Strategy recommendation robot determination method and device, strategy recommendation method and device, equipment and medium

Abstract

The invention discloses a strategy recommendation robot determination method, and the method comprises the steps: screening out screening data from all sample data according to a preset index in a preset index library, and determining a sample feature corresponding to each screening data; performing strategy prediction on all the sample features through a Gini impurity algorithm in a preset robot to obtain prediction strategies corresponding to the screening data; performing release testing on all the prediction strategies through the strategy release platform to obtain test results corresponding to the prediction strategies; all the test results are fed back to the preset robot, and when it is confirmed that the preset robot conducts iterative training according to all the test results and a preset convergence condition is met, the converged preset robot is determined as the strategy recommendation robot. By training the strategy recommendation robot, the strategy meeting the expected target is recommended, the strategy release platform test is linked, and the strategy recommendation efficiency and accuracy of the strategy recommendation robot are improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a strategy recommendation robot determination, strategy recommendation method, device, equipment and medium. Background Art

[0002] With the rapid development of artificial intelligence, filtering out the most relevant strategies for users based on their needs is a core part of Internet services. However, user needs are complex and diverse, which places higher demands on the accuracy of recommended strategies. At present, the recommendation of strategy rules generally relies on experienced strategy personnel to write and update based on experience, which cannot accurately and quickly recommend the strategies required by users, resulting in low efficiency and high requirements for strategy personnel. Therefore, the recommendation process of strategy rules has problems such as low efficiency and long time consumption. Summary of the invention

[0003] The embodiments of the present invention provide a strategy recommendation robot determination method, device, equipment and medium to solve the problems in the prior art of high manual requirements, low efficiency and long time consumption for writing strategies.

[0004] A method for determining a strategy recommendation robot, comprising: Filtering out screening data from all sample data according to preset indicators in a preset indicator library, and determining sample features corresponding to each of the screening data; Performing strategy prediction on all the sample features by using the Gini impurity algorithm in the preset robot to obtain a prediction strategy corresponding to each of the screening data; All the prediction strategies are tested by the strategy release platform to obtain test results corresponding to each prediction strategy; All the test results are fed back to the preset robot, and when it is confirmed that the preset robot is iteratively trained according to all the test results and reaches a preset convergence condition, the preset robot after convergence is determined as the strategy recommendation robot.

[0005] A strategy recommendation robot determination device, comprising: A data screening module, used to screen out screening data from all sample data according to preset indicators in a preset indicator library, and determine sample features corresponding to each of the screening data; A strategy prediction module, used to perform strategy prediction on all the sample features by using the Gini impurity algorithm preset in the robot, and obtain a prediction strategy corresponding to each of the screening data; A release test module, used to perform release tests on all the prediction strategies through a strategy release platform to obtain test results corresponding to each prediction strategy; The iterative convergence module is used to feed back all the test results to the preset robot, and when it is confirmed that the preset robot is iteratively trained according to all the test results and reaches the preset convergence condition, the preset robot after convergence is determined as the strategy recommendation machine.

[0006] A strategy recommendation method, comprising: Acquiring user data, wherein the user data includes at least one target feature; Retrieve a strategy recommendation robot, wherein the strategy recommendation robot is trained by the strategy recommendation robot determination method described above; The user data is input into a strategy recommendation robot, and the strategy recommendation robot performs strategy recommendation on the target features in the user data to obtain a strategy recommendation result.

[0007] A strategy recommendation device, comprising: A data acquisition module, used to acquire user data; A robot calling module, used to call a strategy recommendation robot, wherein the strategy recommendation robot is trained by the strategy recommendation robot determination method described above; The strategy recommendation module is used to input the user into the strategy recommendation robot, and the strategy recommendation robot performs strategy recommendation on the user data to obtain a recommended strategy.

[0008] A computer device comprises a memory, a controller and a computer program stored in the memory and executable on the controller, wherein the controller is used to execute the above-mentioned strategy recommendation robot determination method or strategy recommendation method.

[0009] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the above-mentioned strategy recommendation robot determination method or strategy recommendation method.

