Client credit evaluation method, device and equipment and storage medium

Abstract

The invention discloses a customer credit evaluation method and device, equipment and a storage medium. The method comprises the following steps: firstly processing a target customer personal characteristic information set to obtain a processing data set, scanning to obtain a candidate 1-item set, and filtering according to a support degree to form a first frequent data set; and hashing the first frequent data to the improved hash table bucket to obtain a candidate 2-item set, and storing the candidate 2-item set in the one-dimensional array. And filtering according to the number of the hash buckets, the support degree and the array record support degree to obtain a second frequent data set. And integrating the two frequent data sets to form a frequent item set. And for each newest frequent data set, if any (k-1) th frequent data set is not empty, performing self-connection to generate a candidate k-item set, filtering to obtain a kth frequent data set, integrating the kth frequent data set into a frequent item set, and performing iteration until the newest frequent data set is empty. And any candidate p-item set generated subsequently is recorded in the one-dimensional array. And finally, generating a target customer association rule based on the frequent item set to evaluate customer credit. The scanning frequency can be reduced, and the efficiency is improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a customer credit evaluation method, device, equipment and storage medium. Background Art

[0002] In the context of big data, traditional risk management methods based on statistics and regulations are no longer sufficient and cannot meet the current banking industry's requirements for risk prevention and control. In order to better serve customers and improve risk management capabilities, the banking industry has put forward higher requirements for customer credit assessment systems. In this context, applying relevant technologies such as the Apriori algorithm to conduct customer credit assessment has become an option.

[0003] The Apriori algorithm is a classic algorithm for mining frequent itemsets and association rules. Its core idea is to find frequent itemsets from a data set through continuous iteration. However, in practical applications, the Apriori algorithm needs to scan the database multiple times to calculate the support count of candidate itemsets. This frequent database scanning operation has a great impact on efficiency, especially when the database contains a large number of transactions. Frequent database scanning will significantly reduce the efficiency of customer credit evaluation. Each scan needs to traverse all transaction records. When the amount of data is particularly large, a complete scan will consume a lot of time, causing the entire mining process to become very time-consuming. Summary of the invention

[0004] Based on the above problems, the present application provides a customer credit evaluation method, device, equipment and storage medium, which can reduce the need for multiple database scans, thereby significantly improving processing efficiency.

[0005] The embodiments of the present application disclose the following technical solutions:

[0006] A customer credit evaluation method, the method comprising:

[0007] Processing the target customer's personal feature information set to obtain a processed data set, and scanning the processed data set to obtain a candidate 1-item set;

[0008] Using a preset support and the support of each item in the candidate 1-item set, filtering the candidate 1-item set to obtain a first frequent data set;

[0009] Hash the first frequent data set into the improved hash table bucket to obtain a candidate 2-item set, and record the data in the candidate 2-item set into a one-dimensional array A[M*N]; M represents the total number of item sets, and N represents the total number of transactions;

[0010] Filter the candidate 2-item set according to the number of hash buckets of the improved hash table bucket, the preset support, and the support of each item in the candidate 2-item set recorded in the one-dimensional array to obtain a second frequent data set, and use the union of the first frequent data set and the second frequent data set as a frequent item set set;

[0011] For each latest frequent data set, if any k-1th frequent data set is not an empty set, a self-connection operation is performed on the k-1th frequent data set to generate a potential candidate k-item set, and the candidate k-item set is filtered according to the preset support and the support of each data in the candidate k-item set recorded in the one-dimensional array to obtain the kth frequent data set, and the kth frequent data set is merged into the frequent item set set to obtain a new frequent item set set, and this step is continued to iterate until the latest frequent data set is an empty set; k is an integer greater than 2;

[0012] Generate association rules for the target customers using the latest frequent item set set, and use the association rules to perform credit evaluation on the target customers;

[0013] Wherein, for any candidate p-item set generated subsequently, the data in the candidate p-item set are recorded in a one-dimensional array A[M*N]; p is an integer greater than 2.

[0014] In a possible implementation, hashing the first frequent data set into an improved hash table bucket to obtain a candidate 2-item set includes:

[0015] Determining the number of hash buckets and the hash function of the improved hash table bucket;

[0016] Based on the number of hash buckets and the hash function of the improved hash table bucket, each data in the frequent data set is hashed and mapped to a corresponding bucket;

[0017] The hashed data in the first frequent data set are traversed, and the candidate 2-item sets are generated according to the combination rule of the frequent 1-item sets.

[0018] In a possible implementation, the hash function of the improved hash table bucket is: W is the number of the first frequent data set, x, y are the serial numbers of the data items in the first frequent data set after they are placed in sequence;

[0019] The improved hash table bucket is:

[0020] In a possible implementation, the processing of the target customer's personal feature information set to obtain a processed data set includes:

[0021] Cleaning, reducing and converting the data in the personal characteristic information set one by one to obtain a complete data set with complete data;

[0022] Discretization processing is performed on the continuous personal characteristic information of the complete data set to obtain the processed data set.

