Risk control method and system, storage medium and electronic equipment
By performing variable screening and algorithm model evaluation on risk assessment project data, combining the comprehensive evaluation of implementation costs and predictive effects, selecting the optimal risk control model, the problems of blindness and cost neglect in the existing technology are solved, and more scientific and efficient risk management is achieved.
Patent Information
- Application Number
- CN202510040238.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-10
- Publication Date
- 2025-05-06
AI Technical Summary
When selecting a risk control model, the prior art ignores variable factors that may appear in actual operations, resulting in blindness in model selection, and only considers the prediction effect and ignores implementation costs, which makes financial institutions unable to achieve the optimal effect of risk management.
By obtaining the project data to be evaluated for risk, performing variable screening to obtain the set of in-modulated variables, determining the algorithmic model used to evaluate the cost of implementation, and determining the predictive effect indicators to be evaluated. Then, an evaluation function is constructed to comprehensively evaluate the implementation cost and prediction effect, and the algorithm model with the highest comprehensive ranking is selected as the final model for risk assessment.
A more comprehensive model evaluation is achieved, taking into account the implementation cost and prediction effect during risk control, and selecting the optimal risk control model through a data-driven method, improving the scientificity and efficiency of decision-making.
Smart Images
Figure CN119940934A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of risk control technology, and in particular to a risk control method, system, storage medium, and electronic equipment. Background Art
[0002] With the development of big data and artificial intelligence technology, risk control models have been widely used in various industries. However, in practical applications, how to select the optimal risk control model to achieve the best risk control effect and cost control is a major challenge.
[0003] At present, the industry's evaluation of the effectiveness of risk control models mainly relies on historical data testing, ignoring the variables that may appear in actual operations, resulting in a certain degree of blindness in model selection; at the same time, the existing model selection methods often only consider the model's prediction effect, while ignoring the implementation cost. This may lead financial institutions to choose models with high prediction performance but high costs, thus failing to achieve the optimal effect of risk management. Summary of the invention
[0004] In view of the deficiencies of the prior art, the purpose of the present invention is to provide a risk control method, aiming to solve the technical problems mentioned in the background technology.
[0005] In order to achieve the above object, the present invention is implemented by the following technical solutions:
[0006] A risk control method comprises the following steps:
[0007] S10, obtaining project data that needs to be risk assessed, and performing variable screening on the project data to obtain a model input variable set;
[0008] S20, determining an algorithm model for evaluating the implementation cost, and determining the prediction effect indicators to be evaluated;
[0009] S30, determining initial model input variables from the model input variable set;
[0010] S40, inputting the initial model input variables into the algorithm model, and then calculating the values of the implementation cost and the prediction effect index;
[0011] S50, constructing an evaluation function for comprehensively evaluating the implementation cost and the prediction effect, and obtaining a comprehensive ranking of all the algorithm models through the evaluation function based on the values of the implementation cost and the prediction effect index calculated by each algorithm model;
[0012] S60, selecting the algorithm model with the highest comprehensive ranking as the final algorithm model, and performing risk assessment through the final algorithm model.
[0013] Furthermore, the variable screening is performed through a machine learning algorithm, and the input model variable set is composed of a plurality of single input model variable sets, and each of the single input model variables has a corresponding product and the price of the product.
[0014] Furthermore, the algorithm models include logistic regression model, support vector machine model, xgb, Lightgbm, and neural network model, and the prediction effect indicators include KS, AUC, accuracy, recall rate, and F1 value.
[0015] Furthermore, the specific steps of step S30 include:
[0016] The single entry model variable is selected by using the forward stepwise regression method;
[0017] The single model entry variable selected in the first several rounds is used as the initial model entry variable.
[0018] Furthermore, the algorithm model includes a training set for calculating the implementation cost and a test set for calculating the value of the prediction effect index. The specific steps of step S40 include:
[0019] Input the initial model input variable into the algorithm model, and cyclically add all combinations of the remaining single model input variables into the algorithm model;
[0020] Calculate the total price of the product corresponding to the single input model variable added in the current cycle, and obtain the implementation cost through the training set;
[0021] Simultaneously calculating the value of the prediction effect index through the training set and the test set;
[0022] Determine whether the number of single input model variables, the number of products, and the algorithm model in the algorithm model meet the statistical requirements.
