Service method and device for customer group of financial institution and storage medium

Abstract

The invention discloses a service method and device for a customer group of a financial institution and a storage medium. The method relates to the field of artificial intelligence, and comprises the steps: obtaining a to-be-classified customer data set under the condition of obtaining customer authorization; performing dimension reduction processing on the client data set to obtain a dimension-reduced client data set; the customer data set after dimension reduction is classified, a first type of customer data set and a second type of customer data set are obtained, and the abnormal value of the transaction information of the first type of customers corresponding to the first type of customer data set is lower than the abnormal value of the transaction information of the second type of customers corresponding to the second type of customer data set; and performing clustering analysis on the first type of customers corresponding to the first type of customer data set to obtain customer group classification information corresponding to the first type of customers. Through application of the method and the device, the problem that the service experience for the customer group is relatively poor due to low classification accuracy of manual customer classification in related technologies is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of artificial intelligence, in particular to a service method and device for a customer group of a financial institution and a storage medium. BACKGROUND

[0002] In the field of financial services, traditional customer classification methods rely heavily on human experience and combine credit ratings, transaction records and other indicators for classification. However, this method has many limitations when faced with massive, multi-dimensional data, including rough classification, low accuracy, resulting in insufficient customer service experience and low risk management efficiency. The main problem is that manual classification is difficult to fully capture customer characteristics and is easily influenced by subjective judgment, especially in credit approval and risk assessment, which is insufficient for differentiated services for homogeneous customer groups and inadequate risk control for heterogeneous groups. For example, when processing credit business, if the classification is not detailed and accurate enough, the services provided for customer groups with similar credit conditions may differ too much, or the same service strategy is adopted for customers with different credit risks, which not only reduces customer satisfaction, but also increases the operating costs and potential risks of financial institutions.

[0003] The related art has not yet proposed an effective solution to the problem of low classification accuracy based on manual customer classification, resulting in poor service experience for customer groups. SUMMARY

[0004] The main purpose of the present application is to provide a service method and device for a customer group of a financial institution and a storage medium to solve the problem of low classification accuracy based on manual customer classification in the related art, resulting in poor service experience for customer groups.

[0005] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a service method for a customer group of a financial institution is provided. The method comprises: obtaining a customer data set to be classified under the condition of obtaining customer authorization; performing dimension reduction processing on the customer data set to obtain a dimension-reduced customer data set; classifying the dimension-reduced customer data set to obtain a first customer data set and a second customer data set, wherein the abnormal value of the transaction information of the first customer corresponding to the first customer data set is lower than the abnormal value of the transaction information of the second customer corresponding to the second customer data set; performing cluster analysis on the first customer corresponding to the first customer data set to obtain customer group classification information corresponding to the first customer; determining a service process corresponding to each customer group in the customer group classification information, and providing financial services to the corresponding customer group using the service process.

[0006] Further, the dimension reduction processing is performed on the customer dataset to obtain a dimension-reduced customer dataset, including: performing centering processing on the customer dataset to be classified based on a principal component analysis algorithm to obtain a centered customer dataset; calculating a covariance matrix of the centered customer dataset; calculating eigenvalues of the covariance matrix and eigenvectors corresponding to the eigenvalues; and obtaining the dimension-reduced customer dataset based on the eigenvalues and the eigenvectors corresponding to the eigenvalues.

[0007] Further, the dimension-reduced customer dataset is classified to obtain a first-class customer dataset and a second-class customer dataset, including: determining the number of isolated trees, the maximum value N of sampling customer samples for each isolated tree, and a preset threshold value through an isolated forest algorithm, wherein each customer sample corresponds to a sample point; training each isolated tree; calculating the path length of each sample point in the isolated tree in each trained isolated tree; calculating the outlier value of each sample point in each trained isolated tree through an outlier function based on the path length of each sample point in the isolated tree in each trained isolated tree; and classifying the dimension-reduced customer dataset through the outlier value of each sample point in each trained isolated tree and the preset threshold value to obtain the first-class customer dataset and the second-class customer dataset.

