Risk control rule effect prediction method and system based on big data

Abstract

The invention discloses a risk control rule effect prediction method and system based on big data, and relates to the technical field of artificial intelligence, and the method comprises the steps: collecting historical risk control rule data and real-time behavior data, extracting semantic feature vectors from the historical risk control rule data, and carrying out the clustering analysis, and forming historical clustering insurance types; based on the semantic feature vector and the real-time behavior data, considering the prediction of insurance premium and claim settlement risk, and constructing a risk control rule effect prediction model; according to a prediction result output by the risk control rule effect prediction model, dynamically adjusting a division threshold, dividing risk levels, and dynamically updating a risk customer list; according to the invention, through efficient prediction and management of the risk control rule effect, the accuracy and real-time performance of prediction are improved, the flexibility and intelligent level of the system are enhanced, and the method is of great significance for improving the risk management capability of the insurance industry.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method and system for predicting the effect of risk control rules based on big data. Background Art

[0002] With the rapid development of big data technology and artificial intelligence, risk management in the field of auto insurance claims faces unprecedented challenges and opportunities. In recent years, the insurance industry has increasingly relied on data-driven methods to optimize business processes, especially in the claims process, by analyzing large amounts of historical data and real-time behavioral data to predict potential risks and fraud. However, most of the existing risk control rule effect prediction methods are based on static rules, lacking flexibility and real-time performance, and unable to respond to market changes and fluctuations in individual behavior in a timely manner. In addition, these methods often focus on single-dimensional data analysis, ignoring the more comprehensive risk assessment perspective brought about by multi-source data fusion. Summary of the invention

[0003] In view of the above existing problems, the present invention is proposed.

[0004] Therefore, the present invention provides a risk control rule effect prediction method and system based on big data to solve the problems of insufficient flexibility and inability to respond to market changes and personal behavior fluctuations in real time in existing risk control rule effect prediction methods.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] In a first aspect, an embodiment of the present invention provides a risk control rule effect prediction method based on big data, which includes collecting historical risk control rule data and real-time behavior data, extracting semantic feature vectors from the historical risk control rule data, and performing cluster analysis to form historical clustered insurance types;

[0007] Based on semantic feature vectors and real-time behavioral data, we consider the prediction of premium and claim risks and build a risk control rule effect prediction model;

[0008] Utilize the prediction results output by the risk control rule effect prediction model to dynamically adjust the classification thresholds, classify risk levels, and dynamically update the risk customer list.

[0009] As a preferred solution of the risk control rule effect prediction method based on big data described in the present invention, wherein: extracting semantic features from historical risk control rule data includes the following steps:

[0010] Perform data cleaning, missing value processing and outlier processing on historical risk control rule data, and perform word segmentation on the processed historical risk control rule data to remove stop words;

[0011] Use TF-IDF to calculate the frequency of each word in the document;

[0012] Arrange the TF-IDF values ​​of each word in the document in descending order, select the word with the highest TF-IDF value as the target word, and use the continuous bag of words model to extract semantic features, expressed as:

[0013]

[0014] Among them, v t represents the semantic features of the target word t, W t represents the word set in the context window of the target word t, v w represents the vector representation of the context word w, |W t | represents the number of words in the context window;

[0015] Integrate the semantic features in each document to form a semantic feature vector.

[0016] As a preferred solution of the risk control rule effect prediction method based on big data described in the present invention, clustering analysis of semantic feature vectors refers to obtaining cluster center clusters, analyzing each cluster center cluster according to the clustering results, extracting the insurance type with the highest frequency in the cluster as the representative insurance type in the cluster, and naming the cluster center clusters to form historical clustered insurance types.