[0023] A customer credit evaluation device, comprising:

[0024] A processing unit, used for processing the personal characteristic information set of the target customer to obtain a processed data set;

[0025] A scanning unit, configured to scan the processed data set to obtain a candidate 1-item set;

[0026] A first filtering unit, configured to filter the candidate 1-item set to obtain a first frequent data set by using a preset support and the support of each item in the candidate 1-item set;

[0027] A hash unit, used for hashing the first frequent data set into an improved hash table bucket to obtain a candidate 2-item set;

[0028] A recording unit, used for recording the data in the candidate 2-item set into a one-dimensional array A[M*N], where M represents the total number of item sets and N represents the total number of transactions;

[0029] A second filtering unit, configured to filter the candidate 2-item set according to the number of hash buckets of the improved hash table bucket, the preset support, and the support of each item in the candidate 2-item set recorded in the one-dimensional array to obtain a second frequent data set;

[0030] a merging unit, configured to take a union of the first frequent data set and the second frequent data set as a frequent item set set;

[0031] an iterative unit, for each latest obtained frequent data set, if any k-1th frequent data set is not an empty set, then a self-join operation is performed on the k-1th frequent data set to generate a potential candidate k-item set, the candidate k-item set is filtered according to the preset support and the support of each item in the candidate k-item set recorded in the one-dimensional array to obtain a kth frequent data set, the kth frequent data set is merged into the frequent item set set to obtain a new frequent item set set, and this step is continued to iterate until the latest obtained frequent data set is an empty set; k is an integer greater than 2;

[0032] An association rule generating unit, used for generating association rules for the target customers by using the latest frequent item set set;

[0033] An evaluation unit, used for performing credit evaluation on the target customer using the association rule;

[0034] Wherein, for any candidate p-item set generated subsequently, the data in the candidate p-item set are recorded in a one-dimensional array A[M*N]; p is an integer greater than 2.

[0035] In a possible implementation, the hash unit specifically includes:

[0036] A determination unit, used to determine the number of hash buckets and the hash function of the improved hash table bucket;

[0037] A mapping unit, configured to perform a hash operation on each data in the frequent data set and map it to a corresponding bucket based on the number of hash buckets and the hash function of the improved hash table bucket;

[0038] The candidate item set generating unit is used to traverse the data in the hashed first frequent data set and generate the candidate 2-item set according to the combination rule of the frequent 1-item set.

[0039] In a possible implementation, the hash function of the improved hash table bucket is: W is the number of the first frequent data set, x, y are the serial numbers of the data items in the first frequent data set after they are placed in sequence;

[0040] The improved hash table bucket is:

[0041] In a possible implementation, the processing unit specifically includes:

[0042] A complete data set acquisition unit, used to clean, reduce and convert the data in the personal characteristic information set one by one to obtain a complete data set with complete data;

[0043] The discretization processing unit is used to discretize the continuous personal characteristic information of the complete data set to obtain the processed data set.

[0044] A customer credit evaluation device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the customer credit evaluation method as described above is implemented.

[0045] A computer-readable storage medium stores instructions, and when the instructions are executed on a terminal device, the terminal device executes the customer credit evaluation method as described above.

[0046] Compared with the prior art, this application has the following beneficial effects:

[0047] The present application provides a method, device, equipment and storage medium for evaluating customer credit. Specifically, when executing the customer credit evaluation method provided by the embodiment of the present application, first, the personal feature information set of the target customer can be processed to obtain a processed data set, and the candidate 1-item set can be obtained after scanning, and the preset support and the support of each data item are used to filter to form a first frequent data set. The first frequent data set is hashed to an improved hash table bucket to obtain a candidate 2-item set and stored in a one-dimensional array. The candidate 2-item set is then filtered according to the number of hash buckets, the preset support and the data support recorded in the one-dimensional array to obtain a second frequent data set. The first and second frequent data sets are merged as a frequent item set set. Subsequently, for each latest frequent data set, if any k-1th frequent data set is not empty, a self-join operation is performed to generate a potential candidate k-item set, and it is filtered according to the preset support and the data support stored in the one-dimensional array to obtain the kth frequent data set, and integrated into the frequent item set set, and this step is iterated until the latest frequent data set is an empty set (k is an integer greater than 2), and any candidate p-item set obtained in this process is stored in a one-dimensional array. Finally, the latest frequent item set is used to generate association rules for the target customer, and these rules are applied to the credit evaluation of the target customer. This application significantly improves the processing efficiency by storing frequent item sets in a one-dimensional array, reducing the need to scan the database multiple times. At the same time, the use of improved hash table buckets to store frequent item sets reduces unnecessary candidate set generation and reduces computational complexity. In addition, by reducing the number of scans and optimizing candidate set generation, the frequent item set mining process can be completed faster, improving the overall processing speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in this embodiment or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0049] Figure 1 A method flow chart of a customer credit evaluation method provided in an embodiment of the present application;

[0050] Figure 2 A schematic diagram of the structure of a customer credit evaluation device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0051] To facilitate understanding of the technical solutions provided by the embodiments of the present application, the background technology involved in the embodiments of the present application will be described below.

