the rule and the list of users hit by the rule, and visualization is performed according to the Rule-Tree metafile of the target user, thereby generating a Rule-Tree.

[0030] In an example scenario, the generated rule tree can be Figure 2 As shown in , the time axis represents the time from the first triggering of the rule by user A to the end of the above preset time period (including: time period 1, time period 2, time period 3, time period 4). Figure 2 The middle dotted line represents the division of time periods, which can be divided by month by default. Dictionary items can be used to select (half year, quarter, month, ten days), which can be selected according to actual needs, and this specification embodiment does not limit this.

[0031] In this scenario example, each circle represents a rule system corresponding to a business system. Different shading patterns in the circle represent detailed classifications within the rule system, such as different major / medium / minor categories. Only those belonging to the same category have connections.

[0032] S103: Determine similar rules in the rule tree based on the user list hit by the rule.

[0033] In this embodiment, in order to determine the validity of the rules and thus determine the redundant rules, similar rules in the rule tree may be determined based on the user list hit by the rules.

[0034] In this embodiment, it can be considered that similar users will be hit by similar rules, and similar rules will hit similar users. Users in the same cluster come from the fact that they have the same potential attributes. If two rules can hit users in the same cluster, it means that the two rules have potential similarities. Based on this principle, similar rules in the rule tree can be determined.

[0035] In this embodiment, it is also possible to determine whether two rules are similar rules based on the semantic similarity between the rule descriptions of different rules. This can be determined specifically based on actual conditions, and this embodiment of the specification does not limit this.

[0036] S104: Determine the validity of similar rules in the rule tree by using similar rules in the rule tree and the authenticity of users hit by the rules.

[0037] In this embodiment, since there may be a rule in the similar rules that can be replaced, or the similar rule itself cannot be accurately hit and can be deleted, the similar rules in the rule tree and the authenticity of the users hit by the rules can be used to determine the validity of the similar rules in the rule tree.

[0038] In this embodiment, the rule in which the authenticity of the user hit by the rule is negatively evaluated among similar rules can be regarded as a rule with poor effectiveness, and the rule can be deleted as a redundant rule. In the case where similar rules can be accurately hit, it can be determined whether the hit effects of the two are similar, that is, whether they can be replaced with each other. If so, the rule with relatively poor effectiveness can be regarded as a redundant rule. Of course, the method for determining the effectiveness of similar rules is not limited to the above examples. Under the inspiration of the technical essence of the embodiments of this specification, technical personnel in the relevant field may also make other changes, but as long as the functions and effects achieved are the same or similar to the embodiments of this specification, they should be covered within the protection scope of the embodiments of this specification.

[0039] From the above description, it can be seen that the embodiment of this specification realizes the following technical effects: target rule information sets of multiple business systems can be obtained; wherein the target rule information set contains attribute information of the rule, a list of users hit by the rule, and the authenticity of the user hit by the rule, and the authenticity is used to indicate whether the user hit by the rule truly complies with the rule. A rule tree of the target user can be generated according to the attribute information of the rule and the list of users hit by the rule; wherein the target user is a user hit by the rule in the list of users hit by the rule, and the rule tree contains at least one rule that hits the target user. Further, similar rules in the rule tree can be determined based on the list of users hit by the rule. Since there may be a rule in the similar rules that can be replaced, or the similar rules themselves cannot be accurately hit and can be deleted, etc., the similar rules in the rule tree and the authenticity of the users hit by the rule can be used to determine the validity of similar rules in the rule tree. Thus, the validity of rules in multiple business systems can be determined efficiently and accurately.

[0040] In one embodiment, obtaining a target rule information set for multiple business systems within a preset time period may include: collecting a first rule information set from the multiple business systems; wherein the first rule information set contains attribute information of each rule deployed in the multiple business systems and a list of users hit by the rules. Determine the authenticity of the users hit by the rules based on the behavior data of the users hit by the rules within a preset time period. Furthermore, the authenticity of the users hit by the rules may be added to the first rule information set to obtain a second rule information set. Determine the labels of each rule; wherein the labels of the rules are used to represent the characteristics and classification of the rules; and add the labels of each rule to the second rule information set to obtain the target rule information set.

[0041] In this embodiment, the rule data deployed in different business systems are collected by system docking to obtain the first rule information set. Because the hit user does not necessarily mean that the behavior or event described in the rule (for example, fraud, breach of contract, etc.) will actually occur, the authenticity of each user hit by the rule needs to be marked.

