A decision tree algorithm model training method and a page configuration conflict resolution method

Abstract

This invention provides a training method for a decision tree algorithm model and a page configuration conflict resolution method. By pre-obtaining the effect scores of sample resolution schemes and iteratively adjusting model parameters based on reducing the difference between the predicted effect scores and the sample effect scores, the decision tree algorithm model can learn the core logic of conflict resolution, reduce the deviation between the predicted resolution scheme and the optimal solution of the resolution scheme, and improve the efficiency and accuracy of conflict resolution.

Description

Technical Field

[0001] This invention relates to the field of Internet technology, and in particular to a training method for a decision tree algorithm model and a method for resolving page configuration conflicts. Background Technology

[0002] The page involves configuration items across multiple dimensions. Taking the payment page as an example, it includes options such as package selection, payment method, and add-on purchases. The complexity and diversity of these configuration items increase the risk of logical conflicts during configuration, directly impacting user experience and transaction success rates. For instance, configuring a "new user's first month 6 yuan" package while simultaneously enabling "continuous monthly subscription by default" and "no vouchers supported" may lead to user complaints; adding-on purchases exceeding half the screen display threshold may not automatically trigger the collapsing rule, negatively affecting user experience.

[0003] Resolving page configuration conflicts relies on conflict detection. In the traditional model, conflict detection is mainly done manually, and conflicts are then manually resolved after detection. This method is not only inefficient but also has a high error rate. As business scales up and configuration complexity increases, manual detection and resolution can no longer meet the needs. Summary of the Invention

[0004] The purpose of this invention is to provide a training method for a decision tree algorithm model and a page configuration conflict resolution method, so as to improve the efficiency and accuracy of conflict resolution. The specific technical solution is as follows:

[0005] In a first aspect of this invention, a method for training a decision tree algorithm model is provided, the method comprising:

[0006] Obtain the sample effect scores of conflicting sample configuration items and sample resolution schemes; wherein, the sample resolution schemes are used to resolve the conflicts existing in the sample configuration items;

[0007] The sample configuration items are input into the original decision tree algorithm model to obtain the resolution scheme output by the original decision tree algorithm model, which is then used as the predicted resolution scheme.

[0008] Based on the sample configuration items and the prediction resolution scheme, a value function is constructed to obtain the effect score of the prediction resolution scheme, which is used as the prediction effect score;

[0009] The model parameters of the original decision tree algorithm model are adjusted in the direction of reducing the difference to obtain the target decision tree algorithm model; wherein, the difference is the difference between the predicted effect score and the sample effect score.

[0010] In one possible implementation, the method further includes:

[0011] Obtain conflict association information of the conflicts existing in the sample configuration item; wherein, the conflict association information includes at least one of the conflict type, the severity score of the conflict, and the business scenario label of the sample configuration item;

[0012] The step of inputting the sample configuration items into the original decision tree algorithm model and obtaining the resolution scheme output by the original decision tree algorithm model as the predicted resolution scheme includes:

[0013] The sample configuration items and the conflict association information are input into the original decision tree algorithm model to obtain the resolution scheme output by the original decision tree algorithm model, which is then used as the predicted resolution scheme.

[0014] In one possible implementation, the method further includes:

[0015] The score of the first dimension is obtained by querying the number of sample configuration items associated with the conflict and determining the score according to the preset number range corresponding rules; wherein, the first dimension is used to characterize the conflict association range of the conflict.

[0016] The score for the second dimension is determined by the impact probability output by the impact prediction model based on the sample configuration items, and is used as the second score; wherein, the second dimension is used to characterize the degree of impact of the conflict on the user;

[0017] The score for the third dimension is determined by matching the preset risk level with the corresponding score for the type of business impact caused by the conflict; wherein, the third dimension is used to characterize the degree of business risk caused by the conflict.

[0018] The first score, the second score, and the third score are weighted and calculated to obtain the result, which is used as the severity score of the conflict.

[0019] In a second aspect of the present invention, a method for resolving page configuration conflicts is also provided, the method comprising:

[0020] A configuration knowledge graph is constructed based on the configuration items and configuration item rules displayed on the target page; wherein, the configuration item rules are the constraints that each configuration item must comply with; the configuration knowledge graph includes nodes for representing each configuration item, and for any two configuration items that have the constraints, the knowledge graph also includes an edge for representing the nodes of the two configuration items, and the edge is used to represent the constraints;

[0021] Detect the topological structures in the configuration knowledge graph that satisfy the preset conflict conditions, and use them as conflict topological structures; and determine the configuration items represented by each node in the conflict topological structure, and use them as the conflict configuration items corresponding to the conflict topological structure.

[0022] The conflict configuration items are input into the target decision tree algorithm model to obtain a conflict resolution scheme; wherein, the target decision tree algorithm model is trained in advance using any of the method steps described in the first aspect;

[0023] The configuration items of the target page are adjusted according to the conflict resolution scheme.

[0024] In one possible implementation, detecting the topological structures in the configured knowledge graph that satisfy preset conflict conditions includes:

[0025] The system detects nodes in the configuration knowledge graph whose number exceeds the interface display threshold, and / or edges whose connected nodes represent configuration items that do not satisfy the constraints represented by the edges.

[0026] In one possible implementation, the method further includes:

[0027] Obtain the severity score of the conflict for the conflicting configuration item;

[0028] The conflicts are sorted according to the principle of the highest severity score to obtain a conflict sequence;

[0029] The step of adjusting the configuration items of the target page according to the conflict resolution scheme includes:

[0030] According to the order in the conflict sequence, the configuration items of the target page are adjusted sequentially according to the conflict resolution scheme corresponding to each conflict configuration item.

[0031] In a third aspect of the present invention, a training apparatus for a decision tree algorithm model is also provided, the apparatus comprising:

[0032] The first acquisition module is used to acquire conflicting sample configuration items and sample effect scores of sample resolution schemes; wherein, the sample resolution scheme is used to resolve the conflicts existing in the sample configuration items;

[0033] The first prediction module is used to input the sample configuration items into the original decision tree algorithm model, obtain the resolution scheme output by the original decision tree algorithm model, and use it as the predicted resolution scheme.

[0034] The first scoring module is used to construct a value function based on the sample configuration items and the prediction resolution scheme to obtain the effect score of the prediction resolution scheme, which is used as the prediction effect score.

[0035] The first adjustment module is used to adjust the model parameters of the original decision tree algorithm model in the direction of reducing the difference, so as to obtain the target decision tree algorithm model; wherein, the difference is the difference between the prediction effect score and the sample effect score.

[0036] In one possible implementation, the device further includes:

[0037] The second acquisition module is used to acquire conflict association information of the conflicts existing in the sample configuration item; wherein, the conflict association information includes at least one of the conflict type, the severity score of the conflict, and the business scenario label of the sample configuration item;

[0038] The first prediction module includes: a first prediction submodule, used to input the sample configuration items and the conflict association information into the original decision tree algorithm model, and obtain the resolution scheme output by the original decision tree algorithm model as the predicted resolution scheme.