[0017] As a preferred solution of the risk control rule effect prediction method based on big data described in the present invention, wherein: based on semantic feature vectors and real-time behavior data, considering the prediction of premium and claim risk, a risk control rule effect prediction model is constructed, including the following steps:

[0018] Standardize the real-time behavior data, extract the real-time behavior features, and construct the real-time behavior feature vector;

[0019] Using semantic feature vectors and real-time behavior feature vectors, considering the prediction of premium and claim risk, we build a risk control rule effect prediction model and predict the effect score, expressed as,

[0020]

[0021] Among them, S represents the prediction effect score, n s Represents the number of semantic feature vectors, n b represents the number of real-time behavior feature vectors, v si Represents the semantic feature vector v s The i-th semantic feature in v bj Represents the real-time behavior feature vector v b The jth real-time behavior feature in s represents the semantic feature vector, v brepresents the real-time behavior feature vector, λ represents the influence adjustment parameter of the semantic feature, i represents the index of the semantic feature, and j represents the index of the real-time behavior feature.

[0022] As a preferred solution of the risk control rule effect prediction method based on big data described in the present invention, wherein: based on the prediction results output by the risk control rule effect prediction model, the average prediction effect score is calculated, including the following steps:

[0023] The average prediction effect score is calculated by combining the prediction effect score with the historical clustering insurance types, which is expressed as:

[0024]

[0025] in, represents the average prediction effect score of the k-th historical cluster insurance type, N k represents the number of samples of the kth historical clustered insurance type, S q represents the prediction effect score of the qth sample, q represents the index of the sample, and k represents the index of the historical clustering insurance type.

[0026] As a preferred solution of the risk control rule effect prediction method based on big data described in the present invention, the average prediction effect score of historical clustered insurance types is used to calculate the division threshold, which is expressed as:

[0027]

[0028] Among them, T1 represents the low risk level threshold, T2 represents the medium risk level threshold, and T3 represents the high risk level threshold. and They respectively represent the average prediction effect score of the first historical cluster insurance type and the average prediction effect score of the second historical cluster insurance type.

[0029] As a preferred solution of the risk control rule effect prediction method based on big data described in the present invention, the risk level is divided based on the calculated division threshold, and the risk customer list is dynamically updated, including the following steps:

[0030] When T1, it means that the predicted risk control rule effect is at a low risk level;

[0031] When T2, it means that the predicted risk control rule effect is at a medium risk level;

[0032] When T3, it means that the predicted risk control rule effect is at a high risk level;

[0033] When T3, it means that the predicted risk control rule effect is at an extremely high risk level;

[0034] For customers with high risk levels, they are added to the risk customer list. The risk customer list is continuously updated according to changes in the classification thresholds. For customers with extremely high risk levels, the risk control team is notified to take action.

[0035] In a second aspect, the present invention provides a risk control rule effect prediction system based on big data, including: a data preparation module, which is responsible for collecting historical risk control rule data and real-time behavior data, extracting semantic feature vectors from the historical risk control rule data, and performing cluster analysis to form historical clustered insurance types;

[0036] The risk control rule effect prediction model building module is responsible for building a risk control rule effect prediction model based on semantic feature vectors and real-time behavior data, taking into account the prediction of premiums and claims risks;

[0037] The risk management module is responsible for using the prediction results output by the risk control rule effect prediction model to dynamically adjust the classification thresholds, divide the risk levels, and dynamically update the risk customer list.

[0038] In a third aspect, an embodiment of the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the risk control rule effect prediction method based on big data as described in the first aspect of the present invention is implemented.

[0039] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the risk control rule effect prediction method based on big data as described in the first aspect of the present invention.

[0040] The beneficial effects of the present invention are as follows: by collecting historical risk control rule data and real-time behavior data, and extracting semantic feature vectors therefrom, clustering analysis is performed to form historical clustered insurance types, thereby achieving effective integration and structuring of historical data, and providing key basic information for the subsequent construction of prediction models. Based on semantic feature vectors and real-time behavior data, a risk control rule effect prediction model is constructed. By calculating the prediction effect score, the impact of historical data and real-time data can be comprehensively considered, thereby improving the accuracy and reliability of the prediction. By using the results output by the prediction model, the division threshold is dynamically adjusted, the risk level is divided, and the risk customer list is dynamically updated, thereby achieving dynamic risk assessment and management based on the prediction results, improving the efficiency and accuracy of risk management, and improving the security and stability of the business. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] 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 will be briefly introduced below. Obviously, the accompanying drawings described below 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 work.