[0042] In this embodiment, the behavior data of the user hit by the rule within a preset time period can be obtained, wherein the behavior data can be the behavior data in the business system to which the rule that hits the user belongs. If one of the following two situations is met, the authenticity of the user is marked as 1: (1) The user is hit by the rule under the business system within the target time period and the event described by the rule actually occurs; (2) The user is hit by the deterministic rule (for example: rigid control rule, high-risk control rule) under the business system within the target time period, and control is performed in advance before the event occurs. In this case, the risk is already very high, which can be considered equivalent to the event described by the rule actually occurring. Otherwise, the authenticity of the user is marked as 0.

[0043] In this embodiment, the preset time period may be a pre-set time period, which may be one year, half a year, three months, etc. The specific time period may be determined according to actual conditions, and this specification does not limit this.

[0044] In this embodiment, a rule tag library can be established, and the rules can be tagged with rule tags according to the tag list in the tag library. A unified tag list system can be defined first, so that the rules can be classified and processed later. In some embodiments, the tag list system can be divided into large, medium and small categories. The large categories are market risk, credit risk, and operational risk; the medium categories are market risk (bonds, funds, stocks, etc.), credit risk (legal persons, institutions, individuals, etc.). Operational risk (anti-money laundering, anti-fraud, employee operations); small categories are credit risk-legal persons-real estate, etc.

[0045] In this implementation, since the description of the rule itself is relatively complex, some simple tags can be used to describe the main features of the rule. The rule tag can be used to characterize the main features of the rule. For example, if a rule is to screen out companies whose loan funds flow to real estate in the past 180 days, the corresponding rule tag can be credit risk-legal person-real estate.

[0046] In one implementation, the attribute information of a rule may include: a system rule of the rule, a rule description, and a property of the rule.

[0047] In this embodiment, the attribute information of the rule can be used to characterize the rule and can be used to distinguish different rules. The attribute information of the rule can include: the system rules of the rule (for example, the division of the rule into large, medium and small categories), the rule description, the nature of the rule (for example: prompting rules, verification rules, rigid control rules, high-risk rules, etc.). Of course, it can be understood that the attribute information of the rule can also include other parameters, such as the business system described in the rule, the time when the rule was created, etc., which can be determined according to the actual situation, and the embodiments of this specification do not limit this.

[0048] In one embodiment, before generating the rule tree of the target user according to the attribute information of the rule and the list of users hit by the rule, it may also include: obtaining the feature vector of each rule using word2Vec according to the rule label in the target rule information set. Generate a training sample set for each business system according to the feature vector of each rule, the list of users hit by the rule and the authenticity of the users hit by the rule; wherein the training sample set contains multiple groups of training samples, each group of training samples contains at least one feature vector of a rule and the authenticity of the corresponding users hit by the rule. Furthermore, based on the training sample sets of each business system, a prediction model of each business system can be obtained by training using a Tree-LSTM model; wherein the prediction model is used to predict the authenticity of the users hit by the rule according to the feature vector of the input rule.

[0049] In this embodiment, since the rule description is a text description, it is necessary to digitize the rule, and the rule description can be processed by word segmentation, stop word removal, etc. to obtain the feature vector of the rule. It is also possible to use word2Vec according to the rule label to obtain the feature vector of each rule. Among them, Word2Vec is a tool for generating word vectors, and word vectors are closely related to language models.

[0050] In this embodiment, since different business systems have different rule systems, a corresponding prediction model can be generated for each business system, and corresponding training sample sets can be generated for each business system.

[0051] In this embodiment, a training sample information set can be generated based on the above target rule information set. The target sample information set records the relevant information of the users that the rules have historically hit. For example, the target rule information set records three pieces of information: rule A hits user C at time B, with an authenticity of 1; rule A hits user M at time N, with an authenticity of 0; and rule H hits user C at time T, with an authenticity of 1. Then, three groups of training samples can be generated accordingly: (feature vector of user C, rule A, 1), (feature vector of user M, rule A, 0), and (feature vector of user C, rule H, 1). Of course, the method of generating training samples is not limited to the above examples. Technicians in the relevant field may make other changes based on the technical essence of the embodiments of this specification. However, as long as the functions and effects achieved are the same or similar to those of the embodiments of this specification, they should be covered within the protection scope of the embodiments of this specification.