[0010] The above-mentioned strategy recommendation robot determination method, device, equipment and medium, the present invention screens out screening data through preset indicators, realizes efficient selection of sample data, avoids mistakes in manual screening, solves the problem of time-consuming and high cost in manual screening, and thus improves the efficiency of determining sample characteristics. By using the Gini impurity algorithm in the preset robot to perform strategy prediction on all sample characteristics, the prediction strategy is obtained, and the strategy recommendation efficiency and accuracy of the strategy recommendation robot are improved. All prediction strategies are released and tested through the strategy release platform to ensure that the prediction strategy meets the requirements and recommend strategies that meet the expected goals, thereby improving the accuracy of strategy recommendations. The preset robot is iteratively trained through the test results to achieve the acquisition of the strategy recommendation robot, ensuring that the strategy recommendation robot has a high accuracy rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0012] Figure 1 is a flow chart of a method for determining a strategy recommendation robot in one embodiment of the present invention; Figure 2 is a flowchart of step S10 of a method for determining a strategy recommendation robot in one embodiment of the present invention; Figure 3 is a flowchart of step S10 of a method for determining a strategy recommendation robot in another embodiment of the present invention; Figure 4 is a flowchart of step S20 of a method for determining a strategy recommendation robot in one embodiment of the present invention; Figure 5 is a flowchart of step S30 of a method for determining a strategy recommendation robot in one embodiment of the present invention; Figure 6 is a flow chart of a strategy recommendation method in one embodiment of the present invention; Figure 7 is a principle block diagram of a strategy recommendation robot determination device in one embodiment of the present invention; Figure 8 is a principle block diagram of a strategy recommendation device in one embodiment of the present invention; Fig. 9 is a schematic diagram of a computer device in one embodiment of the present invention. DETAILED DESCRIPTION

[0013] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0014] In one embodiment, if Figure 1 As shown, a method for determining a strategy recommendation robot is provided, comprising the following steps: S10: Filtering out screening data from all sample data according to preset indicators in a preset indicator library, and determining sample features corresponding to each of the screening data.

[0015] Understandably, the filtered data refers to the sample data that meets the indicator requirements. The sample feature refers to the information in the filtered data that meets the feature requirements. The preset indicator refers to the field information set in advance for filtering data. The preset indicator library refers to the collection of all preset indicators.

[0016] Specifically, a preset indicator library is obtained, and all preset indicators in the preset indicator library are determined. Then, all sample data are obtained, and all sample data are screened by the preset indicators, that is, the preset indicators are matched with the information in the sample data to detect whether the preset indicators exist in the sample data. When the information in the sample data fails to match all the preset indicators, it is determined that the sample data does not have information matching the preset indicators, and the sample data is deleted. When the information in the sample data successfully matches the preset indicators, it is determined that the sample data has information matching the preset indicators, the sample data is retained, and the retained data is determined as the screening data. In this way, the screening data that meets the requirements is screened out from all sample data by all the preset indicators in the preset indicator library. Further, the features of all the screening data are determined, that is, whether the features required for policy recommendation are detected in the screening data, and matching is performed by the preset features and the features in the screening data. When the features corresponding to the preset features are not matched in the screening data, it is determined that there are no features matching the preset features in the screening data. When the features corresponding to the preset features are matched in the screening data, it is determined that there are features matching the preset features in the screening data, and the features are determined as the sample features corresponding to the screening data. In this way, it is determined that the corresponding sample features are screened out from all the screened data.

[0017] S20: Performing strategy prediction on all the sample features through the Gini impurity algorithm in the preset robot to obtain a prediction strategy corresponding to each of the screening data.

[0018] Understandably, the preset robot refers to the model structure selected in advance, including initial parameters and loss function, etc. The prediction strategy refers to the strategy recommended by the preset robot based on sample features, for example, loan strategy, repayment strategy.

[0019] Specifically, the preset robot is obtained, that is, the corresponding preset robot can be obtained through information such as robot ID, robot type or state, and application scenario. Then, all sample features are input into the preset robot, and the strategy prediction is performed on all sample features through the Gini impurity algorithm in the preset robot, that is, the Gini impurity is calculated for all sample features through the preset robot, so as to obtain the sample Gini impurity value corresponding to each sample feature. All sample Gini impurity values ​​are compared, the sample feature corresponding to the smallest sample Gini impurity value is selected, and all sample features corresponding to the smallest Gini impurity value are segmented to obtain feature segmentation points. Sample features of the same type are divided into different subsets through feature segmentation points, and each subset is determined as a node or branch. In this way, each subset is segmented again through feature selection and segmentation until all subsets and newly obtained subsets cannot be segmented anymore, and the results are obtained according to all segmentations, and a strategy decision tree is constructed according to the segmentation order. The combined rule strategy is extracted from each node in the strategy decision tree, and then the risk assessment is performed on all the combined rule strategies, and the risk assessment results are compared with the predicted interception target to screen out the nodes (combined rule strategies) that meet the requirements. Next, all the combined rule strategies corresponding to the nodes that meet the requirements are combined, and the probability of each combined rule strategy is calculated using distributed parallelism. The combined sample strategy with the largest probability is selected to obtain the prediction strategy corresponding to the screened data.