[0039] In one possible implementation, the device further includes:

[0040] The first calculation module is used to obtain the score of the first dimension by querying the number of sample configuration items associated with the conflict and determining the score of the first dimension according to the preset number range corresponding rules, and use the first dimension as the first score; wherein, the first dimension is used to characterize the conflict association range of the conflict;

[0041] The second calculation module is used to determine the score of the second dimension by using the impact probability output by the impact prediction model based on the sample configuration item, and use it as the second score; wherein, the second dimension is used to characterize the degree of impact of the conflict on the user;

[0042] The third calculation module is used to determine the score of the third dimension by matching the preset risk level matching rules and the corresponding score of the business impact type caused by the conflict, and use the third dimension as the third score; wherein, the third dimension is used to characterize the degree of business risk caused by the conflict;

[0043] The fourth calculation module is used to perform a weighted calculation on the first score, the second score, and the third score to obtain a calculation result, which serves as the severity score of the conflict.

[0044] In a fourth aspect of the invention, a page configuration conflict resolution apparatus is also provided, the apparatus comprising:

[0045] The graph construction module is used to construct a configuration knowledge graph based on the configuration items and configuration item rules displayed on the target page; wherein, the configuration item rules are the constraints that the configuration items must follow; the configuration knowledge graph includes nodes for representing each configuration item, and for any two configuration items that have the constraints, the knowledge graph also includes an edge for representing the nodes of the two configuration items, and the edge is used to represent the constraints.

[0046] The conflict detection module is used to detect topological structures in the configuration knowledge graph that meet preset conflict conditions, and to identify them as conflict topological structures; and to determine the configuration items represented by each node in the conflict topological structure, as the conflict configuration items corresponding to the conflict topological structure.

[0047] The second prediction module is used to input the conflict configuration items into the target decision tree algorithm model to obtain a conflict resolution scheme; wherein, the target decision tree algorithm model is trained in advance using any of the method steps described in the first aspect;

[0048] The second adjustment module is used to adjust the configuration items of the target page according to the conflict resolution scheme.

[0049] In one possible implementation, the collision detection module includes:

[0050] The first detection submodule is used to detect nodes in the configuration knowledge graph whose number of nodes exceeds the interface display threshold, and / or edges whose configuration items represented by the nodes connected by the edges do not satisfy the constraints represented by the edges.

[0051] In one possible implementation, the device further includes:

[0052] The third acquisition module is used to acquire the severity score of the conflict existing in the conflict configuration item; the conflict sorting module is used to sort the conflicts according to the principle of the highest severity score to obtain the conflict sequence.

[0053] The second adjustment module includes:

[0054] The first submodule is adjusted to adjust the configuration items of the target page according to the order in the conflict sequence and the conflict resolution scheme corresponding to each conflict configuration item.

[0055] In another aspect of the present invention, a computer-readable storage medium is also provided, wherein a computer program is stored therein, and when the computer program is executed by a processor, it implements the training method of any of the decision tree algorithm models described above or the page configuration conflict resolution method.

[0056] In another aspect of the present invention, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute the training method of any of the decision tree algorithm models described above or the page configuration conflict resolution method.

[0057] This invention provides a training method for a decision tree algorithm model and a page configuration conflict resolution method. The method involves obtaining conflicting sample configuration items and sample performance scores for conflict resolution schemes. The sample resolution schemes are used to resolve conflicts in the sample configuration items. The sample configuration items are input into the original decision tree algorithm model to obtain the resolution schemes output by the original model, which are then used as predicted resolution schemes. A value function is constructed based on the sample configuration items and the predicted resolution schemes to obtain the performance scores of the predicted resolution schemes, which are then used as predicted performance scores. The model parameters of the original decision tree algorithm model are adjusted to reduce the discrepancy, resulting in a target decision tree algorithm model. The discrepancy refers to the difference between the predicted performance score and the sample performance score. By using this embodiment, the performance scores of the sample resolution schemes are obtained in advance, and the model parameters are iteratively adjusted based on reducing the difference between the predicted performance score and the sample performance score. This allows the decision tree algorithm model to learn the core logic of conflict resolution, reducing the deviation between the predicted resolution scheme and the optimal solution, thereby improving the efficiency and accuracy of conflict resolution. Attached Figure Description

[0058] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0059] Figure 1 This is a flowchart illustrating a training method for a decision tree algorithm model provided in an embodiment of the present invention.

[0060] Figure 2 This is another flowchart illustrating the training method of the decision tree algorithm model provided in this embodiment of the invention;

[0061] Figure 3 This is a schematic diagram of the severity score calculation process provided in an embodiment of the present invention;

[0062] Figure 4 This is a flowchart illustrating a page configuration conflict resolution method provided in an embodiment of the present invention.

[0063] Figure 5 This is another flowchart illustrating the page configuration conflict resolution method provided in this embodiment of the invention;

[0064] Figure 6 This is another flowchart illustrating the page configuration conflict resolution method provided in this embodiment of the invention;

[0065] Figure 7 This is a schematic diagram of the conflict resolution system provided in an embodiment of the present invention;

[0066] Figure 8 This is a schematic diagram of the structure of the decision tree algorithm model training device provided in this embodiment of the invention;

[0067] Figure 9 This is a schematic diagram of the page configuration conflict resolution device provided in this embodiment of the invention;

[0068] Figure 10 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0069] The technical solutions of the present invention will now be described with reference to the accompanying drawings in the embodiments of the present invention.

[0070] This invention provides a training method for a decision tree algorithm model, see [link to relevant documentation]. Figure 1 The methods include:

[0071] S101, Obtain the sample configuration items with conflicts and the sample effect scores of the sample resolution schemes; wherein, the sample resolution schemes are used to resolve the conflicts existing in the sample configuration items; S102, Input the sample configuration items into the original decision tree algorithm model, and obtain the resolution schemes output by the original decision tree algorithm model as the predicted resolution schemes; S103, Construct a value function based on the sample configuration items and the predicted resolution schemes to obtain the effect scores of the predicted resolution schemes as the predicted effect scores; S104, Adjust the model parameters of the original decision tree algorithm model in the direction of reducing the difference to obtain the target decision tree algorithm model; wherein, the difference is the difference between the predicted effect score and the sample effect score. In this embodiment, by pre-obtaining the effect scores of the sample resolution schemes and iteratively adjusting the model parameters based on reducing the difference between the predicted effect score and the sample effect score, the decision tree algorithm model can learn the core logic of conflict resolution, reduce the deviation between the predicted resolution scheme and the optimal solution of the resolution scheme, and improve the efficiency and accuracy of conflict resolution.

[0072] The following will explain steps S101-S104:

[0073] In step S101 of this embodiment, conflicting configuration items and their resolution schemes' effectiveness scores can be obtained from a pre-built historical case library, serving as sample effectiveness scores for conflicting configuration items and sample resolution schemes. The historical case library includes configuration items that generated conflicts within a historical period, conflict resolution schemes, and the effectiveness scores of the resolution schemes.

[0074] The conflict resolution scheme is used to resolve conflicts in the configuration items. In one possible embodiment, the conflict resolution scheme may include only the successful resolution scheme. In another possible embodiment, the conflict resolution scheme may include both the successful resolution scheme and the failed resolution scheme.

[0075] In step S102 of this embodiment, the decision tree algorithm model can be any model that can obtain a conflict resolution scheme based on configuration items.