[0052] In this embodiment, since the structure of the rule tree is similar to the structure of the Tree-LSTM model, and the rule tree has the properties of a time series, the Tree-LSTM model is a time series model. Therefore, the prediction model of each business system can be obtained by training the Tree-LSTM model. In the above-mentioned Tree-LSTM model tree structure long short-term memory network model, LSTM includes a forget gate, an input gate, and an output gate. The forget gate represents whether the rule will affect the event described in the user's rule, the input gate represents the degree of influence of the rule itself, and the output gate represents the degree of influence of its own influence on the subsequent time period.

[0053] In one embodiment, based on the attribute information of the rule and the list of users hit by the rule, determining similar rules in the rule tree includes: clustering users in the list of users hit by the rule to determine a similar rule information set across business systems; wherein the rules of different business systems hit by users under each cluster belong to the similar rule information set across business systems; and determining similar rules in different business systems in the rule tree that hit the target user in the same time period based on the similar rule information set across business systems. Correspondingly, determining the validity of similar rules in the rule tree using similar rules in the rule tree and the authenticity of users hit by the rules may include: determining the target business system to which the similar rules belong, and obtaining feature vectors of rules in the rule tree that belong to the target business system; wherein the feature vectors of rules that belong to the target business system are arranged according to the time when they hit the user. Further, the validity of similar rules in different business systems in the rule tree in the same time period may be determined based on the feature vectors of rules in the rule tree that belong to the target business system and the authenticity of users hit by the rules using the prediction model of the target business system.

[0054] In this implementation, the rules in the target rule information set may be annotated for similarity, and may be divided into three types of similar rule information sets: similar rule information sets in the same business system, similar rule information sets across business systems, and semantically similar rule information sets.

[0055] In this embodiment, it can be considered that similar users will be hit by similar rules, and similar rules will hit similar users. Users in the same cluster come from the fact that they have the same potential attributes. If two rules can hit users in the same cluster, it means that the two rules have potential similarities. Based on this principle, similar rules in the rule tree can be determined.

[0056] In this implementation, the hit users may be clustered, and the rules in different rule systems that hit users in the same cluster may be classified as cross-business system similar rules.

[0057] In this embodiment, in order to determine the redundant rules in different business systems, since the rule tree has a time sequence feature, the rule tree of the target user can be horizontally evaluated, and similar rules in different business systems in the rule tree of the target user that hit the target user in the same time period can be determined according to the cross-business system similar rule information set (for example: Figure 2 Rules S13, S33).

[0058] In this embodiment, since similar rules across business systems belong to different business systems, their validity needs to be determined separately. The target business system to which the similar rules belong can be determined, and the target business system can be the business system to which any one of the at least two similar rules belongs, and the method for determining the validity of different rules is the same.

[0059] In one embodiment, according to the feature vector of the rule belonging to the target business system in the rule tree and the authenticity of the user hit by the rule, the validity of similar rules in different business systems in the rule tree in the same time period is determined by using the prediction model of the target business system, which may include: based on the feature vector of the rule belonging to the target business system in the rule tree, using the prediction model of the target business system, obtaining the predicted value of the authenticity of the user hit by the rule in the target business system; determining the accuracy of the prediction before removing similar rules according to the actual value and predicted value of the authenticity of the user hit by the rule in the target business system; based on the feature vector of the rule after similar rules are removed in the target business system, using the prediction model of the target business system, obtaining the predicted value of the authenticity of the user hit by the rule in the target business system after similar rules are removed; further, the accuracy of the prediction after similar rules are removed may be determined according to the actual value and predicted value of the authenticity of the user hit by the rule in the target business system after similar rules are removed. The difference between the accuracy of the prediction before similar rules are removed and the accuracy of the prediction after similar rules are removed is used as the validity of similar rules in different business systems in the rule tree in the same time period.

[0060] In this embodiment, since the rule tree may contain rules from multiple business systems, the rules under the same business system in the rule tree may be taken as a whole to generate a feature vector, wherein the feature vectors of each rule in the feature vector are arranged in chronological order of hitting the target user.

[0061] In this embodiment, the prediction value of the authenticity of the user hit by the rule in the target business system can be obtained based on the feature vector of the rule belonging to the target business system in the rule tree and using the prediction model of the target business system. The information recorded in the target rule information set can be the actual value of the authenticity of the user hit by the rule in the target business system. Since the target user may be hit by multiple rules in the target business system, the overall authenticity can be determined based on the authenticity of the user hit by each rule. For example: the target user is hit by rules A, B, and C in different time periods, and the corresponding authenticity is 1, 0, and 1 respectively. Since the target user is truly hit as a whole, the overall authenticity is 1.