[0052] In the context of big data, traditional risk management methods based on statistics and regulations are no longer sufficient and cannot meet the current banking industry's requirements for risk prevention and control. In order to better serve customers and improve risk management capabilities, the banking industry has put forward higher requirements for customer credit assessment systems. In this context, applying relevant technologies such as the Apriori algorithm to conduct customer credit assessment has become an option.

[0053] The Apriori algorithm is a classic algorithm for mining frequent item sets and association rules. Its core idea is to find frequent item sets from a data set through continuous iteration. However, in practical applications, the Apriori algorithm needs to scan the database multiple times to calculate the support count of candidate item sets. This frequent database scanning operation has a great impact on efficiency, especially when the database contains a large number of transactions. Frequent database scanning will significantly reduce the efficiency of customer credit evaluation. Each scan needs to traverse all transaction records. When the amount of data is particularly large, a complete scan will consume a lot of time, causing the entire mining process to become very time-consuming.

[0054] In order to solve this problem, a method, device, equipment and storage medium for evaluating customer credit are provided in an embodiment of the present application. First, the customer's personal feature information set is processed to generate a processed data set, and the candidate 1-item set is obtained by scanning the data set. Then, the candidate 1-item set is filtered using the preset support threshold and the actual support of each item in the candidate 1-item set to extract the first frequent data set. Subsequently, we hash the first frequent data set into the improved hash table bucket to generate a candidate 2-item set, and record the item set in a one-dimensional array. According to the number of hash buckets of the improved hash table bucket, the preset support and the support of each item in the candidate 2-item set recorded in the one-dimensional array, the second frequent data set is further filtered. Then, the first frequent data set and the second frequent data set are merged to form a preliminary frequent item set set. For each subsequent frequent data set, if the k-1th frequent data set is not an empty set, a self-join operation is performed on it to generate a potential candidate k-item set. According to the preset support and the support of each data in the candidate k-item set recorded in the one-dimensional array, the candidate k-item set is filtered to obtain the kth frequent data set and merge it into the frequent item set set. This process continues to iterate until the newly generated frequent data set is an empty set (k is an integer greater than 2), and any candidate p-item set obtained in this process is stored in a one-dimensional array. Finally, the latest frequent item set is used to generate association rules for target customers, and the target customers are evaluated for credit based on these association rules. This application not only improves the efficiency of frequent item set mining, but also reduces unnecessary calculations, thereby more accurately evaluating customer credit.

[0055] 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 only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0056] See also Figure 1 , which is a method flow chart of a customer credit evaluation method provided by an embodiment of the present application, such as Figure 1 As shown, the customer credit evaluation method may include steps S101-S106:

[0057] S101: Processing a target customer's personal feature information set to obtain a processed data set, and scanning the processed data set to obtain a candidate 1-item set.

[0058] Processing the target customer's personal characteristic information set to obtain a processed data set, and then scanning the processed data set means integrating, cleaning, filtering, and other operations on the personal characteristic information provided by the target customer through pre-designed data processing algorithms and rules to form a processed data set. This processing may include steps such as removing duplicate data, filling missing values, and data standardization to ensure the accuracy and consistency of the data. Scanning the processed data set means traversing or checking the sorted data set to identify all possible 1-item sets that appear in the data set, that is, item sets that contain only a single element. Through this series of operations, you can prepare for the subsequent data mining and analysis process, which helps to discover potential association rules and patterns between target customers.

[0059] In a possible implementation, a data set containing some personal information of customers (i.e., a personal characteristic information set) can be collected from a big data platform and a third-party data company. The data set contains about 50,000 pieces of data. The initial data set is stored in an Excel file. The data set contains at least 8 types of customer information, which are: age represents the customer's age, sex represents the customer's gender, region represents the customer's living area, income represents the customer's income, children represents whether the customer has children, car represents whether the user has a car, mortgage represents whether the customer has a mortgage, and pep represents whether the user's credit is good.

[0060] In a possible implementation, the processing of the target customer's personal feature information set to obtain a processed data set includes:

[0061] The data in the personal characteristic information set are cleaned, reduced and converted one by one to obtain a complete data set with complete data, and then the continuous personal characteristic information of the complete data set is discretized to obtain the processed data set.

[0062] The obtained personal feature information set is first cleaned up. For example, there are some missing values ​​in the age field, and some age values ​​are not in the normal age range, such as some age values ​​are less than 18 years old, and there are some obvious wrong values. Some fields have some missing values ​​and isolated values ​​that are not in the common range. After the previous preliminary processing, a relatively complete data set is obtained.