[0076] In step S103 of this embodiment, the target decision tree algorithm model is obtained by training through reinforcement learning (Q-Learning). Reinforcement learning is implemented through a value function (Q function). The core of reinforcement learning is the mapping of "state-action-reward". The value function Q(s,a) is used to evaluate "the expected value of performing the resolution action a under a certain conflict state s". Specifically, the value function is constructed based on the sample configuration items and the predicted resolution scheme to obtain the effect score of the predicted resolution scheme. That is, the expected value of performing the predicted resolution scheme when there is a conflict in the sample configuration items, that is, the effect score of the predicted resolution scheme.

[0077] For example, taking conflicting sample configuration items on the checkout page as an example, the value function is set as follows: State s is the input feature, i.e., the sample configuration item. Action a is the action to be performed, i.e., the prediction and resolution solution. For example, a1: Fold up-priced purchases that exceed the threshold and display the "Expand" button; a2: Adjust the sorting of up-priced purchases, prioritizing the display of high-value products; a3: Reduce the size of up-priced purchase cards, forcing all products to be displayed on half the screen. The expected value of executing the predicted resolution solution is determined by the reward function r, which is the core of the value function. According to the "resolution effect score" setting, a positive reward (r = 1~3 points) is given for successful resolution without secondary conflicts (e.g., after execution of a1, the add-on purchase is displayed normally without other configuration conflicts). A resolution effect score r of 3 points is given for successful resolution but with a slight impact on user experience (e.g., after execution of a3, the card size is reduced but operation is not affected). A zero reward (r = 0 points) is given for an ineffective resolution solution that does not cause additional conflicts (e.g., after execution of a2, the add-on purchase still exceeds the screen limit). A negative reward (r = -2 points) is given for a resolution solution that causes new conflicts (e.g., after execution of a3, the add-on purchase text overflows, triggering a display anomaly conflict). The resolution effect score r is the predicted effect score corresponding to the predicted resolution solution.

[0078] The value function Q(s,a) is calculated using the following formula (1):

[0079] ...Formula (1)

[0080] in, Historical Q-values is the attenuation coefficient, with a value of 0.9, and r is the prediction performance score for executing the current configuration item.

[0081] For example, in the initial state, Q(a1)=0; after executing a1, the conflict is resolved without secondary conflict (r is 3 points), and the new Q value is 0×0.9+3=3. When encountering the same state in the future, the model will prioritize a1 with higher Q value (rather than a2 / a3 with lower Q value) and gradually increase the weight of a1 in the decision tree model.

[0082] The Q-value matrix is ​​updated monthly through reinforcement learning, increasing the weight of "high Q-value (high-value) actions (solutions)" and decreasing or even eliminating the weight of "low Q-value (low-value) actions", ultimately improving the accuracy and efficiency of the target decision tree algorithm model.

[0083] In step S104 of this embodiment, the model parameters of the original decision tree algorithm model are adjusted in the direction of reducing the difference between the predicted effect score and the sample effect score to obtain the target decision tree algorithm model. Here, the sample effect score is the effect score of the optimal solution of the solution. Adjusting the model parameters in the direction of reducing the difference reduces the deviation between the predicted solution and the optimal solution.

[0084] In one possible embodiment, see Figure 2 Compared to Figure 1 The example shown, Figure 2 Add step S105, and refine step S102 into step S1021. Figure 2 The methods shown include:

[0085] S101, Obtain the sample configuration items with conflicts and the sample effect scores of the sample resolution schemes; wherein, the sample resolution schemes are used to resolve the conflicts existing in the sample configuration items; S105, Obtain the conflict association information of the conflicts existing in the sample configuration items; wherein, the conflict association information includes at least one of the conflict type, conflict severity score, and business scenario label of the sample configuration item; S1021, Input the sample configuration items and conflict association information into the original decision tree algorithm model, and obtain the resolution scheme output by the original decision tree algorithm model as the predicted resolution scheme; S103, Construct a value function based on the sample configuration items and the predicted resolution scheme to obtain the effect score of the predicted resolution scheme as the predicted effect score; S104, Adjust the model parameters of the original decision tree algorithm model in the direction of reducing the difference to obtain the target decision tree algorithm model; wherein, the difference is the difference between the predicted effect score and the sample effect score. In this embodiment, sample configuration items and conflict association information including conflict type, severity score, and business scenario label are automatically obtained and input into the original decision tree model to obtain a predicted resolution solution. This provides the model with comprehensive input to output a solution that fits the actual needs, reduces bias, improves the accuracy and generation efficiency of the resolution solution, and thus improves the efficiency and accuracy of conflict resolution.

[0086] Understandably, in this article Figure 2 The steps S101 and S105 shown are executed in parallel. This execution order is only one possible example. In other embodiments, step S101 may be executed first, followed by step S105, or step S105 may be executed first, followed by step S101. Alternatively, step S101 and step S105 may be executed alternately. No specific limitation is made here.

[0087] Because steps S101, S103 and S104 are related to Figure 1 The example shown is the same, so please refer to [link / reference]. Figure 1 The relevant explanations will not be repeated here; the following text will only address... Figure 2 The newly added step S105 and the refined step S1021 are explained below:

[0088] In step S105 of this embodiment, the conflict types include conditional conflicts and display conflicts. Conditional conflicts refer to situations where core configuration items do not match business rules, user tags, etc., such as a mismatch between pricing strategies and user tagging systems, resulting in the configuration scheme failing to adapt to the target user group. Display conflicts refer to situations where the display format and quantity of configuration items exceed the system's preset display rules or screen efficiency thresholds, such as the number of packages exceeding the screen display efficiency threshold, causing display abnormalities and a decline in user experience.

[0089] The types of conflicts can be classified using a random forest classifier. For example, a random forest classifier can be built using Scikit-learn, with training data from historical configuration error cases. The error cases are labeled with conflict types. Taking the aforementioned display conflicts and conditional conflicts as an example, the conflict type can be characterized as a conditional conflict by labeling it as a price contradiction or logical error, and the conflict type can be characterized as a display conflict by labeling it as a display violation.

[0090] During the training of the random forest classifier, 15-dimensional features, such as node attributes (e.g., price, quantity) and relationships between nodes (e.g., mandatory association, conflicting association), are extracted based on the training data and used for model training.

[0091] After classifying conflict types using a random forest classifier, different types of conflicts can be displayed in different ways. Specifically, different colors, fonts, etc., can be used to distinguish different types of conflicts. For example, red can be used to mark conditional conflicts (such as a mismatch between pricing strategies and user tags), and yellow can be used to mark display conflicts (such as the number of packages exceeding the screen efficiency threshold).

[0092] In another possible embodiment, a model with the same classification function as the random forest classifier can be used to classify conflicts, without being specifically limited here.

[0093] In one possible embodiment, the severity score of the conflict is pre-assessed by a professional based on work experience or industry regulations, taking into account the conflicts that exist between sample configuration items, and is used for conflict prioritization and handling strategy formulation.

[0094] Business scenario tags are standardized classifications of business application scenarios, used to identify the applicable scenarios for configuration items, such as new user acquisition campaign scenarios, regular business configuration scenarios, and holiday marketing scenarios.