[0062] In this embodiment, the ROC value can be calculated according to the actual value and predicted value of the authenticity of the user hit by the rule in the target business system, and the calculated ROC value is used as the accuracy of the prediction before removing similar rules. ROC stands for receiver operator characteristic, which is a curve in a two-dimensional coordinate axis, with the horizontal axis being the false positive rate and the vertical axis being the true positive rate.

[0063] In this embodiment, the feature vector of the similar rule under the target business system in the above feature vector can be removed, and the feature vector after removal can be used to reuse the prediction model of the target business system to obtain the predicted value of the authenticity of the user hit by the rule in the target business system after removing the similar rule. Furthermore, the overall authenticity after removing the similar rule can be determined based on the authenticity of the user hit by the rule recorded in the target rule information set. For example: the target user is hit by rules A, B, and C in different time periods, and the corresponding authenticity is 1, 0, and 1 respectively, and the similar rule is C. Since the target user is truly hit as a whole, the overall authenticity is 1.

[0064] In this embodiment, the ROC value can be calculated according to the actual value and predicted value of the authenticity of the users hit by the rules in the target business system after removing similar rules, and the ROC value is used as the accuracy of the prediction after removing similar rules.

[0065] In this embodiment, the change in accuracy can be determined based on the calculation result of the predicted ROC before removing similar rules minus the predicted ROC after removing similar rules. If the calculation result is positive, it means that this type of rule has a deviation in the evaluation of whether the user will have the corresponding time, and it is recorded in the negative rule set; if the calculation result is negative, it means that this type of rule is effective in evaluating whether the user will have the corresponding event, and the larger the absolute value of the negative value, the greater the impact of removing this rule on the accuracy rate and needs to be retained. At the same time, considering that the three rules before and after this rule may have a mutual influence relationship, the three rules before and after this rule are retained.

[0066] In one embodiment, based on the attribute information of the rule and the list of users hit by the rule, determining similar rules in the rule tree may include: clustering the users in the list of users hit by the rule to determine a set of similar rules information for the same business system; wherein the rules of the same business system hit by the users in each cluster belong to the set of similar rules information for the same business system. Based on the set of similar rules information for the same business system, similar rules in the same business system in the rule tree are determined, and the same rules appearing at different times in the rule tree are regarded as similar rules.

[0067] In this implementation, the hit users may be clustered, and the rules that hit users in the same cluster in the same rule system may be classified as similar rules of the same business system, thereby obtaining an information set of similar rules of the same business system.

[0068] In this implementation, in order to determine the redundant rules in the same business system, the rule tree of the target user may be vertically evaluated, which may specifically include two categories:

[0069] The first category: according to the similar rule information set of the same business system, determine the similar rules in the same business system in the rule tree of the target user (for example: Figure 2 Rules S32, S33 in Figure 2For example, the feature vectors of rules S31, S33, S32, and S34 can be determined, and the prediction model of the corresponding target business system can be used to obtain the predicted value of the authenticity of the users hit by the rules in the target business system. ROC1 is calculated based on the predicted value and true value of the authenticity of the users hit by the rules S31, S33, S32, and S34 in the target business system. Rules S33 and S32 in the feature vector are removed to obtain a new feature vector. The prediction model of the corresponding target business system is used to obtain the predicted value of the authenticity of the users hit by the rules in the target business system after removing similar rules S33 and S32. ROC2 is calculated based on the predicted value and true value of the authenticity of the users hit by the rules S31 and S34 in the target business system. If ROC1-ROC2 is positive, it means that there is a deviation in the evaluation of whether the corresponding event will occur to the user by rules S33 and S32, and it is recorded in the negative rule set. If the ROC1-ROC2 value is negative, it means that rules S33 and S32 are effective in evaluating whether the corresponding event will occur to the user, and the larger the absolute value of the negative value, the higher the effectiveness. The three rules before and after rules S33 and S32 on the timeline are recorded.

[0070] Category 2: For the same business system, remove the same rule that appears in different time periods in turn (reference Figure 2 In rule S21), the same method as the first category is used to determine the degree of influence of the rule on the accuracy of the user's authenticity prediction in different time periods. The goal is to record in which time series the rule is most effective. If it is determined to be effective, the three rules before and after the rule on the timeline are recorded.