[0023] Furthermore, after step 40, the method further includes:
[0024] The implementation costs and the values of the prediction effect indicators calculated in all the algorithm models are sorted respectively, so as to obtain the P of each algorithm model. score Cost score, M score Prediction effect index score;
[0025] The values of the prediction effect indicators of the training set and the test set in the algorithm model are subtracted and the absolute value is taken to obtain the difference, and the difference values in all the algorithm models are sorted to obtain the DIFF of each algorithm model. score Difference score.
[0026] Furthermore, the specific steps of step S50 include:
[0027] Determine the P according to actual needs score Cost score, the M socpe Prediction effect index score, and the DIFF score The weight of the difference score;
[0028] An evaluation function for comprehensively evaluating the implementation cost and the prediction effect is constructed, and the formula is as follows:
[0029] P M(X )=Σ(α*P score +β*M score +γ*DIFF score )*T
[0030] Where α is the P score The weight of the cost score, β is the M score The weight of the prediction effect index score, γ is the DIFF score The weight of the difference score, T, is whether the algorithm model meets the statistical requirements, with a value of 1 if it meets the requirements and a value of 0 if it does not meet the requirements;
[0031] The P calculated by each algorithm model score Cost score, M score Prediction effect index score and DIFF score Substituting the difference into the evaluation function to obtain a comprehensive score of the algorithm model;
[0032] The comprehensive scores of each algorithm model are sorted to obtain a comprehensive ranking of each algorithm model.
[0033] The present invention also provides a risk control system, comprising:
[0034] Acquisition module: used to acquire the project data that needs to be risk assessed, and obtain the input variable set after variable screening of the project data;
[0035] In the acquisition module, the variable screening is performed by a machine learning algorithm, and the input model variable set is composed of a plurality of single input model variable sets, each of which has a corresponding product and the price of the product;
[0036] Screening module: used to determine the algorithm model used to evaluate the implementation cost and determine the prediction effect indicators to be evaluated;
[0037] In the screening module, the algorithm models include logistic regression model, support vector machine model, xgb, Lightgbm, and neural network model, and the prediction effect indicators include KS, AUC, accuracy, recall, and F1 value;
[0038] Determining module: used for determining initial model input variables from the model input variable set;
[0039] The determination module is specifically used for:
[0040] The single entry model variable is selected by using the forward stepwise regression method;
[0041] The single model input variable selected in the previous rounds is used as the initial model input variable;
[0042] Calculation module: used for inputting the initial model input variables into the algorithm model, and then obtaining the statistical values of the implementation cost and the prediction effect index;
[0043] The calculation module is specifically used for:
[0044] Input the initial model input variable into the algorithm model, and cyclically add all combinations of the remaining single model input variables into the algorithm model;
[0045] Calculate the total price of the product corresponding to the single input model variable added in the current cycle, and obtain the implementation cost through the training set;
[0046] Simultaneously calculating the value of the prediction effect index through the training set and the test set;
[0047] Determine whether the number of single input model variables, the number of products, and the algorithm model in the algorithm model meet the statistical requirements;
[0048] Sorting module: used to sort the implementation costs and the values of the prediction effect indicators calculated in all the algorithm models, and then obtain the P of each algorithm model. score Cost score, M score Prediction effect index score;
[0049] The values of the prediction effect indicators of the training set and the test set in the algorithm model are subtracted and the absolute value is taken to obtain the difference, and the difference values in all the algorithm models are sorted to obtain the DIFF of each algorithm model. score Difference score;
[0050] Evaluation module: used to construct an evaluation function for comprehensively evaluating the implementation cost and the prediction effect, and based on the values of the implementation cost and the prediction effect index calculated by each algorithm model, obtain a comprehensive ranking of all the algorithm models through the evaluation function;
[0051] The evaluation module is specifically used for:
[0052] Determine the P according to actual needs score Cost score, the M socre Prediction effect index score, and the DIFF score The weight of the difference score;
[0053] An evaluation function for comprehensively evaluating the implementation cost and the prediction effect is constructed, and the formula is as follows:
[0054] P M(X) =Σ(α*P score +β*M score +γ*DIFF soore )*T
[0055] Where α is the P score The weight of the cost score, β is the M score The weight of the prediction effect index score, γ is the DIFF score The weight of the difference score, T, is whether the algorithm model meets the statistical requirements, with a value of 1 if it meets the requirements and a value of 0 if it does not meet the requirements;
[0056] The P calculated by each algorithm model socre Cost score, M score Prediction effect index score and DIFF score Substituting the difference into the evaluation function to obtain a comprehensive score of the algorithm model;
[0057] Sorting the comprehensive scores of each algorithm model to obtain a comprehensive ranking of each algorithm model;
[0058] Selection module: used to select the algorithm model with the highest comprehensive ranking as the final algorithm model, and perform risk assessment through the final algorithm model.