[0095] In step S1021 of this embodiment, the sample configuration items and conflict association information are input into the original decision tree algorithm model, and the original decision tree algorithm model outputs a predicted resolution scheme based on the sample configuration items and conflict association information.

[0096] In one possible embodiment, the data input to the original decision tree algorithm model also includes the conflict association information obtained in step S105. Therefore, the obtained predicted resolution scheme and its score are also affected by the conflict type, the conflict severity score, and the business scenario label of the sample configuration item. Based on this, the state s of the value function Q(s,a), i.e., the input features, also includes at least one of the conflict type, severity score, and business scenario label.

[0097] In another possible embodiment, see Figure 3 The severity score of the aforementioned conflict was obtained using the following methods:

[0098] S301, by querying the number of sample configuration items associated with the conflict and determining the score of the first dimension according to the preset number interval corresponding rules, the first dimension is used to characterize the scope of the conflict association. S302, by determining the score of the second dimension based on the impact probability output by the impact prediction model based on the sample configuration items, the second dimension is used to characterize the degree of impact of the conflict on users. S303, by determining the score of the third dimension based on the preset risk level matching rules and the corresponding score of the business impact type caused by the conflict, the third dimension is used to characterize the degree of business risk caused by the conflict. S304, by weighting the first, second, and third scores, the calculation result is obtained as the severity score of the conflict. In this embodiment, the conflict severity score quantifies the severity of the conflict through three dimensions (scope of impact, user impact, and business risk). Based on the conflict and its severity score, a target decision tree algorithm model is trained to generate a more targeted resolution solution, improving the efficiency and accuracy of conflict resolution.

[0099] The following will explain the aforementioned steps S301-S304:

[0100] In steps S301-S304 of this embodiment, a three-dimensional evaluation system is designed, including a first dimension for characterizing the scope of conflict association, a second dimension for characterizing the degree of impact of conflict on users, and a third dimension for characterizing the degree of business risk caused by conflict. The severity of the conflict is evaluated from these three dimensions.

[0101] For example, taking a score range of 0-10 points for each score, a preset number range is set for the first score. When the total number of configuration items is 100, the first score is 0 when the number of sample configuration items associated with the conflict is 0; when the number is 1-10, the first score is 1 point... and so on. When the number is 100, the first score is 10 points.

[0102] For the second score, a prediction model trained on historical complaint data is used as input for the conflict type and user traffic (e.g., 100,000 users per day), outputting the complaint probability and converting it into a score. For example, if the complaint probability for "display conflict" is 50%, then the second score is 50% × 10 = 5 points.

[0103] For the third score, a score is assigned based on preset risk level matching rules and the type of business impact caused by the conflict. For example, a preset risk level of completely blocking payment (such as payment method misconfiguration leading to payment failure) is Level 1, with a corresponding third score of 10 points; a preset risk level of affecting the use of discounts (such as mismatch between add-on purchase and package price) is Level 2, with a corresponding third score of 7 points; and a preset risk level of only displaying anomalies (such as misaligned package description) is Level 3, with a corresponding third score of 3 points. For example, "condition conflict" directly blocks the transaction, with a third score of 10 points.

[0104] Specifically, completely blocking payment, affecting the use of discounts, and displaying abnormalities are determined through the following methods:

[0105] The logic for completely blocking payments is that conflicting configuration items are directly associated with the "core payment process" (such as payment method activation status, deduction interface permissions, and order generation rules), causing users to be unable to complete payments.

[0106] For example, the operator configured "XX payment is the only payment method" but mistakenly selected "XX payment is not currently supported"; or configured "order amount ≥ 0 yuan is required for payment" but the package price field is empty. The processor detects that the conflicting configuration item is a "required item in the payment process" and that the conflict will cause the payment interface call to fail / the order cannot be generated, which is judged as "completely blocking payment".

[0107] The logic for determining whether a discount is applicable is that conflicting configuration items are associated with the "discount deduction, price 0 calculation" process. This does not affect the payment itself, but users will not be able to enjoy the expected discount.

[0108] For example, if a "new user's first month 6 yuan package" is configured, and "vouchers not supported" is enabled, but the user has already claimed the voucher for this package; or if there is a conflict between the add-on purchase discount rules and the package discount stacking logic (such as "add-on purchase direct discount of 20 yuan" but stacking permission is not configured), the processor will identify that the conflict involves "discount rule configuration items" and will cause the discount to fail, and will determine it as "affecting the use of discounts".

[0109] The logic for determining display anomalies involves associating conflicting configuration items with "interface display rules" (such as the number of package cards, the folding threshold for add-on purchases, and text display). This does not affect business logic (payment, discounts) but only the visual experience.

[0110] For example, if the add-on purchase quantity is configured to 3, but the "slide display" rule is not enabled (the default half-screen display is a maximum of 2), the 3rd add-on purchase will be truncated; or the package description length exceeds the card width, causing text overflow. The computer detects that the conflict belongs to "display-related configuration items" and does not involve the core transaction process, and judges it as "display abnormality".

[0111] After obtaining the scores for the three dimensions mentioned above, the severity score is calculated. In one possible implementation, different weights are set for each dimension based on requirements to improve the accuracy of conflict resolution. For example, the analytic hierarchy process (AHP) can be used to determine the weights of each dimension (e.g., 0.4 for the first dimension, 0.3 for the second dimension, and 0.3 for the third dimension) to generate a severity score of 0-10.

[0112] The following explains how to use a hierarchical approach to determine the weights of each dimension. First, the review team consists of business experts (operations managers), technical experts (backend architects), and data experts (data analysts). When determining weights, each expert independently completes a pairwise comparison matrix (e.g., comparing the importance of "scope of influence" and "user impact"). Then, the matrix consistency ratio (CR) is calculated, requiring CR < 0.1 (if this is not met, the matrix is ​​re-completed). Finally, a weighted average method is used to calculate the final weights, and the results are stored in a weight table. The weights are periodically reviewed to determine if adjustments are needed.

[0113] In another possible embodiment, the average of the scores of the three dimensions mentioned above can also be calculated as the severity score during the calculation of the severity score.

[0114] Based on the aforementioned decision tree algorithm model training method, this embodiment of the invention also provides a page configuration conflict resolution method, see [link to relevant documentation]. Figure 4 The methods include:

[0115] S401, Construct a configuration knowledge graph based on the configuration items and rules displayed on the target page; wherein, the configuration item rules are the constraints that each configuration item must adhere to; the configuration knowledge graph includes nodes for representing each configuration item, and for any two configuration items with constraints, the knowledge graph also includes edges for representing the nodes between the two configuration items, and the edges are used to represent the constraints; S402, Detect the topological structures in the configuration knowledge graph that satisfy the preset conflict conditions, and designate them as conflict topological structures; and determine the configuration items represented by each node in the conflict topological structure, as the conflict configuration items corresponding to the conflict topological structure; S403, Input the conflict configuration items into the target decision tree algorithm model to obtain a conflict resolution scheme; wherein, the target decision tree algorithm model is trained in advance using the aforementioned training method steps for the decision tree algorithm model; S404, Adjust the configuration items on the target page according to the conflict resolution scheme. Using this embodiment, by constructing a knowledge graph containing nodes of configuration items and edges for representing the constraints that each configuration item must adhere to, the scattered configuration items and their logical constraints are transformed into a structured knowledge network. This graph-based representation can capture the dependencies between configuration items, avoiding logical omissions or misjudgments caused by scattered rules during manual analysis. It detects topological structures in the configuration knowledge graph that meet preset conflict conditions (i.e., detects existing conflicts), and generates resolution schemes based on the conflicting configuration items corresponding to these topological structures and the target decision tree algorithm model. This replaces conflict verification and resolution methods that rely on human experience, avoiding subjectivity in manual resolution and providing an automated conflict verification and resolution method, thereby improving the efficiency and accuracy of conflict resolution.