[0042] Figure 1 This is a flow chart of the risk control rule effect prediction method based on big data in Example 1.

[0043] Figure 2 This is a determination diagram of the risk level for predicting the effect of the risk control rules in Example 1. DETAILED DESCRIPTION

[0044] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0045] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0046] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0047] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, and provides a risk control rule effect prediction method based on big data, comprising the following steps:

[0048] S1. Collect historical risk control rule data and real-time behavior data, extract semantic feature vectors from historical risk control rule data, and perform cluster analysis to form historical clustered insurance types.

[0049] The historical risk control rule data includes trigger conditions, execution frequency, applicable scope, and the types of insurance that are compatible with the risk control rule data;

[0050] The real-time behavior data includes user transaction records and driving behavior;

[0051] Perform data cleaning, missing value processing and outlier processing on historical risk control rule data, and perform word segmentation on the processed historical risk control rule data to remove stop words;

[0052] Use TF-IDF to calculate the frequency of each word in the document, expressed as,

[0053]

[0054] TF-IDF(t,d)=TF(t,d)×IDF(t);

[0055] Where IDF(t) represents the inverse document frequency of word t, N represents the total number of documents, df(t) represents the number of documents containing word t, and TF represents IDF(t, d) represents the TF-IDF value of word t in document d;

[0056] The total number of documents and the number of documents containing word t are determined by counting the total number of documents in all historical risk control rule data and the number of documents containing the specific word tt;

[0057] Arrange the TF-IDF values ​​of each word in the document in descending order, select the word with the highest TF-IDF value as the target word, and use the continuous bag of words model to extract semantic features, expressed as:

[0058]

[0059] Among them, v t represents the semantic features of the target word t, W t represents the word set in the context window of the target word t, v w represents the vector representation of the context word w, |W t | represents the number of words in the context window;

[0060] Integrate the semantic features in each document to form a semantic feature vector.

[0061] Cluster analysis of semantic feature vectors refers to obtaining cluster center clusters, analyzing each cluster center cluster according to the clustering results, extracting the insurance type with the highest frequency in the cluster as the representative insurance type in the cluster, and naming the cluster center cluster to form historical cluster insurance types;

[0062] Clustering Algorithm and Number of Clusters The optimal number of clusters can be determined by applying the K-means clustering algorithm and using methods such as the elbow rule or the silhouette coefficient.

[0063] It should be noted that the collection of historical risk control rule data and real-time behavior data has achieved the initial integration of data, providing a rich data foundation for subsequent analysis. By performing data cleaning, missing value processing and outlier processing on the historical risk control rule data, the quality of the data is ensured, providing reliable data support for subsequent analysis. By performing word segmentation processing on the processed historical risk control rule data and removing stop words, the expressiveness of text features is enhanced, which helps to more accurately capture the key information in the text. Using TF-IDF to calculate the word frequency of each word in the document can highlight the important feature words in the text, laying the foundation for subsequent semantic feature extraction. By selecting the word with the highest TF-IDF value as the target word and using the continuous bag of words model to extract semantic features, semantic feature vectors are formed, which can more accurately reflect the core information of the text content. By performing cluster analysis on the semantic feature vectors, historical clustered insurance types are formed, and the structured arrangement of historical risk control rule data is achieved, providing key basic information for the subsequent prediction model construction. This step not only improves the prediction accuracy of the model, but also provides strong support for subsequent real-time behavior data processing and risk assessment.

[0064] S2. Based on semantic feature vectors and real-time behavioral data, consider the prediction of premium and claim risks and build a risk control rule effect prediction model.