[0059] Furthermore, the present invention also provides a storage medium on which a computer program is stored, and when the computer program is executed by a processor, the risk control method as described above is implemented.
[0060] Furthermore, the present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the risk control method as described above when executing the computer program.
[0061] Compared with the prior art, the present invention has the following beneficial effects:
[0062] By screening the variables of the project data to be risk-assessed and obtaining the modeling variable set, the data with low persuasiveness can be effectively eliminated, thereby retaining the better data. By screening the algorithm model used to evaluate the implementation cost and determining the prediction effect index to be evaluated, the implementation cost and prediction effect during risk control are comprehensively considered, thereby achieving a more comprehensive model evaluation. Through steps S30 and S40, the implementation cost of the algorithm model under each combination of modeling variables and the value of the prediction effect index can be efficiently cycled out, each algorithm model is run in turn, and an evaluation function is established. Based on the implementation cost of each algorithm model and the value of the prediction effect index, the optimal algorithm model to be used in this risk assessment can be obtained through the evaluation function.
[0063] Compared with the judgment method of the prior art that relies on experience or intuition to make decisions, the present invention is more scientific by selecting models based on data collection and analysis; the present invention realizes global model selection through evaluation functions, while the prior art may only focus on the performance optimization of a single model; the present invention can dynamically adjust the model selection strategy according to the market environment and data changes, while the prior art may lack such flexibility. BRIEF DESCRIPTION OF THE DRAWINGS
[0064] Figure 1 This is a flow chart of the risk control method in the first embodiment of the present invention;
[0065] Figure 2 It is a structural block diagram of a risk control system in a second embodiment of the present invention;
[0066] Figure 3 is a structural block diagram of a computer device in a third embodiment of the present invention;
[0067] The following specific implementation manner will further illustrate the present invention in conjunction with the above-mentioned drawings. DETAILED DESCRIPTION
[0068] In order to facilitate the understanding of the present invention, the present invention will be described more fully below with reference to the relevant drawings. Several embodiments of the present invention are given in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive.
[0069] It should be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be a central element. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be a central element at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only.
[0070] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.
[0071] Embodiment 1
[0072] See also Figure 1 , which is a risk control method in the first embodiment of the present invention, comprising the following steps S10 to S60:
[0073] S10, obtaining project data that needs to be risk assessed, and performing variable screening on the project data to obtain a model input variable set;
[0074] S20, determining an algorithm model for evaluating the implementation cost, and determining the prediction effect indicators to be evaluated;
[0075] S30, determining initial model input variables from the model input variable set;
[0076] S40, inputting the initial model input variables into the algorithm model, and then calculating the values of the implementation cost and the prediction effect index;
[0077] S50, constructing an evaluation function for comprehensively evaluating the implementation cost and the prediction effect, and obtaining a comprehensive ranking of all the algorithm models through the evaluation function based on the values of the implementation cost and the prediction effect index calculated by each algorithm model;
[0078] S60, selecting the algorithm model with the highest comprehensive ranking as the final algorithm model, and performing risk assessment through the final algorithm model.