[0063] These complete data sets are not necessarily the data format required for association rule mining. Therefore, after preliminary processing, association rules generally cannot process continuous data. For the age and income fields, since these fields are continuous, association rules cannot be directly mined. Therefore, they must be discretized first. For example, the age field can be divided into four stages, adolescents are classified into the interval (18, 29], adults are classified into the interval [30, 39], middle-aged people are classified into the interval [40, 49], and middle-aged and elderly people are classified into the interval [50, 67]. Similarly, continuous income can also be discretized into three levels: low income, middle income, and high income. In the process of discretization, we use characters to identify each discretized interval segment. For example, after the discretization of income, low income corresponds to the identifier I1, middle income corresponds to the identifier I2, and high income corresponds to the identifier I3. Similarly, after the discretization of the age field age, adolescents correspond to the identifier A1, adults correspond to the identifier A2, middle-aged people correspond to the identifier A3, and middle-aged and elderly correspond to the identifier A4.

[0064] S102: Using a preset support and the support of each item in the candidate 1-item set, the candidate 1-item set is filtered to obtain a first frequent data set.

[0065] In the above process, using the preset support means setting a threshold to filter out data items that do not meet the set criteria. By calculating and comparing the support of each data item in the candidate 1-item set, it is possible to determine which items meet the preset support requirements, thereby identifying frequently occurring item sets. The first frequent data set is the frequent item set that meets the support requirements.

[0066] It should be noted that the preset support is the minimum support. In the Apriori algorithm, the minimum support is a user-defined threshold used to filter out frequent item sets. It refers to the ratio between the number of times an item set (including a single item or multiple items) appears in a data set and the total number of transactions. By setting the minimum support threshold, you can control which item sets are considered frequent. Specifically, when the support of an item set (that is, the frequency of the item set appearing in the data set) is greater than or equal to the minimum support threshold, the item set is called a frequent item set. Conversely, an item set with a support lower than the minimum support threshold is called an infrequent item set. In the Apriori algorithm, by setting an appropriate minimum support threshold, you can filter out item sets that appear frequently in the data set, thereby improving efficiency and reducing the search space in the association rule mining process.

[0067] S103: hashing the first frequent data set into the improved hash table bucket to obtain a candidate 2-item set, and recording the data in the candidate 2-item set into a one-dimensional array A[M*N].

[0068] Hashing the first frequent data set into the improved hash table bucket to obtain the candidate 2-item sets means that by mapping the frequent item sets into an updated or optimized hash table bucket structure, the data can be stored and managed more efficiently, reducing search time and improving algorithm efficiency. In this process, each frequent item set will be assigned to the corresponding hash table bucket to help organize and retrieve data. This hashing operation helps to speed up the mining of new candidate item sets, identify potential patterns and regularities, and provide support for the generation of more candidate item sets in subsequent steps. Through this hashing technology, large amounts of data can be processed more efficiently to discover the correlations and regularities between target customers.

[0069] The data in the candidate 2-itemsets are recorded in a one-dimensional array A[M*N]. This means that each candidate 2-itemset is converted into a one-dimensional array to facilitate subsequent data processing and analysis. By recording the candidate 2-itemsets in an array, the data can be organized and managed more efficiently for subsequent mining operations.

[0070] It should be noted that M represents the total number of item sets: M represents the total number of different items in the candidate item set. Each item can be an item set on its own, and M represents the total number of all these items. For example, if there are three items {apple, banana, orange}, then M is 3.

[0071] N represents the total number of transactions: N represents the number of transactions in the entire candidate-item set. In association rule mining, a transaction usually refers to a shopping record or a transaction, which contains information about a series of items. N represents the total number of these transactions. For example, if there are 100 customer shopping records, then N is 100.

[0072] In a possible implementation, to record the data in the candidate 2-item set into a one-dimensional array A[M*N], the following steps may be performed:

[0073] (1) Determine the array size: First, determine the size of the one-dimensional array A, where M represents the total number of item sets and N represents the total number of transactions. The sizes of these two dimensions are set according to the specific situation of the candidate 2-item set.

[0074] (2) Initialize one-dimensional array A: Create a one-dimensional array A of size M*N and initialize it to all 0s or other specified initial values.

[0075] (3) Record data: Traverse the candidate 2-item set, process the items in each record, and mark the corresponding position of the item in the one-dimensional array as 1 or an appropriate value.

[0076] (4) Mapping items to array indexes: Map the items in the candidate 2-item set to the indexes of the one-dimensional array A. This can be achieved by establishing a mapping relationship from items to indices, ensuring that each item can find its corresponding position in the array correctly.

[0077] (5) Storing transaction information: In a one-dimensional array A, every N positions (i.e., one transaction) form a group for storing information about the corresponding transaction. For example, the first N positions may store information about the first transaction, the second N positions may store information about the second transaction, and so on.