[0020] S30: Testing all the prediction strategies through the strategy release platform to obtain test results corresponding to each prediction strategy.

[0021] Understandably, the test results refer to the evaluation results of various indicators of the prediction strategy, such as accuracy, rationality, satisfaction, etc.

[0022] Specifically, after obtaining the prediction strategy, all prediction strategies are released and tested through the strategy release platform, that is, all prediction strategies are input into the strategy release platform, and each prediction strategy is bypass tested through the bypass test module in the strategy release platform, that is, the prediction strategy is tested in a smaller range to evaluate its potential impact and adjust the test plan. Then, each prediction strategy is released and approved through the release approval module in the strategy release platform, that is, the rationality and accuracy of the prediction strategy are tested to determine whether the prediction strategy can be output. Next, each prediction strategy is compared and tested through the comparison experiment module in the strategy release platform, that is, the prediction strategies of different groups are tested for variables, and the strategies are obtained before and after to detect various strategy indicators, so as to obtain the test results corresponding to each prediction strategy.

[0023] S40: Feeding back all the test results to the preset robot, and upon confirming that the preset robot is iteratively trained according to all the test results and reaches a preset convergence condition, determining the preset robot after convergence as the strategy recommendation robot.

[0024] Understandably, the preset convergence condition refers to the convergence condition set in advance, which can be the condition that the number of iterations reaches the preset number, or the loss value is less than the set threshold. The convergence condition can also be the condition that the loss value is very small and will not decrease after 5,000 calculations. The strategy recommendation robot refers to a robot used to recommend strategies. It integrates intelligent analysis, decision-making and recommendation capabilities, and can automatically recommend the most suitable strategy solution for users or systems based on user behavior data.

[0025] Specifically, after obtaining the test results, all test results are fed back to the preset robot. After obtaining the test results, the preset robot adjusts and optimizes the initial parameters according to the test results to obtain the adjusted preset robot. All sample features are re-input into the adjusted preset robot, so that the adjusted preset robot can re-recommend strategies for all sample features, thereby obtaining new strategies, and retesting the new strategies, as well as feedback results, thereby realizing iterative training of the preset robot, so that the results output by the preset robot can continuously approach the accurate results, making the prediction accuracy higher and higher, until the preset robot reaches the preset convergence condition, and the preset robot after convergence is determined as the strategy recommendation robot.

[0026] In the strategy recommendation robot determination method of the present invention, the screening data is screened out by preset indicators, which realizes the efficient selection of sample data, avoids mistakes in manual screening, solves the problem of long time and high cost in manual screening, and thus improves the efficiency of determining sample features. The strategy prediction of all sample features is performed by the Gini impurity algorithm in the preset robot, which realizes the acquisition of prediction strategies and improves the strategy recommendation efficiency and accuracy of the strategy recommendation robot. All prediction strategies are released and tested through the strategy release platform to ensure that the prediction strategies meet the requirements and recommend strategies that meet the expected goals, thereby improving the accuracy of strategy recommendations. The preset robot is iteratively trained through the test results to realize the acquisition of the strategy recommendation robot and ensure that the strategy recommendation robot has a high accuracy rate.

[0027] In one embodiment, if Figure 2 As shown, in step S10, that is, filtering out the screening data from all the sample data according to the preset indicators in the preset indicator library, includes: S101, matching all preset indicators in the preset indicator library with sample information in the sample data to obtain an indicator matching result.

[0028] S102: When the indicator matching result indicates that there is sample information matching at least one preset indicator in the sample data, the sample data is determined as screening data.

[0029] Understandably, preset indicators refer to the requirements for filtering data set in advance, such as field information, field type, etc. The indicator matching result is used to indicate whether the sample information and the preset indicators are matched successfully. Sample information refers to specific fields, information types, etc. in the sample data.