[0079] It can be understood that by screening the variables of the project data to be risk assessed to obtain the model input variable set, the data with low persuasiveness can be effectively eliminated, thereby retaining the better data. By screening the algorithm model used to evaluate the implementation cost and determining the prediction effect index to be evaluated, the implementation cost and prediction effect during risk control are comprehensively considered, thereby achieving a more comprehensive model evaluation. Through steps S30 and S40, the implementation cost of the algorithm model under each combination of model input variables and the value of the prediction effect index can be efficiently cycled out, each algorithm model is run in turn, and an evaluation function is established. Based on the implementation cost of each algorithm model and the value of the prediction effect index, the optimal algorithm model to be used in this risk assessment can be obtained through the evaluation function;
[0080] Compared with the judgment method of the prior art that relies on experience or intuition to make decisions, the present invention is more scientific by selecting models based on data collection and analysis; the present invention realizes global model selection through evaluation functions, while the prior art may only focus on the performance optimization of a single model; the present invention can dynamically adjust the model selection strategy according to the market environment and data changes, while the prior art may lack such flexibility. The technical solution of the present invention can be widely used in risk control decisions in the financial field, such as credit risk assessment, investment portfolio optimization, etc. The present invention helps financial institutions optimize risk management strategies, improve decision-making efficiency and accuracy, and reduce operating costs.
[0081] Furthermore, in step S10, the variable screening is performed by a machine learning algorithm. By screening the single input model variables by the machine learning algorithm, some unconvincing data can be filtered out, thereby leaving data that can better reflect the risk. The input model variable set is composed of a plurality of single input model variable sets, and each of the single input model variables has a corresponding product and the price of the product. In this embodiment, the single input model variables, products, and product prices are shown in the following table:
[0082]
[0083]
[0084] For example, the project data can be regarded as an e-commerce platform, the products are some of the products in the e-commerce platform, and the model input variables are different types of products.
[0085] Furthermore, in step S20, the algorithm model includes a logistic regression model, a support vector machine model, xgb, Lightgbm, and a neural network model, and the prediction effect indicators include KS, AUC, accuracy, recall, and F1 value.
[0086] It is understandable that since different algorithm models have different focuses, the built-in algorithms will be different. By substituting the same set of data into different algorithm models, the results will also be different.
[0087] Exemplarily, in this embodiment, the algorithm model adopts a logistic regression model, and the prediction effect index adopts a KS value.
[0088] Furthermore, the specific steps of step S30 include:
[0089] The single entry model variable is selected by using the forward stepwise regression method;
[0090] The single model entry variable selected in the first several rounds is used as the initial model entry variable.
[0091] Furthermore, the algorithm model includes a training set for calculating the implementation cost and a test set for calculating the value of the prediction effect index. The specific steps of step S40 include:
[0092] Input the initial model input variable into the algorithm model, and cyclically add all combinations of the remaining single model input variables into the algorithm model;
[0093] Calculate the total price of the product corresponding to the single input model variable added in the current cycle, and obtain the implementation cost through the training set;
[0094] Simultaneously calculating the value of the prediction effect index through the training set and the test set;
[0095] Determine whether the number of single input model variables, the number of products, and the algorithm model in the algorithm model meet the statistical requirements.
[0096] Exemplarily, in this embodiment, the single input variables x2, x5, x8, x12, x15, x17, and x20 selected in the previous rounds are used as initial input variables, and then all combinations of the remaining 13 variables are cyclically added based on the initial input variables. At the same time, the total price of the product corresponding to the single input variable added in the current cycle is calculated, the implementation cost is obtained through the training set, and the number of single input variables, the number of products, and the algorithm model in the algorithm model are statistically determined to determine whether they meet the statistical requirements.
[0097] It can be understood that this method can improve the quality of the single input model variables selected in the first several rounds to a certain extent. When the data set is large, this method can speed up the subsequent iteration time, and then quickly combine all the remaining single input model variables into the algorithm model to obtain the implementation cost and the value of the prediction effect index between the single input model variables in different combinations. However, the disadvantage is that the single input model variables selected in the first several rounds are used as the initial input model variables, which may result in the possibility of selection errors, thereby affecting the accuracy of risk assessment.
[0098] In other embodiments, several single input variables may be selected as initial input variables, and then all combinations of other single input variables may be added cyclically. This method may allow all combinations of single input variables to be used as initial input variables in turn, and the risk assessment may be more accurate. However, the disadvantage is that the calculation time may be longer if the data set is large.
[0099] The above method is used to calculate the price situation of all combinations of single-entry model variables in a project data and the value of the corresponding prediction effect index. At the same time, it can be obtained whether the number of single-entry model variables, the number of products, and the algorithm model in the algorithm model meet the statistical requirements.