[0078] (6) Record item occurrence: At the corresponding transaction position, depending on whether the item appears in the transaction, the corresponding position is marked as 1 or other appropriate value to indicate the occurrence of the item in the transaction.

[0079] Through the above steps, the data in the candidate 2-itemset can be effectively recorded into the one-dimensional array A[M*N], which is convenient for the subsequent processing and analysis of frequent item sets and transaction information.

[0080] In a possible implementation, hashing the first frequent data set into the improved hash table bucket to obtain a candidate 2-item set can be implemented through steps A1-A3:

[0081] A1: Determine the number of hash buckets and the hash function of the improved hash table bucket.

[0082] First, determine the number of hash buckets and hash functions for the improved hash table buckets. This step is to plan the structure of the hash table, including determining how many buckets are needed and choosing a suitable hash function to ensure that the data is evenly distributed in the buckets.

[0083] In a possible implementation, the hash function of the improved hash table bucket is: W is the number of the first frequent data set, x, y are the serial numbers of the data items in the first frequent data set after they are placed in sequence;

[0084] The improved hash table bucket is:

[0085] A3: Based on the number of hash buckets and the hash function of the improved hash table bucket, a hash operation is performed on each data in the frequent data set, and the data is mapped to a corresponding bucket.

[0086] Based on the number of hash buckets and hash functions of the improved hash table buckets, each data in the frequent data set is hashed and mapped to the corresponding bucket, which means that each data item in the frequent data set is converted into a hash value through a specific algorithm, and these hash values ​​are assigned to the pre-set hash table buckets according to the designed hash function. This process helps to organize and store data more efficiently, so that data items can be quickly and evenly distributed to each bucket.

[0087] In this process, the hash function is responsible for converting each data item into a unique hash value, and the number of hash buckets determines how many different buckets the data item will be distributed into. The purpose of the hash operation is to speed up data access and improve storage efficiency, while reducing the existence of duplicates, thereby providing higher performance and scalability for subsequent data mining and analysis stages. By mapping data items in frequent data sets to corresponding buckets, a better foundation can be provided for the subsequent data processing process to more effectively discover potential patterns and regularities.

[0088] A3: Traverse the hashed data in the first frequent data set, and generate the candidate 2-itemset according to the combination rule of frequent 1-itemset.

[0089] In the process of traversing the data in the first frequent data set that has been hashed and generating candidate 2-item sets according to the combination rules of frequent 1-item sets, it means checking and analyzing the data in the first frequent data set that has been hashed. Based on the specific combination rules of frequent 1-item sets, attempts will be made to combine different frequent items to construct new candidate 2-item sets.

[0090] By traversing the first frequent data set after hashing, the algorithm will select some combinations of frequent item sets and combine them according to the defined rules. These combinations may be connections between two frequent items or other forms of combination to generate new candidate 2-item sets. This step is to explore the potential patterns and association rules in the data set and provide more possibilities for the next stage of data mining.

[0091] The purpose of generating candidate 2-item sets is to further explore the potential correlations and patterns in the data. Through this process, we can find out which items have frequent co-occurrences or correlations, thus providing more useful information for subsequent analysis and association rule generation. This method can deeply explore the internal structure behind the data and help understand the relationships and behavior patterns between target customers.

[0092] S104: Filter the candidate 2-item sets according to the number of hash buckets of the improved hash table buckets, the preset support, and the support of each item in the candidate 2-item set recorded in the one-dimensional array to obtain a second frequent data set, and use the union of the first frequent data set and the second frequent data set as a frequent item set set.

[0093] According to the number of hash buckets of the improved hash table bucket, the preset support and the support of each item in the candidate 2-item set recorded in the one-dimensional array, the candidate 2-item set is filtered to obtain the second frequent data set, which means that by combining the setting of the hash table bucket, the support threshold and the support information of the specific item set data, the qualified frequent 2-item sets are screened out. This step is to find those frequent 2-item sets that meet the preset support requirements from the candidate item set.

[0094] Taking the union of the first frequent data set and the second frequent data set as the frequent itemset set means combining the frequent itemsets obtained by the first mining and the second frequent itemsets obtained by filtering to form a more complete frequent itemset summary. Through this summary method, more frequent itemsets that meet the preset support standards can be collected, which helps to capture the patterns and regularities in the data more comprehensively.

[0095] Integrating the first frequent data set and the second frequent data set into a frequent item set can help identify more extensive and important data patterns and provide more useful information for subsequent data analysis and association rule generation. This process aims to deeply explore the correlation behind the data in order to better understand the relationship and behavior patterns between target customers and provide support for further data mining and decision making.

[0096] S105: For each latest frequent data set, if any k-1th frequent data set is not an empty set, a self-join operation is performed on the k-1th frequent data set to generate a potential candidate k-item set, the candidate k-item set is filtered according to the preset support and the support of each item in the candidate k-item set recorded in the one-dimensional array to obtain the kth frequent data set, the kth frequent data set is merged into the frequent item set set to obtain a new frequent item set set, and this step is continuously iterated until the latest frequent data set is an empty set.