[0065] Perform minimum and maximum normalization processing on the real-time behavior data, convert the processed real-time behavior data into integer code, extract real-time behavior features, select features that are highly correlated with the target variable (such as claim risk) by calculating the correlation between the real-time behavior features, determine the dimension of the feature vector according to the number of extracted real-time behavior features, and put the extracted real-time behavior features into a vector in order to form a real-time behavior feature vector;

[0066] Using semantic feature vectors and real-time behavior feature vectors, considering the prediction of premium and claim risk, we build a risk control rule effect prediction model and predict the effect score, expressed as,

[0067]

[0068] Among them, S represents the prediction effect score, n s Represents the number of semantic feature vectors, n b represents the number of real-time behavior feature vectors, v si Represents the semantic feature vector v s The i-th semantic feature in v bj Represents the real-time behavior feature vector v b The jth real-time behavior feature in s represents the semantic feature vector, which contains semantic features related to premium and claim risk, v brepresents the real-time behavior feature vector, which contains features extracted from real-time behavior data, such as user transaction records and driving behavior. λ represents the influence adjustment parameter of semantic features. i represents the index of semantic features. j represents the index of real-time behavior features. Represents the semantic feature vector v s The square of the i-th semantic feature in .

[0069] It should be noted that λ is a regulating parameter used to adjust the influence of the semantic feature vector in the prediction model. Its function is to balance the contribution of the semantic feature vector and the real-time behavior feature vector when calculating the prediction effect score. More specifically, λ controls the weight of the semantic feature vector when calculating the prediction effect score. When λ is large, the influence of the semantic feature is amplified; otherwise, it is weakened. In practical applications, the specific value of λ needs to be adjusted according to the actual situation to achieve the best prediction effect. Usually, the optimal value of λ can be determined by methods such as cross-validation. For example, the value of λ can be adjusted on the validation set to observe the change in the prediction effect score and select the λ value that makes the prediction effect the best. In addition, other methods such as random search can be used to find the optimal value of λ.

[0070] It should be noted that the design basis for constructing the risk control rule effect prediction model formula is to consider the prediction of premium and claim risk, comprehensively evaluate the effect of the newly created risk control rules, and take into account the impact of historical data and real-time data. The score is intended to quantify the prediction performance of the newly created risk control rules to facilitate risk level classification and dynamic adjustment. The design principle is to evaluate the prediction effect of the newly created risk control rules by calculating the combined impact of the semantic feature vector and the real-time behavior feature vector. The design considers the different contributions of the semantic feature vector and the real-time behavior feature vector, and the Euclidean norm (i.e., the length of the vector) is used. This part reflects the strength of the semantic feature vector. The larger the value, the greater the impact of the semantic feature on the prediction effect.

[0071] It represents the sum of each feature in the real-time behavior feature vector after being transformed by the sigmoid function. This part reflects the influence of the real-time behavior feature on the prediction effect. It represents the sum of each feature in the semantic feature vector after being transformed by the sigmoid function and weighted by the adjustment parameter λ. This part reflects the influence of the adjusted semantic features on the prediction effect.

[0072] It should be noted that the prediction model of the effect of risk control rules is constructed by using semantic feature vectors and real-time behavior feature vectors, considering the prediction of premium and claim risks. The real-time behavior data is effectively utilized, and the real-time behavior feature vector is constructed by standardizing and extracting the real-time behavior features, thereby enhancing the real-time and adaptability of the model. The risk control rule effect prediction model constructed using semantic feature vectors and real-time behavior feature vectors can comprehensively consider the influence of historical data and real-time data, and improve the accuracy and reliability of the prediction. By calculating the prediction effect score, the effect of the newly created risk control rules can be quantitatively evaluated, which provides an important basis for the subsequent risk level classification. This step not only improves the prediction ability of the model, but also can respond to market changes and personal behavior fluctuations in a timely manner, enhancing the flexibility and real-time performance of the system.

[0073] S3. Use the prediction results output by the risk control rule effect prediction model to dynamically adjust the classification threshold, divide the risk level, and dynamically update the risk customer list.

[0074] The average prediction effect score is calculated by combining the prediction effect score with the historical clustering insurance types, which is expressed as:

[0075]

[0076] in, represents the average prediction effect score of the k-th historical cluster insurance type, N k represents the number of samples of the kth historical clustered insurance type, S q represents the prediction effect score of the qth sample, q represents the index of the sample, and k represents the index of the historical clustering insurance type.