[0116] The conflict resolution method provided in this embodiment of the invention can be applied to any graph database. For ease of explanation, this article only uses the application of the above conflict resolution method to the Neo4j graph database as an example.

[0117] Before implementing the conflict resolution methods described above, a Neo4j graph database environment must first be set up. Specifically, setting up the Neo4j graph database environment includes installing the Neo4j server and configuring database parameters. This ensures the stability and performance of the database. A cluster architecture for the graph database is designed, using a 3-node Neo4j cluster (2 masters and 1 slave) deployed in a Docker container, with load balancing achieved through Nginx. This improves the availability and scalability of the database. Neo4j has efficient graph computation capabilities and a flexible query language, enabling it to support the storage and querying of large-scale graph data.

[0118] After the configuration knowledge graph is built, it is stored using the Neo4j graph database. Specifically, bolt port encryption is configured, and data backup strategies are set, including daily full backups and hourly incremental backups.

[0119] For ease of description, this article uses a half-screen cashier scenario configured in the member transaction backend as an example to illustrate the conflict resolution method provided by the embodiments of the present invention. In this scenario, the target page is the cashier page.

[0120] Steps S401-S404 will be explained below:

[0121] In step S401 of this embodiment, after establishing the Neo4j graph database environment, configuration item data is collected from the member transaction backend configuration system using a data acquisition tool. Configuration items include the products, payment methods, and add-on purchases displayed on the checkout page. Products can be corresponding packages (e.g., a discounted package consisting of a mobile phone, headphones, and a power bank, presented as a whole product configuration item on the checkout page), images (e.g., main product image, detailed images from different angles, used to visually display the product's appearance on the checkout page), and unit prices. Payment methods can be various common electronic payment methods. Add-on purchases can be additional items that consumers can obtain by paying a certain amount extra on top of purchasing the main product; for example, when purchasing a cosmetics set from a certain brand, adding 99 yuan can obtain other products from the same brand.

[0122] For example, the data acquisition tool can be an ETL (Extract Transform Load) tool, which can extract data periodically and automatically to ensure the timeliness and integrity of the data.

[0123] To improve the accuracy of conflict detection, in one possible embodiment, the collected configuration item data can be cleaned. For example, a Python script can be used to clean and preprocess the collected data. The cleaning process includes removing invalid fields, filling in missing values, and converting data types to ensure the validity and consistency of the data.

[0124] After cleaning the data, key attributes are extracted and stored in the `config_items` table of a MySQL database. This table stores configuration item data in a structured manner, facilitating subsequent graph construction and conflict detection. For example, taking the products displayed on the checkout page as corresponding package deals, key attributes include package price, payment method sorting, and add-on quantity.

[0125] After collecting the configuration items, it is also necessary to collect the configuration item rules. The configuration item rules are the constraints that each configuration item must comply with, which may include display rules and logic rules.

[0126] Display rules define how configuration items are presented in the user interface or documentation. For example, a package deal should be displayed after the payment method is selected, and the add-on option should dynamically pop up after the user selects a specific package. Logical rules define the business constraints or dependencies between configuration items. For example, a payment method may only support a specific package, and add-on items should meet price or category matching logic with the main product.

[0127] Configuration rules are extracted by analyzing the business documents (such as PRDs, Product Requirement Documents) of the half-screen checkout system. The analysis employs a semi-automatic process combining manual initial screening and NLP (Neuro-Linguistic Programming)-assisted verification. Specifically, core rules (such as "continuous monthly subscriptions must display the crossed-out price") are manually extracted from the business documents and operations manual to ensure the accuracy of the business logic. After extracting the core rules, a text classification model is used to perform semantic verification, identifying duplicate rules (such as "add-on purchases > 2 require swiping" and "add-on purchases exceeding 2 require scrolling") and prompting manual merging to ultimately form the configuration rules. For example, display rules could be "add-on purchases > 2 require swiping," and logical rules could be "continuous monthly subscriptions must display the crossed-out price."

[0128] After extracting the configuration rules, the extracted rules can be stored in JSON (JavaScript Object Notation) format. JSON format is characterized by its clear structure and ease of parsing, which facilitates subsequent graph construction and conflict detection.

[0129] Each of the aforementioned rules includes fields such as condition (IF), action (THEN), and priority (PRIORITY), which are stored in the rule_knowledge table. This table stores rule data in a structured manner, providing a basis for conflict detection.

[0130] After obtaining the configuration items and configuration item rules, a configuration knowledge graph is constructed based on the configuration items and configuration item rules.

[0131] When constructing a configuration knowledge graph, the first step is to define the nodes within it. Specifically, in the Neo4j graph database, nodes are defined based on each configuration item. Nodes represent configuration items, therefore, they include package nodes, payment method nodes, and add-on purchase nodes. In addition to the nodes mentioned above, nodes can also include rule nodes, tag nodes, etc., without specific limitations here.

[0132] After defining each node, you can also define its attributes. Attributes for the "Package" node include Package ID, price, auto-renewal status, and applicable tags. Attributes for the "Payment Method" node include Payment ID, sorting, whether it's selected by default, and supported discount types. Attributes for the "Add-on Purchase" node include Product ID, quantity, price, and display method. These attributes comprehensively describe the characteristics of the configuration items.

[0133] After defining each node, edges are constructed in the configuration knowledge graph based on each node. Each edge in the configuration knowledge graph represents a configuration item rule, and the node connected to each edge is the node involved in the configuration item rule represented by that edge. In other words, each edge represents the constraints that need to be followed between the connected configuration items.

[0134] Specifically, relationship edges between nodes are constructed based on the configuration rules they satisfy. These edges include mandatory associations (e.g., the mandatory relationship between continuous monthly subscription packages and the crossed-out price display rule), restrictive associations (e.g., the conditional relationship between the quantity of add-on purchases and the sliding display rule), and conflicting associations (e.g., the contradictory relationship between not supporting vouchers and the first-month discount package). Edge construction must accurately reflect the logical relationships within the business rules to ensure accurate conflict detection.

[0135] In step S402 of this embodiment, the preset conflict conditions include, but are not limited to: inconsistencies between display rules and logical rules (e.g., conflict between payment method display and package compatibility), circular dependencies or mutual exclusion between rules, hierarchical or priority conflicts between configuration items, and infinite loops or non-executable situations in rule execution.

[0136] In one possible embodiment, a conflict detection algorithm can be developed to detect topologies in the configuration knowledge graph that satisfy preset conflict conditions. In this paper, the conflicting configuration items corresponding to the conflicting topologies that satisfy the preset conflict conditions are the detected conflicts in the configuration knowledge graph.