[0030] Specifically, a preset indicator library and all sample data are obtained, and then all sample data are screened through the preset indicator library, that is, all preset indicators in the preset indicator library are matched with the sample information in each sample data, that is, the sample information in the sample data is matched with all preset indicators respectively, so as to obtain the indicator matching results corresponding to each sample data. Further, all indicator matching results are tested to determine the meaning of each indicator matching result representation, and when the indicator matching result representation does not match the sample information corresponding to the preset indicator in the sample data, the sample data is deleted. When the indicator matching result represents that there is sample information matching at least one preset indicator in the sample data, the sample data is retained and determined as the screening data.

[0031] That is, in this embodiment, the sample data is screened by using preset indicators, which reduces the screening time, avoids errors caused by manual screening, reduces labor costs, and improves the efficiency of screening the sample data.

[0032] In one embodiment, if Figure 3 As shown, in step S10, that is, determining the sample features corresponding to each of the screening data, includes: S103, obtaining a preset screening rule, performing feature matching on the screening feature in the preset screening rule and the data feature in the screening data, and obtaining a feature matching result.

[0033] S104: When the feature matching result indicates that the screening data has at least one data feature matching the screening feature, all the data features matching the screening feature are determined as sample features corresponding to the screening data.

[0034] Understandably, the preset screening rules refer to feature requirements set in advance for screening. Screening features refer to feature requirements for screening data, such as field type features, etc. The feature matching result is used to indicate whether the screening features and data features are matched successfully.

[0035] Specifically, after obtaining the screening data, the preset screening rules are obtained, and the data features in the screening data are matched by the preset screening rules, that is, the screening features in the preset screening rules are matched with the data features in the screening data respectively, that is, by matching the field information between the screening features and the data features, or calculating the similarity between the screening features and the data features, so as to determine whether the screening features and the data features are successfully matched, and the feature matching results corresponding to each screening data can be obtained. Further, all feature matching results are detected to determine the meaning of each feature matching result representation. When the feature matching result representation does not match a data feature corresponding to the screening feature in the screening data, the screening data is deleted. When the feature matching result represents that the screening data has at least one data feature that matches the screening feature, the screening data is retained, and all data features that match the screening feature are determined as sample features corresponding to the screening data.

[0036] That is, in this embodiment, by presetting the screening rules, the matching of data features is achieved, thereby achieving the determination of feature matching results, and then achieving the determination of sample features of the screened data, thereby improving the screening efficiency of sample features and reducing the cost of manual screening.

[0037] In one embodiment, if Figure 4 As shown, in step S20, the strategy prediction is performed on all the sample features by using the Gini impurity algorithm in the preset robot to obtain the prediction strategy corresponding to each of the screening data, including: S201, performing Gini impurity processing on all the sample features corresponding to each of the screening data by the preset robot to obtain a Gini impurity value corresponding to each of the sample types.

[0038] S202, performing segmentation processing on all sample features corresponding to the smallest Gini impurity value to obtain feature segmentation points.

[0039] S203: construct a strategy decision tree according to all the sample features and all the feature segmentation points.

[0040] S204, extracting the combined rule strategy of each node in the strategy decision tree, and performing risk assessment on all the combined rule strategies to obtain a risk assessment result.

[0041] S205: Filter out target rule policies that meet the preset interception target and nodes corresponding to the target rule policies according to the risk assessment result.

[0042] S206, determine the association relationship between the screened nodes. When the association relationship represents that two nodes between adjacent levels are parent-child nodes, delete the target rule strategy corresponding to the child node in the parent-child node, and perform distributed parallel probability calculation on the target rule strategy corresponding to the parent node in the parent-child node to obtain a prediction strategy corresponding to each of the screened data.

[0043] Understandably, Gini impurity refers to the expected error rate of a data item when a certain result from a set is randomly applied to the set. Gini impurity value refers to the result calculated for the sample features. Strategy decision tree refers to a prediction model used to recommend the optimal strategy based on features. Risk assessment result refers to the risk assessment value of each information of the combined rule strategy. Node refers to a branch on the strategy decision tree. Parent-child node refers to a branch that represents two nodes at the upper and lower levels. Combined rule strategy refers to the strategy corresponding to each node.