[0100] Furthermore, after step 40, the method further includes:
[0101] The implementation costs and the values of the prediction effect indicators calculated in all the algorithm models are sorted respectively, so as to obtain the P of each algorithm model. score Cost score, M score Prediction effect index score;
[0102] The values of the prediction effect indicators of the training set and the test set in the algorithm model are subtracted and the absolute value is taken to obtain the difference, and the difference values in all the algorithm models are sorted to obtain the DIFF of each algorithm model. score Difference score.
[0103] For example, suppose that a total of S algorithm models are built, and the implementation cost calculated by each algorithm model is sorted from high to low, and the top P soore Cost score is 1, and the second-ranked P soore The cost score is 2, and so on, the last ranked P score The cost score is S. Similarly, the values of the prediction effect indicators calculated by each algorithm model are ranked from worst to best. The worst ranked M soore The prediction effect index score is 1, and the second worst is M score The prediction effect index score is 2, and so on. The best ranking M score The score of the prediction effect index is S;
[0104] Then, the values of the prediction effect indicators of the training set and the test set in the algorithm model are subtracted and the absolute value is taken to obtain the difference, and the difference values in all the algorithm models are sorted, and the top DIFF seove The difference score is 1, ranking second in DIFF seove The difference is divided into 2, and so on, the last DIFF score The difference is scored as S.
[0105] Furthermore, the specific steps of step S50 include:
[0106] Determine the P according to actual needs score Cost score, the M socre Prediction effect index score, and the DIFF score The weight of the difference score;
[0107] An evaluation function for comprehensively evaluating the implementation cost and the prediction effect is constructed, and the formula is as follows:
[0108] P M(X) =∑(α*P score +β*M score +γ*DIFF score )*R
[0109] Where α is the P score The weight of the cost score, β is the M score The weight of the prediction effect index score, γ is the DIFF score The weight of the difference score, T, is whether the algorithm model meets the statistical requirements, with a value of 1 if it meets the requirements and a value of 0 if it does not meet the requirements;
[0110] The P calculated by each algorithm model score Cost score, M score Prediction effect index score and DIFF score Substituting the difference into the evaluation function to obtain a comprehensive score of the algorithm model;
[0111] The comprehensive scores of each algorithm model are sorted to obtain a comprehensive ranking of each algorithm model.
[0112] It should be noted that new parameters can be added to the evaluation function according to the actual needs of users.
[0113] In summary, compared with the judgment method of the prior art that relies on experience or intuition to make decisions, a risk control method in the above-mentioned embodiment of the present invention is more scientific by selecting a model based on data collection and analysis; the present invention realizes global model selection through an evaluation function, while the prior art may only focus on the performance optimization of a single model; the present invention can dynamically adjust the model selection strategy according to the market environment and data changes, while the prior art may lack such flexibility.
[0114] Embodiment 2
[0115] Please refer to Figure 2 , shown is a risk control system 40 in a second embodiment of the present invention, comprising:
[0116] Acquisition module 11: used to acquire project data that need to be risk assessed, and obtain a model input variable set after performing variable screening on the project data;
[0117] In the acquisition module 11, the variable screening is performed by a machine learning algorithm, and the input model variable set is composed of a plurality of single input model variable sets, each of which has a corresponding product and the price of the product;
[0118] Screening module 12: used to determine the algorithm model used to evaluate the implementation cost and determine the prediction effect indicators to be evaluated;
[0119] In the screening module 12, the algorithm models include logistic regression model, support vector machine model, xgb, Lightgbm, and neural network model, and the prediction effect indicators include KS, AUC, accuracy, recall, and F1 value;
[0120] Determining module 13: used to determine initial model input variables from the model input variable set;
[0121] The determining module 13 is specifically used for:
[0122] The single entry model variable is selected by using the forward stepwise regression method;
[0123] The single model input variable selected in the previous rounds is used as the initial model input variable;
[0124] Calculation module 14: used to input the initial model input variables into the algorithm model, and then obtain the statistical values of the implementation cost and the prediction effect index;
[0125] The calculation module 14 is specifically used for:
[0126] Input the initial model input variable into the algorithm model, and cyclically add all combinations of the remaining single model input variables into the algorithm model;
[0127] Calculate the total price of the product corresponding to the single input model variable added in the current cycle, and obtain the implementation cost through the training set;
[0128] Simultaneously calculating the value of the prediction effect index through the training set and the test set;