[0137] Specifically, a conflict detection algorithm can be developed based on Neo4j's Cypher query language. Cypher is a query language specifically designed for graph databases.

[0138] Taking display conflicts and conditional conflicts as examples, when the conflict is a display conflict, the preset conflict condition can be the inconsistency between display rules and logical rules. A display conflict detection algorithm is developed to detect display conflicts. During the detection process, the algorithm needs to traverse the relevant nodes and relationship edges in the configuration knowledge graph to determine whether each edge and the node it connects to meets the preset conflict condition. For example, if a maximum of three packages can be displayed in half a screen, the algorithm checks whether the number of packages exceeds this threshold.

[0139] When the conflict is a conditional conflict, the preset conflict conditions can be one or more of the following: circular dependencies or mutual exclusions between rules, hierarchical or priority conflicts between configuration items, and infinite loops or non-executable situations during rule execution. A conditional conflict detection algorithm is developed to detect conditional conflicts. During the detection process, the algorithm also needs to traverse the relevant nodes and relationship edges in the configuration knowledge graph to determine whether each edge and the node it connects to satisfies the preset conflict conditions. For example, node attribute matching can be used to detect logical contradictions between pricing strategies and user tags; for instance, it can detect whether the "new user first month 6 yuan" package has been configured for existing user tags.

[0140] To improve the efficiency of conflict checking and thus conflict resolution, in one possible embodiment, the performance of the conflict detection algorithm can be optimized to ensure that the detection latency is controlled within 50ms.

[0141] In step S403 of this embodiment, after a conflict is detected, the detected conflict is input into the target decision tree algorithm model to generate a conflict resolution scheme.

[0142] In step S404 of this embodiment, after obtaining the conflict resolution solution, the configuration items on the page are adjusted according to the conflict resolution solution, that is, the conflict is resolved. For example, the number, position, and display method of the configuration items are adjusted.

[0143] During the adjustment process, configuration modifications are executed through transaction management, generating configuration change logs. The conflict resolution process must ensure data consistency and integrity to avoid new problems caused by configuration changes. Therefore, the Neo4j database configuration must be modified according to the conflict resolution plan, the configuration knowledge graph must be updated synchronously, and the resolution of conflicts must be verified.

[0144] During this process, a rollback mechanism can be configured to automatically roll back if verification fails. All changes are recorded using AOP (Aspect-Oriented Programming) aspects, ensuring configuration consistency and traceability. The rollback mechanism allows for timely restoration of the original configuration in case of problems with configuration modifications, preventing impact on business operations.

[0145] In another possible embodiment, the existence of a topology in the configuration knowledge graph that satisfies preset conflict conditions can be determined by judging the number of nodes and whether the nodes connected by the edges satisfy the configuration item rules. For details, see [link to relevant documentation]. Figure 5 Compared to Figure 4 The example shown, Figure 5 Step S402 is further refined into step S4021. Figure 5 The methods shown include:

[0146] S401, Construct a configuration knowledge graph based on the configuration items and configuration rules displayed on the target page; wherein, the configuration rules are the constraints that each configuration item must comply with; the configuration knowledge graph includes nodes for representing each configuration item, and for any two configuration items with constraints, the knowledge graph also includes edges for representing the nodes between the two configuration items, and the edges are used to represent constraints; S4021, Detect nodes in the configuration knowledge graph whose number of nodes exceeds the interface display threshold, and / or edges whose connected nodes represent configuration items that do not satisfy the constraints represented by the edges, as conflict topologies; and determine the configuration items represented by each node in the conflict topology as the conflict configuration items corresponding to the conflict topology; S403, Input the conflict configuration items into the target decision tree algorithm model to obtain a conflict resolution scheme; wherein, the target decision tree algorithm model is trained in advance using the aforementioned decision tree algorithm model training method steps; S404, Adjust the configuration items on the target page according to the conflict resolution scheme. By using this embodiment, the existence of a topological structure that satisfies the preset conflict conditions in the configuration knowledge graph is determined by detecting whether the number of nodes exceeds the interface display threshold and / or whether the configuration items represented by the nodes connected by the edges do not meet the constraints represented by the edges. This reduces manual intervention and improves the efficiency and accuracy of conflict resolution.

[0147] Step S4021 is a detailed step of the aforementioned step S402. Steps S401, S403 and S404 have been explained in the preceding text and will not be repeated here.

[0148] The interface display threshold can be set by professional technicians based on their work experience or according to industry standards. This embodiment of the invention does not impose any specific limitations on this.

[0149] In one possible implementation, to further improve the efficiency and reduce the cost of conflict resolution, conditions for triggering conflict verification and resolution can be pre-set. This allows for automated verification of the configuration scheme in response to a release command for the checkout page. Specifically, a hook function is added to the configuration release process. Operators can enter a release command by clicking the "Release" button or using preset controls on the keyboard, and the system automatically executes the aforementioned conflict detection and resolution steps. After conflict resolution, a detection report is generated, containing conflict details, resolution suggestions, and an impact assessment. The report is in Markdown format and includes a conflict node relationship graph (generated using the Neo4j GraphAware plugin) to ensure zero-conflict configuration releases.

[0150] It can also automatically validate configuration schemes in response to changes made to configuration items on the checkout page.

[0151] When configuration items are modified, WebSocket can be integrated into the system. When operators modify configuration items, the front end sends the modified configuration data to the back end in real time. The back end triggers conflict detection based on the configuration knowledge graph and provides feedback on whether a conflict was detected. WebSocket is a protocol for full-duplex communication over a single TCP connection. After a conflict is detected, the conflicting item is displayed in a preset manner, along with resolution suggestions, achieving real-time linkage between "modification-detection-prompt".

[0152] By applying the above embodiments, conflict detection and resolution are only performed upon receiving a release command and / or modification operation, thereby avoiding unnecessary conflict detection and reducing resource waste. Simultaneously, performing detection when the configuration scheme changes allows for timely discovery and resolution of conflicts, improving the timeliness of conflict resolution.

[0153] When multiple conflicts are detected, to prioritize resolving the most severe conflicts, the conflicts can be sorted according to their severity scores, and then resolved in that order. For details, see [link to relevant documentation]. Figure 6 Compared to Figure 4 The example shown, Figure 6 Add steps S405 and S406, and refine step S404 into step S4041. Figure 6 The methods shown include:

[0154] S401, Construct a configuration knowledge graph based on the configuration items and configuration item rules displayed on the target page; wherein, the configuration item rules are the constraints that each configuration item must adhere to; the configuration knowledge graph includes nodes to represent each configuration item, and for any two configuration items with constraints, the knowledge graph also includes edges to represent the nodes between the two configuration items, and the edges are used to represent the constraints; S402, Detect the topological structures in the configuration knowledge graph that satisfy the preset conflict conditions, and designate them as conflict topological structures; and determine the configuration items represented by each node in the conflict topological structure, and make them... S403: Input the conflict configuration items corresponding to the conflict topology into the target decision tree algorithm model to obtain a conflict resolution scheme; wherein, the target decision tree algorithm model is trained in advance using the aforementioned decision tree algorithm model training method steps; S405: Obtain the severity score of the conflicts existing in the conflict configuration items; S406: Sort each conflict according to the principle of the highest severity score to obtain a conflict sequence; S4041: Adjust the configuration items of the target page according to the order in the conflict sequence and the conflict resolution scheme corresponding to each conflict configuration item. Using this embodiment, sorting each conflict topology according to the principle of the highest severity score to obtain a conflict sequence, and resolving each conflict according to the order in the sequence, allows for a more accurate assessment of conflict priority, thereby prioritizing the handling of high-risk conflicts and improving the timeliness of conflict resolution.