[0071] In one embodiment, determining similar rules in the rule tree based on the attribute information of the rules and the user list of rule hits can include: determining the semantic similarity between the rules using the LDA model based on the rule labels in the target rule information set; taking the rules whose semantic similarity is greater than or equal to a preset threshold as semantic similarity rules to obtain a semantic similarity rule set; and determining the semantic similarity rules in the rule tree based on the semantic similarity rule set.

[0072] In this embodiment, the semantic similarity of rules can be analyzed using the LDA model according to the rule labels. Semantically similar rules are classified into the same rule set. Among them, LDA (Latent Dirichlet Allocation) is a document topic generation model, also known as a three-layer Bayesian probability model, which includes a three-layer structure of words, topics and documents. LDA is an unsupervised machine learning technology that can be used to identify hidden topic information in large-scale document collections or corpuses.

[0073] In this embodiment, the semantic similarity rule can be used to determine the similar rules that can be replaced in the rule number according to the effective rules obtained in the horizontal evaluation and the vertical evaluation (for example: Figure 2 In one embodiment, the rules S13 and S31 in Figure 2 S13 and S31 in the above example are swapped. Please use the corresponding system prediction model to determine the predicted value of authenticity and compare the accuracy of the authenticity prediction before and after the replacement rule. Figure 2 The first circle on the left and the first circle on the right should be calculated separately). Take the absolute value of the predicted ROC after replacement minus the predicted ROC before replacement. If the absolute value is less than the preset threshold, it proves that the replacement is effective, and the effective replacement rule is recorded. At the same time, according to the accuracy change (predicted ROC after replacement minus predicted ROC before replacement), the larger the value, the more effective the replaced rule is.

[0074] In this implementation, the preset threshold may be a value greater than 0, such as 0.1, 0.52, etc. The specific value may be determined based on actual conditions, and the embodiments of this specification do not limit this.

[0075] In the embodiment of the present specification, after repeating the above steps for a preset number of target users, a rule validity evaluation result can be obtained, which may include a positive rule set and a negative rule set. The positive rule set contains valid rules, and the negative rule set contains invalid rules.

[0076] In the embodiments of this specification, through the horizontal evaluation of rules, it can be obtained whether the same type of rules across systems are effective for the prediction results. Through the first type of vertical evaluation of rules, it can be obtained whether the same type of rules in the same system are effective for the prediction results, which can be used to determine whether the rule system needs to be adjusted on this type of system to remove invalid rule branches. Through the second type of vertical evaluation of rules, the effectiveness of the rules in different time periods can be obtained, which can be used to determine at which time point the rules are more effective for the prediction results. Through the rule substitution evaluation, the degree of substitutability of similar rules can be obtained, which can be used to streamline the rules to reduce the troubleshooting pressure of front-line employees, and also reduce the system operation pressure of running similar rules repeatedly.

[0077] In the embodiments of this specification, for an effective rule, the three rules that appear before and after it can be counted. By sorting them according to the frequency of occurrence, three types of optimization combinations can be generated:

[0078] Predictive rule set: The TOPN_1 rule that appears most frequently before the effective rule is used as the predictive rule set. This type of rule set may not have a great effect on the prediction results, but the appearance of this type of rule indicates the appearance of effective rules, indicating that the risk event status of the customer is about to change. For similar rules of the same type, the TOPN_2 rule that appears most frequently in the similar rule set is taken as the representative rule.

[0079] Valid rule set: the valid rules generated in the previous steps. Classify similar rules and take the TOPN_3 rules that have the greatest impact on the prediction effect as the representative rules of the similar rules.

[0080] Post-rule set: For the TOPN_4 rules that appear most frequently after the effective rules. Although the description of this type of rules is different from the effective rules in terms of the meaning of the rules, the actual effects are similar. Take the TOPN_5 that appears most frequently in the similar rule set as the representative rule and remove it.

[0081] The representative rules in the prediction rule set and the representative rules in the effective rule set are retained, and the representative rules in the subsequent rule set are eliminated, so that the optimal combination of rules can be obtained.

[0082] In the embodiments of this specification, rule simplification can be achieved through a global evaluation of the effectiveness of the rules and the combination effect. Since most of the rules are calculated in a systematic way, the pressure of system resource consumption can be reduced after simplification, the amount of data storage can be reduced to reduce the demand for storage resources, and intensive management of system resources can be achieved. In addition, the effectiveness of the similarity of different rules can be evaluated from a global cross-system perspective, and the decision-making level of the auxiliary organization can better analyze the rationality and effectiveness of the rule settings, providing a reference for decision-making. Furthermore, the user's workload can be reduced. Only a few parameters need to be set, and the system can realize the full process operation, including data fusion, model generation, and rule combination, reducing the difficulty of user operation.