[0097] For each newly obtained frequent data set, if any k-1th frequent data set is not an empty set, these k-1th frequent data sets will be self-joined to generate potential candidate k-item sets. This means using the items in the previous frequent data set and combining them according to specific rules to form a new candidate item set. Next, the candidate k-item sets are filtered according to the preset support and the support of each item in the candidate k-item set recorded in the one-dimensional array to select the kth frequent data set that meets the support requirements. This step aims to find those frequent k-item sets that have been filtered by support from the generated candidate item sets, that is, item sets with high frequency and potential value.

[0098] Merge the kth frequent data set into the frequent itemset set to obtain a new frequent itemset set. Then continue to iterate this step until the latest frequent data set is an empty set. This process will be repeated, and each iteration will further expand the scope of the frequent itemset and discover more strongly associated patterns and rules. Gradually increasing the length k of the itemset will help discover more complex and deep data associations.

[0099] In a possible implementation, the method further includes:

[0100] For any candidate p-item set, the data in the candidate p-item set is recorded into a one-dimensional array A[M*N]; p is an integer greater than 1.

[0101] S106: Generate association rules for the target customers using the latest frequent item set set, and perform credit evaluation on the target customers using the association rules.

[0102] The association rules of target customers are generated by using the latest frequent item sets, aiming to discover the relationship between different items through data mining technology, so as to provide a basis for further credit evaluation. The generated association rules can help understand the correlation between different attributes or behaviors, and evaluate the credit of customers based on these rules.

[0103] In a possible implementation, the generating of the association rule of the target customer by using the latest frequent item set set and the credit evaluation of the target customer by using the association rule can be implemented by the following steps:

[0104] Generate association rules: Based on the set of frequent itemsets, association rule mining algorithms (such as Apriori algorithm) can be applied to find the associations between frequent items. These association rules express the relationship between different attributes. For example, the simultaneous presence of certain attributes may imply specific behaviors or characteristics.

[0105] Analyze association rules: For the generated association rules, their validity and reliability can be evaluated based on indicators such as confidence and support. A high-confidence rule means that when one attribute appears, the other attribute is likely to appear as well. This correlation can be used to infer the potential behavior or characteristics of customers.

[0106] Credit evaluation: By analyzing the generated association rules, the rules can be matched with the customer's data to determine whether the customer meets these rules. Based on the premises and conclusions in the rules, the customer's credit status can be inferred. For example, in a given rule, if the customer lives in the central city and owns a private car, it may indicate that the customer has good credit, because a large amount of data shows that these two attribute combinations are usually associated with higher credit ratings.

[0107] For example, assume the following attribute fields:

[0108] ep represents whether the user's credit is good or not. YES means the customer has good credit, and NO means the customer has bad credit.

[0109] car indicates whether the customer has a private car. YES means the customer has a private car, and NO means the customer does not have a private car.

[0110] region indicates the customer's residential area, including INNER_CITY (city center), TOWN (town), SUBURBAN (suburbs), and RURAL (rural area).

[0111] Through frequent item set mining, a set of frequent item sets is obtained, and specific association rules are generated from them. For example, the confidence threshold is set to 0.8, and the following association rules are generated:

[0112] (region=INNER_CITYAND car=YES==>(pep=YES);

[0113] If there are 9075 customers in the frequent item set who both live in the city center and own a private car, and 8712 of them meet these two conditions and also purchase insurance (pep = YES), then it means: among all 9075 cases that meet region = INNER_CITY AND car = YES, 8712 cases indicate that the customer has good credit (pep = YES). Therefore, the confidence of this rule is 8712 / 9075≈0.96. 0.96 is greater than 0.8, so the customer has good credit.

[0114] Based on the contents of S101-S106, it can be known that the personal feature information set of the target customer is first processed to generate a processed data set, and the processed data set is scanned to obtain candidate 1-item sets. Then, the preset support threshold is used to filter these candidate 1-item sets to extract the first frequent data set. Subsequently, the first frequent data set is hashed into the improved hash table bucket to generate candidate 2-item sets, and the data in these candidate 2-item sets are recorded in a one-dimensional array. Based on the number of hash buckets of the improved hash table bucket, the preset support, and the support of each item in the candidate 2-item set recorded in the one-dimensional array, the candidate 2-item sets are further filtered to obtain the second frequent data set. Afterwards, the first frequent data set and the second frequent data set are merged to form a more comprehensive set of frequent item sets. For each latest frequent data set, if the k-1th frequent data set is not an empty set, a potential candidate k-item set is generated by a self-join operation. According to the preset support and the support of each item in the candidate k-item set recorded in the one-dimensional array, the candidate k-item sets are filtered to extract the kth frequent data set. The kth frequent data set is merged into the frequent item set set, and the process is iterated until the newly generated frequent data set is an empty set (k is an integer greater than 2). Any candidate p-item set obtained in this process is recorded in a one-dimensional array. Finally, the association rules of the target customers are generated using the latest frequent item set set, and these association rules are used to evaluate the credit of the target customers. This application effectively reduces the number of database scans by optimizing the generation and filtering mechanism of candidate sets, and improves the efficiency and accuracy of processing large-scale data.