[0155] Step S4041 is a detailed step of the aforementioned step S404. Steps S401-S403 have been explained in the preceding text and will not be repeated here.

[0156] Understandably, in this article Figure 6 The steps S405 and S406 shown are executed before step S4041, and can be executed in parallel with any one of the steps S401-S403. This execution order is just one possible example and is not specifically limited here.

[0157] In steps S405, S406, and S4041 of this embodiment, the conflict topologies are sorted in descending order of the severity scores obtained above to obtain a conflict sequence. The conflict topologies in this conflict sequence are then resolved in the order they appear.

[0158] In the aforementioned conflict resolution process, a RESTful interface should be developed to allow operations personnel to trigger conflict detection and resolution operations through a visual interface. The interface needs to be highly secure and easy to use, facilitating its use by operations personnel. The interface documentation should provide a detailed description of its functionality, parameters, return values, and other information to facilitate invocation and integration by developers.

[0159] Corresponding to the aforementioned conflict resolution method, a four-layer architecture is adopted to execute the aforementioned conflict resolution scheme. See [link to relevant documentation]. Figure 7 ,include:

[0160] Data layer 701 is used to store the aforementioned configuration items and configuration item rules;

[0161] Graph layer 702 is used to construct knowledge graphs and relationship networks, which is equivalent to the aforementioned step S401;

[0162] The detection layer 703 is used to realize conflict identification and analysis, which is equivalent to the aforementioned steps S402 and S103;

[0163] Application layer 704 is used to provide configuration management and resolution interfaces to implement the aforementioned steps S401-S404.

[0164] Corresponding to the aforementioned training method for the decision tree algorithm model, this embodiment of the invention also provides a decision tree algorithm model training device, see [link to relevant documentation]. Figure 8 The device includes:

[0165] The first acquisition module 801 is used to acquire conflicting sample configuration items and sample effect scores of sample resolution schemes; wherein, the sample resolution scheme is used to resolve the conflicts existing in the sample configuration items; the first prediction module 802 is used to input the sample configuration items into the original decision tree algorithm model and acquire the resolution scheme output by the original decision tree algorithm model as the predicted resolution scheme; the first scoring module 803 is used to construct a value function based on the sample configuration items and the predicted resolution scheme to obtain the effect score of the predicted resolution scheme as the predicted effect score; the first adjustment module 804 is used to adjust the model parameters of the original decision tree algorithm model in the direction of reducing differences to obtain the target decision tree algorithm model; wherein, the difference is the difference between the predicted effect score and the sample effect score.

[0166] In one possible implementation, the apparatus further includes: a second acquisition module, configured to acquire conflict association information of conflicts existing in the sample configuration item; wherein the conflict association information includes at least one of the conflict type, the severity score of the conflict, and the business scenario label of the sample configuration item; and a first prediction module, comprising: a prediction first submodule, configured to input the sample configuration item and the conflict association information into the original decision tree algorithm model, and acquire the resolution scheme output by the original decision tree algorithm model as the predicted resolution scheme.

[0167] In one possible implementation, the device further includes: a first calculation module, configured to obtain a score for a first dimension by querying the number of sample configuration items associated with the conflict and determining a score for a first dimension based on a preset number range correspondence rule; wherein the first dimension is used to characterize the conflict association range of the conflict; a second calculation module, configured to obtain a score for a second dimension by using an impact prediction model based on the impact probability output by the sample configuration items; wherein the second dimension is used to characterize the degree of impact of the conflict on the user; a third calculation module, configured to obtain a score for a third dimension by using a preset risk level matching rule and matching the corresponding score of the business impact type caused by the conflict; wherein the third dimension is used to characterize the degree of business risk caused by the conflict; and a fourth calculation module, configured to perform a weighted calculation on the first score, the second score, and the third score to obtain a calculation result, which is used as the severity score of the conflict.

[0168] Corresponding to the aforementioned conflict resolution method, this embodiment of the invention also provides a page configuration conflict resolution device, see [link to relevant documentation]. Figure 9 The device includes:

[0169] The graph construction module 901 is used to construct a configuration knowledge graph based on the configuration items and configuration item rules displayed on the target page; wherein, the configuration item rules are the constraints that each configuration item must comply with; the configuration knowledge graph includes nodes for representing each configuration item, and for any two configuration items with the constraints, the knowledge graph also includes an edge for representing the nodes of the two configuration items, and the edge is used to represent the constraints; the conflict detection module 902 is used to detect the topological structures in the configuration knowledge graph that satisfy the preset conflict conditions, as conflict topological structures; and determine the configuration items represented by each node in the conflict topological structure, as the conflict configuration items corresponding to the conflict topological structure; the second prediction module 903 is used to input the conflict configuration items into the target decision tree algorithm model to obtain a conflict resolution scheme; wherein, the target decision tree algorithm model is trained in advance using the training method steps of the decision tree algorithm model; the second adjustment module 904 is used to adjust the configuration items of the target page according to the conflict resolution scheme.

[0170] In one possible implementation, the conflict detection module includes: a first detection submodule, used to detect nodes in the configuration knowledge graph whose number of nodes exceeds the interface display threshold, and / or edges whose configuration items represented by the nodes connected by the edges do not satisfy the constraints represented by the edges.

[0171] In one possible implementation, the device further includes: a third acquisition module, configured to acquire the severity score of the conflict existing in the conflict configuration item; a conflict sorting module, configured to sort each conflict according to the principle of the highest severity score to obtain a conflict sequence; and a second adjustment module, including: an adjustment first submodule, configured to adjust the configuration items of the target page according to the order in the conflict sequence and the conflict resolution scheme corresponding to each conflict configuration item.

[0172] This invention also provides an electronic device, such as... Figure 10 As shown, it includes a processor 1001, a communication interface 1002, a memory 1003, and a communication bus 1004, wherein the processor 1001, the communication interface 1002, and the memory 1003 communicate with each other through the communication bus 1004.

[0173] Memory 1003 is used to store computer programs;

[0174] When processor 1001 executes a program stored in memory 1003, it performs the following steps:

[0175] Obtain the sample effect scores of conflicting sample configuration items and sample resolution schemes; wherein, the sample resolution schemes are used to resolve the conflicts existing in the sample configuration items;

[0176] The sample configuration items are input into the original decision tree algorithm model to obtain the resolution scheme output by the original decision tree algorithm model, which is then used as the predicted resolution scheme.