[0083] Based on the same inventive concept, a device for determining the validity of a rule is also provided in the embodiments of this specification, as described in the following embodiments. Since the principle of solving the problem by the device for determining the validity of a rule is similar to that of the method for determining the validity of a rule, the implementation of the device for determining the validity of a rule can refer to the implementation of the method for determining the validity of a rule, and the repeated parts will not be repeated. As used below, the term "unit" or "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceived. Figure 3 is a structural block diagram of a device for determining the validity of a rule according to an embodiment of this specification, such as Figure 3 As shown, it may include: an acquisition module 301, a generation module 302, a first determination module 303, and a second determination module 304. The structure is described below.

[0084] The acquisition module 301 can be used to acquire target rule information sets of multiple business systems; wherein the target rule information set includes attribute information of the rule, a list of users hit by the rule, and the authenticity of the users hit by the rule, wherein the authenticity is used to indicate whether the users hit by the rule actually comply with the rule;

[0085] The generating module 302 may be used to generate a rule tree of a target user according to the attribute information of the rule and the list of users hit by the rule; wherein the target user is a user hit by the rule in the list of users hit by the rule, and the rule tree contains at least one rule that hits the target user;

[0086] A first determination module 303 may be used to determine similar rules in the rule tree based on a list of users hit by the rule;

[0087] The second determination module 304 may be used to determine the validity of similar rules in the rule tree by using similar rules in the rule tree and the authenticity of users hit by the rules.

[0088] The embodiment of this specification also provides an electronic device, which can be specifically referred to in Figure 4 The electronic device composition structure diagram of the method for determining the validity of rules provided by the embodiment of this specification is shown, and the electronic device may specifically include an input device 41, a processor 42, and a memory 43. Among them, the input device 41 can be specifically used to input the target rule information set of multiple business systems. The processor 42 can be specifically used to obtain the target rule information set of multiple business systems; wherein the target rule information set contains the attribute information of the rule, the list of users hit by the rule, and the authenticity of the user hit by the rule, and the authenticity is used to indicate whether the user hit by the rule really complies with the rule; according to the attribute information of the rule and the list of users hit by the rule, a rule tree of the target user is generated; wherein the target user is a user hit by the rule in the list of users hit by the rule, and the rule tree contains at least one rule that hits the target user; based on the list of users hit by the rule, similar rules in the rule tree are determined; and the validity of similar rules in the rule tree is determined by using similar rules in the rule tree and the authenticity of the user hit by the rule. The memory 43 can be specifically used to store parameters such as the validity of the rule tree and similar rules.

[0089] In this embodiment, the input device may specifically be one of the main devices for information exchange between the user and the computer system. The input device may include a keyboard, a mouse, a camera, a scanner, a light pen, a handwriting input board, a voice input device, etc.; the input device is used to input the original data and the program for processing these numbers into the computer. The input device can also obtain and receive data transmitted from other modules, units, and devices. The processor can be implemented in any appropriate manner. For example, the processor can take the form of a computer-readable medium, a logic gate, a switch, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), a programmable logic controller, and an embedded microcontroller, etc., such as a microprocessor or a processor and a computer-readable program code (such as software or firmware) that can be executed by the (micro) processor. The memory may specifically be a memory device used to store information in modern information technology. The memory may include multiple levels. In a digital system, anything that can store binary data can be a memory; in an integrated circuit, a circuit with a storage function without a physical form is also called a memory, such as a RAM, a FIFO, etc.; in a system, a storage device with a physical form is also called a memory, such as a memory stick, a TF card, etc.

[0090] In this embodiment, the functions and effects specifically realized by the electronic device can be explained in comparison with other embodiments and will not be described in detail here.

[0091] The embodiment of the present specification also provides a computer storage medium based on a method for determining the validity of a rule, wherein the computer storage medium stores computer program instructions, and when the computer program instructions are executed, the following can be achieved: obtaining a target rule information set of multiple business systems; wherein the target rule information set includes attribute information of the rule, a list of users hit by the rule, and the authenticity of the users hit by the rule, and the authenticity is used to indicate whether the users hit by the rule truly comply with the rule; generating a rule tree for the target user based on the attribute information of the rule and the list of users hit by the rule; wherein the target user is a user hit by the rule in the list of users hit by the rule, and the rule tree includes at least one rule that hits the target user; determining similar rules in the rule tree based on the list of users hit by the rule; and determining the validity of similar rules in the rule tree using similar rules in the rule tree and the authenticity of users hit by the rule.