[0077] The average prediction effect score of historical clustered insurance types is used to calculate the partition threshold, which is expressed as:

[0078]

[0080] Among them, T1 represents the low risk level threshold, T2 represents the medium risk level threshold, and T3 represents the high risk level threshold. and They represent the average prediction effect score of the first historical cluster insurance type and the average prediction effect score of the second historical cluster insurance type respectively;

[0081] Assume that there are three historical clustered insurance types, and the average effect scores are 2.0; 0.6; 1.2;

[0085] 2.0.

[0086] When T1, it means that the predicted risk control rule effect is at a low risk level;

[0087] When T2, it means that the predicted risk control rule effect is at a medium risk level;

[0088] When T3, it means that the predicted risk control rule effect is at a high risk level;

[0089] When T3, it means that the predicted risk control rule effect is at an extremely high risk level;

[0090] For customers with high risk levels, they are added to the risk customer list. The risk customer list is continuously updated according to changes in the classification thresholds. For customers with extremely high risk levels, the risk control team is notified to take action.

[0091] It should be noted that the prediction results output by the risk control rule effect prediction model are used to dynamically adjust the threshold, divide the risk level, and dynamically update the risk customer list. Dynamic risk assessment and management based on the prediction results are realized, which improves the efficiency and accuracy of risk management. By calculating the average prediction effect score and setting different risk level thresholds accordingly, customers of different risk levels can be more accurately identified and classified, providing clear standards for subsequent management measures. The function of dynamically updating the risk customer list can respond to changes in high-risk customers in a timely manner, improving the timeliness of risk prevention and control. This step not only enhances the intelligence level of the system, but also effectively reduces risk losses and improves the security and stability of the business.

[0092] This embodiment also provides a risk control rule effect prediction system based on big data, including: a data preparation module, which is responsible for collecting historical risk control rule data and real-time behavior data, extracting semantic feature vectors from the historical risk control rule data, and performing cluster analysis to form historical clustered insurance types;

[0093] The risk control rule effect prediction model building module is responsible for building a risk control rule effect prediction model based on semantic feature vectors and real-time behavior data, taking into account the prediction of premiums and claims risks;

[0094] The risk management module is responsible for using the prediction results output by the risk control rule effect prediction model to dynamically adjust the classification thresholds, divide the risk levels, and dynamically update the risk customer list.

[0095] This embodiment also provides a computer device, which is suitable for the risk control rule effect prediction method based on big data, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the risk control rule effect prediction method based on big data proposed in the above embodiment.

[0096] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0097] The present embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the risk control rule effect prediction method based on big data as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, disk or optical disk.

[0098] In summary, the present invention collects historical risk control rule data and real-time behavior data, extracts semantic feature vectors therefrom, and performs cluster analysis to form historical clustered insurance types, thereby achieving effective integration and structuring of historical data and providing key basic information for the subsequent construction of prediction models. Based on semantic feature vectors and real-time behavior data, a risk control rule effect prediction model is constructed. By calculating the prediction effect score, the influence of historical data and real-time data can be comprehensively considered, thereby improving the accuracy and reliability of the prediction. By using the results output by the prediction model, the division threshold is dynamically adjusted, the risk level is divided, and the risk customer list is dynamically updated, thereby achieving dynamic risk assessment and management based on the prediction results, improving the efficiency and accuracy of risk management, and improving the security and stability of the business.

[0099] Example 2

[0100] Referring to Table 1, which is the second embodiment of the present invention, in order to further verify the advancement of the present invention, experimental simulation data of the risk control rule effect prediction method based on big data is provided.

[0101] In order to verify the effect of the present invention in the field of auto insurance claims, an experimental case containing 1,000 historical auto insurance risk control rule data was designed. These data cover different types of vehicle accidents, such as collisions, scratches, etc., and contain relevant claims data. The experiment used a data set from the real world, which included historical risk control rules and real-time behavior data. The specific steps are as follows:

[0102] Data preprocessing: The historical risk control rule data was cleaned, missing values ​​were processed, and outliers were processed. The processed data was segmented and stop words were removed.