[0044] Specifically, after obtaining the sample features, all sample features are input into a preset robot, and the preset robot performs Gini impurity processing on all sample features, that is, the sample types and types of sample features corresponding to each screening data are counted, and all sample features are classified according to the sample type to obtain sample features corresponding to each sample type, and Gini impurity calculation is performed on all sample features corresponding to each screening data, so as to obtain the Gini impurity value corresponding to each sample feature. Compare all Gini impurity values, select the minimum Gini impurity value, and screen out the sample features corresponding to the minimum Gini impurity value. Then, feature segmentation is performed on all the sample features screened out, so as to determine the feature segmentation point, and all sample features are divided into different subsets based on the feature segmentation point, each subset corresponding to a node (i.e., branch). Then, the above feature selection and feature segmentation process is repeated for each subset until all subsets cannot be segmented anymore, and then a strategy decision tree is constructed through all the obtained nodes. Further, extract the combined rule strategy of each node in the strategy decision tree, that is, extract a combined rule strategy from each node in the strategy decision tree, and perform risk assessment on all combined rule strategies, that is, perform score assessment on each combined rule strategy through the preset risk assessment standard, so as to obtain the risk assessment result corresponding to each combined rule strategy. Then, obtain the preset interception target, and screen all combined rule strategies through the preset interception target, that is, compare and process the risk assessment result corresponding to each combined rule strategy with the preset interception target. When the risk assessment result meets the preset interception target, determine the combined rule strategy corresponding to the risk assessment result as the target rule strategy, and screen out the node corresponding to the target rule strategy. Next, detect the association relationship between the nodes of adjacent levels. When the association relationship between the two nodes of adjacent levels is a superior-subordinate relationship, determine the two nodes between the adjacent levels as parent-child nodes, and delete the target rule strategy corresponding to the child node in the parent-child node according to the preset processing method. When the association relationship between the two nodes of adjacent levels is not a parent-child node, the node is retained. Then, distributed parallel probability calculation is performed on the target rule strategies corresponding to all retained nodes to obtain probability values, which are compared to screen out the optimal prediction strategy.

[0045] That is, in this embodiment, the calculation of the Gini impurity value is realized, thereby realizing the determination of the feature segmentation point, and then realizing the construction of the strategy decision tree. By extracting the combined rule strategy and performing risk assessment, the risk assessment result is obtained, thereby realizing the screening of the target rule strategy and its corresponding nodes, and then realizing the recommendation of the prediction strategy.

[0046] In one embodiment, if Figure 5As shown, in step S30, all the prediction strategies are released and tested through the strategy release platform to obtain test results corresponding to each prediction strategy, including: S301, performing a bypass test on all the prediction strategies to obtain a bypass test result corresponding to each prediction strategy.

[0047] S302: Perform release approval processing on all the prediction strategies to obtain release approval results corresponding to each prediction strategy.

[0048] S303, performing a comparison experiment on all the prediction strategies to obtain a comparison experiment result corresponding to each prediction strategy; S304: Determine a test result corresponding to the prediction strategy according to the bypass test result, the release approval result and the comparison experiment result.

[0049] Understandably, comparative experiments refer to the use of A / B testing technology to compare different strategies and solutions and evaluate their effectiveness. Comparative experiment results refer to the comparison results between different strategies. Release approval results refer to the detection results of whether the prediction strategy is allowed to be released. Bypass testing is to test the effectiveness of the recommended strategy or algorithm by building a parallel or simulated environment. Bypass test results refer to the results of the prediction strategy to see if it meets the test of various indicators.

[0050] Specifically, after obtaining the recommended strategy, bypass testing is performed on all prediction strategies, that is, bypass testing is performed on each prediction strategy separately, that is, each prediction strategy is tested separately by building a simulation environment, and the accuracy, satisfaction, diversity and other indicators of the prediction strategy are tested in the simulation environment. When the content of the prediction strategy meets the requirements of each indicator, a bypass test result that meets the test requirements is obtained. When the content in the prediction strategy does not meet one or more indicator requirements, a bypass test result that does not meet the test requirements is obtained. Then, all prediction strategies are released for approval, that is, each prediction strategy is reviewed according to the preset policy release rules, that is, the content of the prediction strategy is detected and processed through the content in the policy release rules. When all the contents in the prediction strategy meet the policy release rules, a release approval result that allows recommendation is obtained. When one or more contents in the prediction strategy do not meet the policy release rules, a release approval result that does not allow recommendation is obtained. Further, all prediction strategies are subjected to comparison experiment processing, that is, at least one historical strategy scheme corresponding to each sample feature is obtained, and the prediction strategy corresponding to the same sample feature and the historical strategy are compared and processed, so as to obtain the comparison experiment result. Next, the test result corresponding to the prediction strategy is determined based on the bypass test result, the release approval result, and the comparison experiment result, that is, the bypass test result, the release approval result, and the comparison experiment result corresponding to the same prediction strategy are obtained. When all three results indicate passing, the test result that allows the recommended strategy is obtained. When any one of the three results indicates failing, the test result that does not allow the recommended strategy is obtained.