[0092] In this embodiment, the storage medium includes, but is not limited to, a random access memory (RAM), a read-only memory (ROM), a cache, a hard disk (HDD), or a memory card. The memory may be used to store computer program instructions. The network communication unit may be an interface for network connection communication set in accordance with the standard specified by the communication protocol.

[0093] In this embodiment, the functions and effects specifically implemented by the program instructions stored in the computer storage medium can be explained in comparison with other embodiments and will not be repeated here.

[0094] Obviously, those skilled in the art should understand that the modules or steps of the above-mentioned embodiments of this specification can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order from that here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. In this way, the embodiments of this specification are not limited to any specific combination of hardware and software.

[0095] Although the embodiments of this specification provide method operation steps as described in the above embodiments or flowcharts, more or fewer operation steps may be included in the method based on routine or no creative labor. In the steps where there is no necessary causal relationship logically, the execution order of these steps is not limited to the execution order provided by the embodiments of this specification. When the device or terminal product of the method described in practice is executed, it can be executed in sequence or in parallel according to the method shown in the embodiments or drawings (for example, a parallel processor or a multi-threaded processing environment).

[0096] It should be understood that the above description is for illustration and not for limitation. By reading the above description, many embodiments and many applications beyond the examples provided will be apparent to those skilled in the art. Therefore, the scope of the embodiments of this specification should not be determined with reference to the above description, but should be determined with reference to the full scope of the aforementioned claims and the equivalents possessed by these claims.

[0097] The above is only a preferred embodiment of the embodiment of this specification, and is not intended to limit the embodiment of this specification. For those skilled in the art, the embodiment of this specification may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the embodiment of this specification shall be included in the protection scope of the embodiment of this specification.

Claims

1. A method for determining the validity of a rule, characterized in that: include: Obtain target rule information sets of multiple business systems; wherein the target rule information set includes attribute information of the rule, a list of users hit by the rule, and the authenticity of the users hit by the rule, wherein the authenticity is used to indicate whether the users hit by the rule actually comply with the rule; Generate a rule tree of a target user according to the attribute information of the rule and the list of users hit by the rule; wherein the target user is a user hit by the rule in the list of users hit by the rule, and the rule tree contains at least one rule that hits the target user; Determining similar rules in the rule tree based on the user list of rule hits; Determine the validity of similar rules in the rule tree by using similar rules in the rule tree and the authenticity of users hit by the rules; The target rule information set of multiple business systems within a preset time period is obtained, including: Collecting a first rule information set from the multiple business systems; wherein the first rule information set includes attribute information of each rule deployed in the multiple business systems and a list of users hit by the rule; Determine the authenticity of the user hit by the rule based on the behavior data of the user hit by the rule within a preset time period; Adding the authenticity of the user hit by the rule to the first rule information set to obtain a second rule information set; Determine the label of each rule; wherein the label of the rule is used to represent the characteristics and classification of the rule; The labels of the respective rules are added to the second rule information set to obtain the target rule information set.

2. The method according to claim 1, characterized in that The attribute information of the rule includes: the system rules of the rule, the rule description, and the nature of the rule.

3. The method according to claim 1, characterized in that Before generating a rule tree of a target user according to the attribute information of the rule and the list of users hit by the rule, the method further includes: According to the rule labels in the target rule information set, word2Vec is used to obtain the feature vectors of each rule; Generate a training sample set for each business system according to the feature vectors of each rule, the list of users hit by the rule and the authenticity of the users hit by the rule; wherein the training sample set includes multiple groups of training samples, each group of training samples includes at least one feature vector of a rule and the authenticity of the corresponding users hit by the rule; Based on the training sample sets of the various business systems, the prediction models of the various business systems are trained using the Tree-LSTM model; wherein the prediction model is used to predict the authenticity of the users hit by the rules according to the feature vectors of the input rules.