[0103] Extract semantic features: Use the TF-IDF algorithm to calculate the frequency of each word in the document, and sort the TF-IDF values ​​of each word in the document in descending order, and select the top 10 words with the highest TF-IDF values ​​as target words. Use the continuous bag of words model to extract semantic features and form a semantic feature vector.

[0104] Cluster analysis: Cluster analysis was performed on the formed semantic feature vectors, and five cluster center clusters were obtained. The accident type with the highest frequency in each cluster was taken as the representative accident type of the cluster, and the cluster center cluster was named.

[0105] Real-time behavior data processing: The real-time behavior data is standardized, and the real-time behavior features are extracted from it to construct the real-time behavior feature vector.

[0106] Constructing a risk control rule effect prediction model: Using the extracted semantic feature vectors and real-time behavior feature vectors, combined with the prediction of premium and claim risks, a risk control rule effect prediction model was constructed and the prediction results were output.

[0107] Calculate the average prediction effect score: The prediction effect score is combined with the historical clustered accident type to calculate the average prediction effect score for each historical clustered accident type.

[0108] Risk level classification: Based on the calculated average prediction effect score, the thresholds for classification into four risk levels: low, medium, high and very high are determined.

[0109] Dynamically update the risk customer list: Based on the risk levels, the risk customer list is dynamically updated.

[0110] The details are shown in Table 1:

[0111] Table 1 Experimental record table

[0112]

[0113] It can be observed from Table 1 above that in the embodiment, different types of car insurance accidents have different average prediction effect scores, which reflects their risk level differences. For example, the average prediction effect score of spontaneous combustion is 0.91, which is rated as an extremely high risk level, which means that for this type of accident, the implementation of new risk control rules may have a high risk. On the other hand, the average prediction effect scores of scratching and hitting the guardrail are 0.63 and 0.59, respectively, and are both rated as low risk levels, indicating that it is relatively safe to implement new risk control rules on these two types of accidents.

[0114] In addition, by comparing the average prediction effect scores of different accident types, it can be found that accident types with higher scores (such as spontaneous combustion and rollover) usually correspond to higher risk levels. This shows that the method of the present invention can effectively identify and distinguish the risk levels of different accident types, thereby helping insurance companies to better manage risks and avoid potential losses. In particular, for accident types rated as extremely high risk levels, insurance companies can take timely measures through the method of the present invention, such as adding relevant customers to risky customers, or strengthening monitoring mechanisms, to reduce potential risks.

[0115] In summary, the present invention can accurately predict the effect of newly created risk control rules by constructing a risk control rule effect prediction model, and effectively manage risks by grading the average prediction effect scores of different accident types, which reflects its creativity and novelty. This method not only helps to improve the risk management capabilities of insurance companies, but also reduces the economic losses caused by improper risk control.

[0116] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A risk control rule effect prediction method based on big data, characterized by: include, Collect historical risk control rule data and real-time behavior data, extract semantic feature vectors from historical risk control rule data, and perform cluster analysis to form historical clustered insurance types; Based on semantic feature vectors and real-time behavioral data, we consider the prediction of premium and claim risks and build a risk control rule effect prediction model; Utilize the prediction results output by the risk control rule effect prediction model to dynamically adjust the classification thresholds, classify risk levels, and dynamically update the risk customer list.

2. The risk control rule effect prediction method based on big data according to claim 1, characterized in that: Extracting semantic features from historical risk control rule data includes the following steps: Perform data cleaning, missing value processing and outlier processing on historical risk control rule data, and perform word segmentation on the processed historical risk control rule data to remove stop words; Use TF-IDF to calculate the frequency of each word in the document; Arrange the TF-IDF values ​​of each word in the document in descending order, select the word with the highest TF-IDF value as the target word, and use the continuous bag of words model to extract semantic features, expressed as: Among them, v t represents the semantic features of the target word t, W t represents the word set in the context window of the target word t, v w represents the vector representation of the context word w, |W t | represents the number of words in the context window; Integrate the semantic features in each document to form a semantic feature vector.