[0115] See also Figure 2 , Figure 2 This is a schematic diagram of the structure of a customer credit evaluation device provided in an embodiment of the present application. Figure 2 As shown, the customer credit evaluation device includes:

[0116] Processing unit 2001, used for processing the personal characteristic information set of the target customer to obtain a processed data set;

[0117] A scanning unit 2002, configured to scan the processed data set to obtain a candidate 1-item set;

[0118] A first filtering unit 2003, configured to filter the candidate 1-item set by using a preset support and the support of each item in the candidate 1-item set to obtain a first frequent data set;

[0119] A hash unit 2004, configured to hash the first frequent data set into an improved hash table bucket to obtain a candidate 2-item set;

[0120] The recording unit 2005 is used to record the data in the candidate 2-item set into a one-dimensional array A[M*N], where M represents the total number of item sets and N represents the total number of transactions;

[0121] A second filtering unit 2006 is configured to filter the candidate 2-item set according to the number of hash buckets of the improved hash table bucket, the preset support, and the support of each item in the candidate 2-item set recorded in the one-dimensional array to obtain a second frequent data set;

[0122] A merging unit 2007, configured to use a union of the first frequent data set and the second frequent data set as a frequent item set set;

[0123] Iteration unit 2008 is used for, for each latest frequent data set, if any k-1th frequent data set is not an empty set, performing a self-join operation on the k-1th frequent data set to generate a potential candidate k-item set, filtering the candidate k-item set according to the preset support and the support of each item in the candidate k-item set recorded in the one-dimensional array to obtain a kth frequent data set, merging the kth frequent data set into the frequent item set set to obtain a new frequent item set set, and continuing to iterate this step until the latest frequent data set is an empty set; k is an integer greater than 2;

[0124] An association rule generating unit 2009, used to generate association rules of the target customers using the latest frequent item set;

[0125] An evaluation unit 2010, configured to use the association rule to perform credit evaluation on the target customer;

[0126] Wherein, for any candidate p-item set generated subsequently, the data in the candidate p-item set are recorded in a one-dimensional array A[M*N]; p is an integer greater than 2.

[0127] In a possible implementation, the hash unit specifically includes:

[0128] A determination unit, used to determine the number of hash buckets and the hash function of the improved hash table bucket;

[0129] A mapping unit, configured to perform a hash operation on each data in the frequent data set and map it to a corresponding bucket based on the number of hash buckets and the hash function of the improved hash table bucket;

[0130] The candidate item set generating unit is used to traverse the data in the hashed first frequent data set and generate the candidate 2-item set according to the combination rule of the frequent 1-item set.

[0131] In a possible implementation, the hash function of the improved hash table bucket is: W is the number of the first frequent data set, x, y are the serial numbers of the data items in the first frequent data set after they are placed in sequence;

[0132] The improved hash table bucket is:

[0133] In a possible implementation, the processing unit specifically includes:

[0134] A complete data set acquisition unit, used to clean, reduce and convert the data in the personal characteristic information set one by one to obtain a complete data set with complete data;

[0135] The discretization processing unit is used to discretize the continuous personal characteristic information of the complete data set to obtain the processed data set.

[0136] In addition, an embodiment of the present application also provides a customer credit evaluation device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the customer credit evaluation method described above is implemented.

[0137] In addition, an embodiment of the present application further provides a computer-readable storage medium, in which instructions are stored. When the instructions are executed on a terminal device, the terminal device executes the customer credit evaluation method as described above.

[0138] The above is a detailed introduction to a customer credit evaluation method, device, equipment and storage medium provided by the present application. The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same and similar parts between the embodiments can refer to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can refer to the method part description. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of the present application, several improvements and modifications can be made to the present application, and these improvements and modifications also fall within the scope of protection of the claims of the present application.