[0177] Based on the sample configuration items and the prediction resolution scheme, a value function is constructed to obtain the effect score of the prediction resolution scheme, which is used as the prediction effect score;

[0178] The model parameters of the original decision tree algorithm model are adjusted in the direction of reducing the difference to obtain the target decision tree algorithm model; wherein, the difference is the difference between the predicted effect score and the sample effect score;

[0179] or,

[0180] A configuration knowledge graph is constructed based on the configuration items and configuration item rules displayed on the target page; wherein, the configuration item rules are the constraints that each configuration item must comply with; the configuration knowledge graph includes nodes for representing each configuration item, and for any two configuration items that have the constraints, the knowledge graph also includes an edge for representing the nodes of the two configuration items, and the edge is used to represent the constraints;

[0181] Detect the topological structures in the configuration knowledge graph that satisfy the preset conflict conditions, and use them as conflict topological structures; and determine the configuration items represented by each node in the conflict topological structure, and use them as the conflict configuration items corresponding to the conflict topological structure.

[0182] The conflict configuration items are input into the target decision tree algorithm model to obtain a conflict resolution scheme; wherein, the target decision tree algorithm model is obtained by training in advance using any of the steps described in the method for training decision tree algorithm models;

[0183] The configuration items of the target page are adjusted according to the conflict resolution scheme.

[0184] The communication bus mentioned above can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used to represent it in the diagram, but this does not mean that there is only one bus or one type of bus.

[0185] The communication interface is used for communication between the aforementioned terminal and other devices.

[0186] The memory may include random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0187] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0188] In another embodiment of the present invention, a computer-readable storage medium is also provided, wherein a computer program is stored therein, and when the computer program is executed by a processor, it implements the training method of the decision tree algorithm model or the page configuration conflict resolution method described in any of the above embodiments.

[0189] In another embodiment of the present invention, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute the training method of the decision tree algorithm model or the page configuration conflict resolution method described in any of the above embodiments.

[0190] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0191] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0192] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0193] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of protection of the present invention.

Claims

1. A training method for a decision tree algorithm model, characterized in that, The method includes: Obtain the sample effect scores of conflicting sample configuration items and sample resolution schemes; wherein, the sample resolution schemes are used to resolve the conflicts existing in the sample configuration items; The sample configuration items are input into the original decision tree algorithm model to obtain the resolution scheme output by the original decision tree algorithm model, which is then used as the predicted resolution scheme. Based on the sample configuration items and the prediction resolution scheme, a value function is constructed to obtain the effect score of the prediction resolution scheme, which is used as the prediction effect score; The model parameters of the original decision tree algorithm model are adjusted in the direction of reducing the difference to obtain the target decision tree algorithm model; wherein, the difference is the difference between the predicted effect score and the sample effect score.

2. The method according to claim 1, characterized in that, The method further includes: Obtain conflict association information of the conflicts existing in the sample configuration item; wherein, the conflict association information includes at least one of the conflict type, the severity score of the conflict, and the business scenario label of the sample configuration item; The step of inputting the sample configuration items into the original decision tree algorithm model and obtaining the resolution scheme output by the original decision tree algorithm model as the predicted resolution scheme includes: The sample configuration items and the conflict association information are input into the original decision tree algorithm model to obtain the resolution scheme output by the original decision tree algorithm model, which is then used as the predicted resolution scheme.

3. The method according to claim 2, characterized in that, The method further includes: The score of the first dimension is obtained by querying the number of sample configuration items associated with the conflict and determining the score according to the preset number range corresponding rules; wherein, the first dimension is used to characterize the conflict association range of the conflict. The score for the second dimension is determined by the impact probability output by the impact prediction model based on the sample configuration items, and is used as the second score; wherein, the second dimension is used to characterize the degree of impact of the conflict on the user; The score for the third dimension is determined by matching the preset risk level with the corresponding score for the type of business impact caused by the conflict; wherein, the third dimension is used to characterize the degree of business risk caused by the conflict. The first score, the second score, and the third score are weighted and calculated to obtain the result, which is used as the severity score of the conflict.

4. A method for resolving page configuration conflicts, characterized in that, The method includes: A configuration knowledge graph is constructed based on the configuration items and configuration item rules displayed on the target page; wherein, the configuration item rules are the constraints that each configuration item must comply with; the configuration knowledge graph includes nodes for representing each configuration item, and for any two configuration items that have the constraints, the knowledge graph also includes an edge for representing the nodes of the two configuration items, and the edge is used to represent the constraints; Detect the topological structures in the configuration knowledge graph that satisfy the preset conflict conditions, and use them as conflict topological structures; and determine the configuration items represented by each node in the conflict topological structure, and use them as the conflict configuration items corresponding to the conflict topological structure. The conflict configuration items are input into the target decision tree algorithm model to obtain a conflict resolution scheme; wherein, the target decision tree algorithm model is obtained in advance by training using the method steps of any one of claims 1-3; The configuration items of the target page are adjusted according to the conflict resolution scheme.

5. The method according to claim 4, characterized in that, The detection of topological structures in the configured knowledge graph that satisfy preset conflict conditions includes: The system detects nodes in the configuration knowledge graph whose number exceeds the interface display threshold, and / or edges whose connected nodes represent configuration items that do not satisfy the constraints represented by the edges.

6. The method according to claim 4, characterized in that, The method further includes: Obtain the severity score of the conflict for the conflicting configuration item; The conflicts are sorted according to the principle of the highest severity score to obtain a conflict sequence; The step of adjusting the configuration items of the target page according to the conflict resolution scheme includes: According to the order in the conflict sequence, the configuration items of the target page are adjusted sequentially according to the conflict resolution scheme corresponding to each conflict configuration item.

7. A training device for a decision tree algorithm model, characterized in that, The device includes: The first acquisition module is used to acquire conflicting sample configuration items and sample effect scores of sample resolution schemes; wherein, the sample resolution scheme is used to resolve the conflicts existing in the sample configuration items; The first prediction module is used to input the sample configuration items into the original decision tree algorithm model, obtain the resolution scheme output by the original decision tree algorithm model, and use it as the predicted resolution scheme. The first scoring module is used to construct a value function based on the sample configuration items and the prediction resolution scheme to obtain the effect score of the prediction resolution scheme, which is used as the prediction effect score. The first adjustment module is used to adjust the model parameters of the original decision tree algorithm model in the direction of reducing the difference, so as to obtain the target decision tree algorithm model; wherein, the difference is the difference between the prediction effect score and the sample effect score.

8. A page configuration conflict resolution device, characterized in that, The device includes: The graph construction module is used to construct a configuration knowledge graph based on the configuration items and configuration item rules displayed on the target page; wherein, the configuration item rules are the constraints that the configuration items must follow; the configuration knowledge graph includes nodes for representing each configuration item, and for any two configuration items that have the constraints, the knowledge graph also includes an edge for representing the nodes of the two configuration items, and the edge is used to represent the constraints. The conflict detection module is used to detect topological structures in the configuration knowledge graph that meet preset conflict conditions, and to identify them as conflict topological structures; and to determine the configuration items represented by each node in the conflict topological structure, as the conflict configuration items corresponding to the conflict topological structure. The second prediction module is used to input the conflict configuration items into the target decision tree algorithm model to obtain a conflict resolution scheme; wherein, the target decision tree algorithm model is trained in advance using the method steps of any one of claims 1-3; The second adjustment module is used to adjust the configuration items of the target page according to the conflict resolution scheme.

9. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; A processor, when executing a program stored in memory, implements the steps of the method described in any one of claims 1-3 or 4-6.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the method described in any one of claims 1-3 or 4-6.

Images