4. The method according to claim 3, characterized in that Determining similar rules in the rule tree based on the attribute information of the rule and the user list hit by the rule includes: Clustering the users in the user list hit by the rule to determine a cross-business system similar rule information set; wherein the rules of different business systems hit by users in each cluster belong to the cross-business system similar rule information set; Determine similar rules that hit the target user in the same time period in different business systems in the rule tree according to the cross-business system similar rule information set; Correspondingly, using similar rules in the rule tree and the authenticity of users hit by the rules, determining the validity of similar rules in the rule tree includes: Determine the target business system to which the similarity rule belongs; Acquire feature vectors of rules belonging to the target business system in the rule tree; wherein the feature vectors of rules belonging to the target business system are arranged according to the time when they hit the user; According to the feature vector of the rule belonging to the target business system in the rule tree and the authenticity of the user hit by the rule, the prediction model of the target business system is used to determine the validity of similar rules in the rule tree in different business systems in the same time period.

5. The method according to claim 4, characterized in that According to the feature vector of the rule belonging to the target business system in the rule tree and the authenticity of the user hit by the rule, the validity of similar rules in different business systems in the rule tree in the same time period is determined by using the prediction model of the target business system, including: Based on the feature vector of the rule belonging to the target business system in the rule tree, using the prediction model of the target business system, obtaining a prediction value of the authenticity of the user hit by the rule in the target business system; Determining the accuracy of the prediction before removing similar rules based on the actual value and the predicted value of the authenticity of the user hit by the rule in the target business system; Based on the feature vector of the rule after similar rules are removed in the target business system, using the prediction model of the target business system, obtain the prediction value of the authenticity of the user hit by the rule in the target business system after similar rules are removed; Determine the accuracy of the prediction after removing the similar rules according to the actual value and the predicted value of the authenticity of the users hit by the rules in the target business system after removing the similar rules; The difference between the prediction accuracy before removing the similar rules and the prediction accuracy after removing the similar rules is used as the validity of the similar rules in the rule tree in different business systems in the same time period.

6. The method according to claim 3, characterized in that Determining similar rules in the rule tree based on the attribute information of the rule and the user list hit by the rule includes: Clustering the users in the user list hit by the rule to determine a set of similar rule information for the same business system; wherein the rules of the same business system hit by the users in each cluster belong to the set of similar rule information for the same business system; Determining similar rules in the same business system in the rule tree according to the similar rule information set of the same business system; The same rule appearing at different times in the rule tree is regarded as a similar rule.

7. The method according to claim 3, characterized in that Determining similar rules in the rule tree based on the attribute information of the rule and the user list hit by the rule includes: Determine the semantic similarity between the rules using the LDA model according to the rule labels in the target rule information set; The rules whose semantic similarity is greater than or equal to a preset threshold are taken as semantic similarity rules to obtain a semantic similarity rule set; According to the semantic similarity rule set, semantic similarity rules in the rule tree are determined.

8. A device for determining the validity of a rule, characterized in that: include: An acquisition module, used to acquire target rule information sets of multiple business systems; wherein the target rule information set includes attribute information of the rule, a list of users hit by the rule, and the authenticity of the users hit by the rule, wherein the authenticity is used to indicate whether the users hit by the rule actually comply with the rule; A generating module, configured to generate a rule tree of a target user according to the attribute information of the rule and the list of users hit by the rule; wherein the target user is a user hit by the rule in the list of users hit by the rule, and the rule tree contains at least one rule that hits the target user; A first determination module, configured to determine similar rules in the rule tree based on a user list of rule hits; A second determination module, used to determine the validity of similar rules in the rule tree by using similar rules in the rule tree and the authenticity of users hit by the rules; The target rule information set of multiple business systems within a preset time period is obtained, including: Collecting a first rule information set from the multiple business systems; wherein the first rule information set includes attribute information of each rule deployed in the multiple business systems and a list of users hit by the rule; Determine the authenticity of the user hit by the rule based on the behavior data of the user hit by the rule within a preset time period; Adding the authenticity of the user hit by the rule to the first rule information set to obtain a second rule information set; Determine the label of each rule; wherein the label of the rule is used to represent the characteristics and classification of the rule; The labels of the respective rules are added to the second rule information set to obtain the target rule information set.

9. A device for determining the validity of a rule, characterized in that: The method comprises a processor and a memory for storing processor-executable instructions, wherein the processor implements the steps of the method according to any one of claims 1 to 7 when executing the instructions.

10. A computer-readable storage medium, characterized in that: Computer instructions are stored thereon, and when the instructions are executed, the steps of the method according to any one of claims 1 to 7 are implemented.

Images