[0051] That is, by bypassing the test of each prediction strategy, the prediction strategy is evaluated, ensuring that the prediction strategy meets the index requirements. By approving the release of each prediction strategy, the prediction strategy is tested, ensuring that the released prediction strategy meets the approval rules. By conducting comparative experiments on the prediction strategies, the comparison of different strategies is achieved, thereby achieving the determination of the comparative experimental results, and then achieving the determination of the test results, ensuring that the recommendation ability of the preset robot is continuously improved.

[0052] In one embodiment, if Figure 6 As shown, a strategy recommendation method is provided, comprising the following steps: S11, acquiring user data, wherein the user data includes at least one target feature.

[0053] S12, calling a strategy recommendation robot, wherein the strategy recommendation robot is trained by the above-mentioned strategy recommendation robot determination method.

[0054] S13, inputting the user data into a strategy recommendation robot, and having the strategy recommendation robot perform strategy recommendation on the target features in the user data to obtain a strategy recommendation result.

[0055] Understandably, user data refers to user behavior data, such as the amount of loans in the past month, the number of successful deductions in the past two months, etc. Target features refer to relevant behavior information in user data, such as loan repayment information in the past month, etc. Strategy recommendation results refer to the strategy planning recommended to the user, such as loan strategy or repayment strategy, etc.

[0056] Specifically, user data including at least one target feature is obtained. Then, a policy recommendation robot is retrieved and the user data is input into the policy recommendation robot. The policy recommendation robot makes policy recommendations for the target features in the user data using the policy recommendation capabilities learned during training, that is, through the policy decision tree in the policy recommendation robot, each target feature is matched with a node in the policy decision tree. When the match is successful, the corresponding policy information is obtained from the node that successfully matches the target feature, and a risk assessment is performed on all policy information, that is, the policy information is compared with the preset interception target, so as to screen out nodes that meet the requirements. Next, all policies corresponding to the nodes that meet the requirements are combined, and the probability of each policy combination is calculated by distributed parallel calculation, that is, the interception rate of each policy combination is calculated, so as to select the combination strategy with the highest probability, and obtain the policy recommendation result corresponding to the user data.

[0057] That is, in this embodiment, the strategy recommendation robot performs strategy recommendation on user data, thereby recommending strategies that meet the expected goals to users, achieving more precise strategy recommendations, and thereby improving the strategy recommendation efficiency and accuracy of the strategy recommendation robot.

[0058] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0059] In one embodiment, a strategy recommendation device is provided, which corresponds one-to-one to the strategy recommendation method in the above embodiment. Figure 8 As shown, the strategy recommendation device includes a data acquisition module 11, a robot retrieval module 12 and a strategy recommendation module 13. The functional modules are described in detail as follows: A data acquisition module 11 is used for the data acquisition module to acquire user data, wherein the user data includes at least one target feature; A robot calling module 12, used for the robot calling module, is used to call a strategy recommendation robot, wherein the strategy recommendation robot is trained by the strategy recommendation robot determination method described above; The strategy recommendation module 13 is used to input the user data into a strategy recommendation robot, and the strategy recommendation robot performs strategy recommendation on the target features in the user data to obtain a strategy recommendation result.

[0060] In one embodiment, a strategy recommendation robot determination device is provided, and the strategy recommendation robot determination device corresponds to the strategy recommendation robot determination method in the above embodiment. Figure 7 As shown, the strategy recommendation robot determination device includes a data screening module 10, a strategy prediction module 20, a release test module 30 and an iterative convergence module 40. The functional modules are described in detail as follows: The data screening module 10 is used to screen out screening data from all sample data according to preset indicators in the preset indicator library, and determine sample features corresponding to each of the screening data; A strategy prediction module 20 is used to perform strategy prediction on all the sample features by using a Gini impurity algorithm preset in the robot to obtain a prediction strategy corresponding to each of the screening data; A release test module 30, used to perform release tests on all the prediction strategies through a strategy release platform to obtain test results corresponding to each prediction strategy; The iterative convergence module 40 is used to feed back all the test results to the preset robot, and when it is confirmed that the preset robot is iteratively trained according to all the test results and reaches a preset convergence condition, determine the preset robot after convergence as the strategy recommendation robot.