3. The risk control rule effect prediction method based on big data according to claim 2, characterized in that: Cluster analysis of semantic feature vectors refers to obtaining cluster center clusters, analyzing each cluster center cluster based on the clustering results, extracting the most frequently occurring insurance type in the cluster as the representative insurance type in the cluster, and naming the cluster center cluster to form historical cluster insurance types.

4. The risk control rule effect prediction method based on big data according to claim 3, characterized in that: Based on semantic feature vectors and real-time behavioral data, considering the prediction of premium and claim risks, a risk control rule effect prediction model is constructed, which includes the following steps: Standardize the real-time behavior data, extract the real-time behavior features, and construct the real-time behavior feature vector; Using semantic feature vectors and real-time behavior feature vectors, considering the prediction of premium and claim risk, we build a risk control rule effect prediction model and predict the effect score, expressed as, Among them, S represents the prediction effect score, n s Represents the number of semantic feature vectors, n b represents the number of real-time behavior feature vectors, v si Represents the semantic feature vector v s The i-th semantic feature in v bj Represents the real-time behavior feature vector v b The jth real-time behavior feature in s represents the semantic feature vector, v b represents the real-time behavior feature vector, λ represents the influence adjustment parameter of the semantic feature, i represents the index of the semantic feature, and j represents the index of the real-time behavior feature.

5. The risk control rule effect prediction method based on big data according to claim 4, characterized in that: Based on the prediction results output by the risk control rule effect prediction model, the average prediction effect score is calculated, including the following steps: The average prediction effect score is calculated by combining the prediction effect score with the historical clustering insurance types, which is expressed as: in, represents the average prediction effect score of the k-th historical cluster insurance type, N k represents the number of samples of the kth historical clustering insurance type, S q represents the prediction effect score of the qth sample, q represents the index of the sample, and k represents the index of the historical clustering insurance type.

6. The risk control rule effect prediction method based on big data according to claim 5, characterized in that: The average prediction effect score of historical clustered insurance types is used to calculate the partition threshold, which is expressed as: Among them, T1 represents the low risk level threshold, T2 represents the medium risk level threshold, and T3 represents the high risk level threshold. and They respectively represent the average prediction effect score of the first historical cluster insurance type and the average prediction effect score of the second historical cluster insurance type.

7. The risk control rule effect prediction method based on big data according to claim 6, characterized in that: Based on the calculated classification threshold, the risk level is divided and the risk customer list is dynamically updated, including the following steps: When S≤T 1 When , it indicates that the predicted risk control rule effect is at a low risk level; When T 1 <S≤T 2 When , it indicates that the predicted risk control rule effect is at a medium risk level; When T 2 <S≤T 3 When , it indicates that the predicted risk control rule effect is at a high risk level; When S>T 3 When , it indicates that the predicted risk control rule effect is at an extremely high risk level; For customers with high risk levels, they are added to the risk customer list. The risk customer list is continuously updated according to changes in the classification thresholds. For customers with extremely high risk levels, the risk control team is notified to take action.

8. A risk control rule effect prediction system based on big data, based on the risk control rule effect prediction method based on big data according to any one of claims 1 to 7, characterized in that: include, The data preparation module is responsible for collecting historical risk control rule data and real-time behavior data, extracting semantic feature vectors from historical risk control rule data, and performing cluster analysis to form historical clustered insurance types; The risk control rule effect prediction model building module is responsible for building a risk control rule effect prediction model based on semantic feature vectors and real-time behavior data, taking into account the prediction of premiums and claims risks; The risk management module is responsible for using the prediction results output by the risk control rule effect prediction model to dynamically adjust the classification thresholds, divide the risk levels, and dynamically update the risk customer list.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the risk control rule effect prediction method based on big data described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the risk control rule effect prediction method based on big data described in any one of claims 1 to 7 are implemented.

Images