[0139] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0140] It should also be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

Claims

1. A customer credit evaluation method, characterized in that: The method comprises: Processing the target customer's personal feature information set to obtain a processed data set, and scanning the processed data set to obtain a candidate 1-item set; Using a preset support and the support of each item in the candidate 1-item set, filtering the candidate 1-item set to obtain a first frequent data set; Hash the first frequent data set into the improved hash table bucket to obtain a candidate 2-item set, and record the data in the candidate 2-item set into a one-dimensional array A[M*N]; M represents the total number of item sets, and N represents the total number of transactions; Filter the candidate 2-item set according to the number of hash buckets of the improved hash table bucket, the preset support, and the support of each item in the candidate 2-item set recorded in the one-dimensional array to obtain a second frequent data set, and use the union of the first frequent data set and the second frequent data set as a frequent item set set; For each latest frequent data set, if any k-1th frequent data set is not an empty set, a self-connection operation is performed on the k-1th frequent data set to generate a potential candidate k-item set, and the candidate k-item set is filtered according to the preset support and the support of each data in the candidate k-item set recorded in the one-dimensional array to obtain the kth frequent data set, and the kth frequent data set is merged into the frequent item set set to obtain a new frequent item set set, and this step is continued to iterate until the latest frequent data set is an empty set; k is an integer greater than 2; Generate association rules for the target customers using the latest frequent item set set, and use the association rules to perform credit evaluation on the target customers; Wherein, for any candidate p-item set generated subsequently, the data in the candidate p-item set are recorded in a one-dimensional array A[M*N]; p is an integer greater than 2.

2. The method according to claim 1, characterized in that The step of hashing the first frequent data set into the improved hash table bucket to obtain a candidate 2-item set includes: Determining the number of hash buckets and the hash function of the improved hash table bucket; Based on the number of hash buckets and the hash function of the improved hash table bucket, each data in the frequent data set is hashed and mapped to a corresponding bucket; The hashed data in the first frequent data set are traversed, and the candidate 2-item sets are generated according to the combination rule of the frequent 1-item sets.

3. The method according to claim 1 or 2, characterized in that: The hash function of the improved hash table bucket is: W is the number of the first frequent data set, x, y are the serial numbers of the data items in the first frequent data set after they are placed in sequence; The improved hash table bucket is:

4. The method according to claim 1, characterized in that: The processing of the target customer's personal feature information set to obtain a processed data set includes: Cleaning, reducing and converting the data in the personal characteristic information set one by one to obtain a complete data set with complete data; Discretization processing is performed on the continuous personal characteristic information of the complete data set to obtain the processed data set.

5. A customer credit evaluation device, characterized in that: The device comprises: A processing unit, used for processing the personal characteristic information set of the target customer to obtain a processed data set; A scanning unit, configured to scan the processed data set to obtain a candidate 1-item set; A first filtering unit, configured to filter the candidate 1-item set to obtain a first frequent data set by using a preset support and the support of each item in the candidate 1-item set; A hash unit, used for hashing the first frequent data set into an improved hash table bucket to obtain a candidate 2-item set; A recording unit, used for recording the data in the candidate 2-item set into a one-dimensional array A[M*N], where M represents the total number of item sets and N represents the total number of transactions; A second filtering unit, configured to filter the candidate 2-item set according to the number of hash buckets of the improved hash table bucket, the preset support, and the support of each item in the candidate 2-item set recorded in the one-dimensional array to obtain a second frequent data set; a merging unit, configured to take a union of the first frequent data set and the second frequent data set as a frequent item set set; an iterative unit, for each latest obtained frequent data set, if any k-1th frequent data set is not an empty set, then a self-join operation is performed on the k-1th frequent data set to generate a potential candidate k-item set, the candidate k-item set is filtered according to the preset support and the support of each item in the candidate k-item set recorded in the one-dimensional array to obtain a kth frequent data set, the kth frequent data set is merged into the frequent item set set to obtain a new frequent item set set, and this step is continued to iterate until the latest obtained frequent data set is an empty set; k is an integer greater than 2; An association rule generating unit, used for generating association rules for the target customers by using the latest frequent item set set; An evaluation unit, used for performing credit evaluation on the target customer using the association rule; Wherein, for any candidate p-item set generated subsequently, the data in the candidate p-item set are recorded in a one-dimensional array A[M*N]; p is an integer greater than 2.

6. The device according to claim 5, characterized in that The hash unit specifically includes: A determination unit, used to determine the number of hash buckets and the hash function of the improved hash table bucket; A mapping unit, configured to perform a hash operation on each data in the frequent data set and map it to a corresponding bucket based on the number of hash buckets and the hash function of the improved hash table bucket; The candidate item set generating unit is used to traverse the data in the hashed first frequent data set and generate the candidate 2-item set according to the combination rule of the frequent 1-item set.

7. The device according to claim 5 or 6, characterized in that The hash function of the improved hash table bucket is: W is the number of the first frequent data set, x, y are the serial numbers of the data items in the first frequent data set after they are placed in sequence; The improved hash table bucket is:

8. The device according to claim 5, characterized in that The processing unit specifically includes: A complete data set acquisition unit, used to clean, reduce and convert the data in the personal characteristic information set one by one to obtain a complete data set with complete data; The discretization processing unit is used to discretize the continuous personal characteristic information of the complete data set to obtain the processed data set.

9. A customer credit evaluation device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the customer credit evaluation method according to any one of claims 1 to 4 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed on the terminal device, the terminal device executes the customer credit evaluation method according to any one of claims 1 to 4.

Images