[0061] Optionally, the strategy prediction module 20 includes: A Gini impurity processing unit, configured to perform Gini impurity processing on all the sample features corresponding to each of the screening data through the preset robot to obtain a Gini impurity value corresponding to each of the sample types; A feature segmentation processing unit, used for segmenting the sample feature corresponding to the minimum Gini impurity value to obtain a feature segmentation point; A decision tree construction unit, used for constructing a strategy decision tree according to all the sample features and all the feature segmentation points; A risk assessment unit, used for extracting the combined rule strategy of each node in the strategy decision tree, and performing risk assessment on all the combined rule strategies to obtain a risk assessment result; A policy node screening unit, used to screen out target rule policies that meet preset interception targets and nodes corresponding to the target rule policies according to the risk assessment results; A strategy prediction unit is used to determine the association relationship between the screened nodes. When the association relationship represents that two nodes between adjacent levels are parent-child nodes, the target rule strategy corresponding to the child node in the parent-child node is deleted, and the target rule strategy corresponding to the parent node in the parent-child node is subjected to distributed parallel probability calculation to obtain a prediction strategy corresponding to each of the screened data.

[0062] Optionally, the release test module 30 includes: A bypass test unit, used to perform a bypass test on all the prediction strategies to obtain a bypass test result corresponding to each prediction strategy; A release approval processing unit, used to perform release approval processing on all the forecast strategies and obtain a release approval result corresponding to each forecast strategy; A comparison experiment processing unit, used for performing comparison experiment processing on all the prediction strategies to obtain comparison experiment results corresponding to each prediction strategy; A test result determination unit is used to determine the test result corresponding to the prediction strategy according to the bypass test result, the release approval result and the comparison experiment result.

[0063] Optionally, the data screening module 10 includes: An indicator matching unit, used for matching all preset indicators in the preset indicator library with the sample information in the sample data to obtain an indicator matching result; The screening data unit is configured to determine the sample data as screening data when the indicator matching result indicates that there is sample information matching at least one preset indicator in the sample data.

[0064] Optionally, the data screening module 10 further includes: A feature matching unit, used to obtain a preset screening rule, perform feature matching on the screening feature in the preset screening rule and the data feature in the screening data, and obtain a feature matching result; The sample feature unit is used to determine all the data features matching the screening feature as sample features corresponding to the screening data when the feature matching result indicates that the screening data has at least one data feature matching the screening feature.

[0065] This application further provides Fig. 9 A computer device is shown, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor is used to execute the strategy recommendation robot determination method or the strategy recommendation method.

[0066] For the specific definition of computer equipment, processors and their units and modules, please refer to the definition of the strategy recommendation robot determination method mentioned above, which will not be repeated here. Each module in the above-mentioned processor can be implemented in whole or in part by software, hardware and a combination thereof. Understandably, the processor includes a processor, a memory, a network interface and a database connected by a device bus. Each module of the processor can be embedded in or independent of the processor in hardware form, or can be stored in the memory in software form, so that the processor can call and execute the operations corresponding to the above modules. Among them, the processor is used to provide computing and control capabilities. The memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating device, a computer program and a database. The internal memory provides an environment for the operation of the operating device and the computer program in the non-volatile storage medium. The database is used to store the data used by the strategy recommendation robot determination method in the above-mentioned embodiment. The network interface is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a strategy recommendation robot determination method or the described strategy recommendation method is implemented.

[0067] In one embodiment, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned strategy recommendation robot determination method is implemented.

[0068] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0069] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0070] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A method for determining a strategy recommendation robot, characterized in that: include: Filtering out screening data from all sample data according to preset indicators in a preset indicator library, and determining sample features corresponding to each of the screening data; Performing strategy prediction on all the sample features by using the Gini impurity algorithm in the preset robot to obtain a prediction strategy corresponding to each of the screening data; All the prediction strategies are tested by the strategy release platform to obtain test results corresponding to each prediction strategy; All the test results are fed back to the preset robot, and when it is confirmed that the preset robot is iteratively trained according to all the test results and reaches a preset convergence condition, the preset robot after convergence is determined as the strategy recommendation robot.