[0129] Determine whether the number of single input model variables, the number of products, and the algorithm model in the algorithm model meet the statistical requirements;
[0130] Sorting module: used to sort the implementation costs and the values of the prediction effect indicators calculated in all the algorithm models, and then obtain the P of each algorithm model. scove Cost score, M scove Prediction effect index score;
[0131] The values of the prediction effect indicators of the training set and the test set in the algorithm model are subtracted and the absolute value is taken to obtain the difference, and the difference values in all the algorithm models are sorted to obtain the DIFF of each algorithm model. scove Difference score;
[0132] Evaluation module 15: used to construct an evaluation function for comprehensively evaluating the implementation cost and the prediction effect, and based on the values of the implementation cost and the prediction effect index calculated by each algorithm model, obtain a comprehensive ranking of all the algorithm models through the evaluation function;
[0133] The evaluation module 15 is specifically used for:
[0134] Determine the P according to actual needs score Cost score, the M score Prediction effect index score, and the DIFF scove The weight of the difference score;
[0135] An evaluation function for comprehensively evaluating the implementation cost and the prediction effect is constructed, and the formula is as follows:
[0136] P M(X) =∑(α*P score +β*M score +γ*DIFF score )*T
[0137] Where α is the P score The weight of the cost score, β is the M score The weight of the prediction effect index score, γ is the DIFF score The weight of the difference score, T, is whether the algorithm model meets the statistical requirements, with a value of 1 if it meets the requirements and a value of 0 if it does not meet the requirements;
[0138] The P calculated by each algorithm model score Cost score, M score Prediction effect index score and DIFF score Substituting the difference into the evaluation function to obtain a comprehensive score of the algorithm model;
[0139] Sorting the comprehensive scores of each algorithm model to obtain a comprehensive ranking of each algorithm model;
[0140] Selection module 16: used to select the algorithm model with the highest comprehensive ranking as the final algorithm model, and perform risk assessment through the final algorithm model.
[0141] The functions or operation steps implemented when the above modules and units are executed are generally the same as those in the above method embodiments, and will not be repeated here.
[0142] The risk control system provided in the embodiment of the present invention has the same implementation principle and technical effects as those of the aforementioned method embodiment. For the sake of brief description, for matters not mentioned in the device embodiment, reference may be made to the corresponding contents in the aforementioned method embodiment.
[0143] Embodiment 3
[0144] The present invention also provides a computer device, see Figure 3 , shown is a computer device in the third embodiment of the present invention, including a memory 10, a processor 20, and a computer program 30 stored in the memory 10 and executable on the processor 20, and the processor 20 implements the above-mentioned risk control method when executing the computer program 30.
[0145] The memory 10 includes at least one type of storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory, etc.), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 10 may be an internal storage unit of a computer device, such as a hard disk of the computer device. In other embodiments, the memory 10 may also be an external storage device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card, etc. Further, the memory 10 may also include both an internal storage unit of a computer device and an external storage device. The memory 10 may be used not only to store application software and various types of data installed in the computer device, but also to temporarily store data that has been output or is to be output.
[0146] Among them, in some embodiments, the processor 20 can be an electronic control unit (Electronic Control Unit, abbreviated as ECU, also known as a vehicle computer), a central processing unit (Central Processing Unit, CPU), a controller, a microcontroller, a microprocessor or other data processing chip, used to run the program code stored in the memory 10 or process data, such as executing access restriction programs, etc.
[0147] It should be pointed out that Figure 3 The structure shown does not constitute a limitation on the computer device. In other embodiments, the computer device may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.
[0148] An embodiment of the present invention further provides a readable storage medium having a computer program stored thereon, and the computer program implements the risk control method as described above when executed by a processor.
[0149] Those skilled in the art will appreciate that the logic and / or steps represented in the flowchart or otherwise described herein, for example, may be considered as an ordered list of executable instructions for implementing logical functions, and may be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For purposes of this specification, a "computer-readable medium" may be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.
[0150] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.
[0151] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or a combination thereof: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.
[0152] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.
[0153] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.