2. The strategy recommendation robot determination method according to claim 1, characterized in that: The method of performing strategy prediction on all the sample features by using the Gini impurity algorithm in the preset robot to obtain a prediction strategy corresponding to each of the screening data includes: Performing Gini impurity processing on all the sample features corresponding to each of the screening data by the preset robot to obtain a Gini impurity value corresponding to each of the sample types; Segmenting the sample feature corresponding to the smallest Gini impurity value to obtain a feature segmentation point; Constructing a strategy decision tree according to all the sample features and all the feature segmentation points; Extracting the combined rule strategy of each node in the strategy decision tree, and performing risk assessment on all the combined rule strategies to obtain a risk assessment result; Filter out target rule strategies that meet preset interception targets and nodes corresponding to each target rule strategy according to the risk assessment result; Determine the association relationship between the screened nodes. When the association relationship represents that two nodes between adjacent levels are parent-child nodes, delete the target rule strategy corresponding to the child node in the parent-child node, and perform distributed parallel probability calculation on the target rule strategy corresponding to the parent node in the parent-child node to obtain a prediction strategy corresponding to each of the screened data.

3. The strategy recommendation robot determination method according to claim 1, characterized in that: The release test of all the prediction strategies is carried out through the strategy release platform to obtain the test results corresponding to each prediction strategy, including: Performing bypass testing on all the prediction strategies to obtain bypass test results corresponding to each prediction strategy; Performing release approval processing on all the forecasting strategies to obtain release approval results corresponding to each forecasting strategy; Performing comparative experimental processing on all the prediction strategies to obtain comparative experimental results corresponding to each prediction strategy; The test result corresponding to the prediction strategy is determined according to the bypass test result, the release approval result and the comparison experiment result.

4. The strategy recommendation robot determination method according to claim 1, characterized in that: The step of filtering out the filtered data from all the sample data according to the preset indicators in the preset indicator library includes: Matching all preset indicators in the preset indicator library with the sample information in the sample data to obtain an indicator matching result; When the indicator matching result indicates that there is sample information matching at least one preset indicator in the sample data, the sample data is determined as screening data.

5. The strategy recommendation robot determination method according to claim 1, characterized in that: The determining of the sample features corresponding to each of the screening data comprises: Obtaining a preset screening rule, performing feature matching on the screening feature in the preset screening rule and the data feature in the screening data, and obtaining a feature matching result; When the feature matching result indicates that the screening data has at least one data feature matching the screening feature, all the data features matching the screening feature are determined as sample features corresponding to the screening data.

6. A strategy recommendation method, characterized in that: include: Acquiring user data, wherein the user data includes at least one target feature; Retrieve a strategy recommendation robot, wherein the strategy recommendation robot is trained by the strategy recommendation robot determination method according to any one of claims 1 to 5; The user data is input into a strategy recommendation robot, and the strategy recommendation robot performs strategy recommendation on the target features in the user data to obtain a strategy recommendation result.

7. A strategy recommendation robot determination device, characterized in that: include: A data screening module, used to screen out screening data from all sample data according to preset indicators in a preset indicator library, and determine sample features corresponding to each of the screening data; A strategy prediction module, used to perform strategy prediction on all the sample features by using the Gini impurity algorithm preset in the robot, and obtain a prediction strategy corresponding to each of the screening data; A release test module, used to perform release tests on all the prediction strategies through a strategy release platform to obtain test results corresponding to each prediction strategy; The iterative convergence module is used to feed back all the test results to the preset robot, and when it is confirmed that the preset robot is iteratively trained according to all the test results and reaches the preset convergence condition, the preset robot after convergence is determined as the strategy recommendation machine.

8. A strategy recommendation device, characterized in that: include: A data acquisition module, used to acquire user data, wherein the user data includes at least one target feature; A robot calling module, used to call a strategy recommendation robot, wherein the strategy recommendation robot is trained by the strategy recommendation robot determination method according to any one of claims 1 to 5; The strategy recommendation module is used to input the user data into a strategy recommendation robot, and the strategy recommendation robot performs strategy recommendation on the target features in the user data to obtain a strategy recommendation result.

9. A computer device, characterized in that: It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor is used to execute the strategy recommendation robot determination method as described in any one of claims 1 to 5 or the strategy recommendation method as described in claim 6.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the strategy recommendation robot determination method according to any one of claims 1 to 5 or the strategy recommendation method according to claim 6 is implemented.

Images