Claims
1. A risk control method, characterized in that: The steps include: S10, obtaining project data that needs to be risk assessed, and performing variable screening on the project data to obtain a model input variable set; S20, determining an algorithm model for evaluating the implementation cost and determining the prediction effect indicators to be evaluated; S30, determining initial model input variables from the model input variable set; S40, inputting the initial model input variables into the algorithm model, and then calculating the values of the implementation cost and the prediction effect index; S50, constructing an evaluation function for comprehensively evaluating the implementation cost and the prediction effect, and obtaining a comprehensive ranking of all the algorithm models through the evaluation function based on the values of the implementation cost and the prediction effect index calculated by each algorithm model; S60, selecting the algorithm model with the highest comprehensive ranking as the final algorithm model, and performing risk assessment through the final algorithm model.
2. The risk control method according to claim 1, characterized in that: The variable screening is performed through a machine learning algorithm, and the input model variable set is composed of a plurality of single input model variable sets, and each of the single input model variables has a corresponding product and the price of the product.
3. The risk control method according to claim 1, characterized in that: The algorithm models include logistic regression model, support vector machine model, xgb, Lightgbm, and neural network model, and the prediction effect indicators include KS, AUC, accuracy, recall rate, and F1 value.
4. The risk control method according to claim 2, characterized in that: The specific steps of step S30 include: The single entry model variable is selected by using the forward stepwise regression method; The single model entry variable selected in the first several rounds is used as the initial model entry variable.
5. The risk control method according to claim 4, characterized in that: The algorithm model includes a training set for calculating the implementation cost and a test set for calculating the value of the prediction effect index. The specific steps of step S40 include: Input the initial model input variable into the algorithm model, and cyclically add all combinations of the remaining single model input variables into the algorithm model; Calculate the total price of the product corresponding to the single input model variable added in the current cycle, and obtain the implementation cost through the training set; Simultaneously calculating the value of the prediction effect index through the training set and the test set; Determine whether the number of single input model variables, the number of products, and the algorithm model in the algorithm model meet the statistical requirements.
6. The risk control method according to claim 5, characterized in that: After step 40, the method further includes: The implementation costs and the values of the prediction effect indicators calculated in all the algorithm models are sorted respectively, so as to obtain the P of each algorithm model. score Cost score, M score Prediction effect index score; The values of the prediction effect indicators of the training set and the test set in the algorithm model are subtracted and the absolute value is taken to obtain the difference, and the difference values in all the algorithm models are sorted to obtain the DIFF of each algorithm model. score Difference score.
7. The risk control method according to claim 6, characterized in that: The specific steps of step S50 include: Determine the P according to actual needs score Cost score, the M score Prediction effect index score, and the DIFF score The weight of the difference score; An evaluation function for comprehensively evaluating the implementation cost and the prediction effect is constructed, and the formula is as follows: P M(X) =∑(α*F score +β*M score +γ*DIFF score )*T Where α is the P score The weight of the cost score, β is the M score The weight of the prediction effect index score, γ is the DIFF score The weight of the difference score, T, is whether the algorithm model meets the statistical requirements, with a value of 1 if it meets the requirements and a value of 0 if it does not meet the requirements; The P calculated by each algorithm model score Cost score, M score Prediction effect index score and DIFF score Substituting the difference into the evaluation function to obtain a comprehensive score of the algorithm model; The comprehensive scores of each algorithm model are sorted to obtain a comprehensive ranking of each algorithm model.
8. A risk control system, characterized in that: include: Acquisition module: used to acquire the project data that needs to be risk assessed, and obtain the input variable set after variable screening of the project data; Screening module: used to screen the algorithm models used to evaluate the implementation cost and determine the prediction effect indicators to be evaluated; Determining module: used for determining initial model input variables from the model input variable set; Calculation module: used for inputting the initial model input variables into the algorithm model, thereby obtaining the values of the implementation cost and the prediction effect index; Evaluation module: used to construct an evaluation function for comprehensively evaluating the implementation cost and the prediction effect, and based on the values of the implementation cost and the prediction effect index calculated by each algorithm model, obtain a comprehensive ranking of all the algorithm models through the evaluation function; Selection module: used to select the algorithm model with the highest comprehensive ranking as the final algorithm model, and perform risk assessment through the final algorithm model.
9. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the risk control method according to any one of claims 1 to 7 is implemented.
10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the risk control method according to any one of claims 1 to 7 is implemented.