[0008] Further, the training of each isolated tree includes: randomly selecting no more than N customer samples from the customer dataset as the root node of the isolated tree; randomly assigning a data feature dimension and a random cutting point of the data feature dimension; configuring the sample points less than the random cutting point in the current node's left branch and the sample points greater than the random cutting point in the current node's right branch under the current data feature dimension; repeating the steps of randomly assigning a data feature dimension and a random cutting point of the data feature dimension, configuring the sample points less than the random cutting point in the current node's left branch and the sample points greater than the random cutting point in the current node's right branch under the current data feature dimension, and constructing a new leaf node until there is only one sample point on the leaf node of the isolated tree or the current isolated tree has reached a preset tree depth, and confirming the completion of the training of the current isolated tree.

[0009] Further, cluster analysis is performed on the first type of customers corresponding to the first type of customer dataset to obtain customer group classification information for the first type of customers, including: randomly selecting a customer sample from the first type of customer dataset as the initial cluster center based on the K-means clustering algorithm; selecting K cluster centers based on the shortest distance between each customer sample in the first type of customer dataset and the existing cluster centers; calculating the distance from each customer sample in the first type of customer dataset to the K cluster centers, and assigning each customer sample in the first type of customer dataset to the nearest cluster; using the mean of the customer samples in each cluster as the cluster center of each cluster; repeating the steps of calculating the distance from each customer sample in the first type of customer dataset to the K cluster centers, assigning each customer sample in the first type of customer dataset to the nearest cluster, and using the mean of the customer samples in each cluster as the cluster center of each cluster, until the preset conditions are met; calculating the distance from each customer sample in the second type of customer dataset to each cluster, and assigning each customer sample in the second type of customer dataset to the nearest cluster; and treating the customer samples in each cluster as a category to obtain customer group classification information.

[0010] Furthermore, selecting K cluster centers based on the shortest distance between each customer sample in the first type of customer dataset and the existing cluster centers includes: calculating the shortest distance between each customer sample in the first type of customer dataset and all cluster centers, and calculating the probability of each customer sample in the first type of customer dataset being a cluster center; selecting the customer sample with the highest probability as the next cluster center; repeating the steps of calculating the shortest distance between each customer sample in the first type of customer dataset and all cluster centers, calculating the probability of each customer sample in the first type of customer dataset being a cluster center, and selecting the customer sample with the highest probability as the next cluster center, until K cluster centers are selected.

[0011] Furthermore, the customer dataset to be classified includes textual information and first numerical information. After obtaining the customer dataset to be classified with the customer's authorization, the method further includes: converting the textual information into numerical information to obtain second numerical information; and processing the first numerical information and the second numerical information to obtain the processed customer dataset.

[0012] To achieve the above objectives, according to another aspect of this application, a service apparatus for a financial institution's customer groups is provided. The apparatus includes: an acquisition unit for acquiring a customer dataset to be classified, with customer authorization; a dimensionality reduction unit for performing dimensionality reduction processing on the customer dataset to obtain a dimensionality-reduced customer dataset; a classification unit for classifying the dimensionality-reduced customer dataset to obtain a first type of customer dataset and a second type of customer dataset, wherein the outlier values ​​of transaction information for the first type of customers in the first type of customer dataset are lower than the outlier values ​​of transaction information for the second type of customers in the second type of customer dataset; a clustering unit for performing cluster analysis on the first type of customers in the first type of customer dataset to obtain customer group classification information corresponding to the first type of customers; and a determination unit for determining the service process corresponding to each customer group in the customer group classification information, and providing financial services to the corresponding customer group using the service process.

[0013] Furthermore, the dimensionality reduction unit includes: a centralization module, used to centralize the customer dataset to be classified based on the principal component analysis algorithm to obtain a centralized customer dataset; a first calculation module, used to calculate the covariance matrix of the centralized customer dataset; a second calculation module, used to calculate the eigenvalues ​​of the covariance matrix and the eigenvectors corresponding to the eigenvalues; and a first determination module, used to obtain the dimensionality-reduced customer dataset based on the eigenvalues ​​and the eigenvectors corresponding to the eigenvalues.

[0014] Furthermore, the classification unit includes: a second determination module, used to determine the number of isolated trees, the maximum value N of customer samples sampled in each isolated tree, and a preset threshold using the isolated forest algorithm, wherein each customer sample corresponds to a sample point; a training module, used to train each isolated tree; a third calculation module, used to calculate the path length of each sample point in each trained isolated tree within the isolated tree; and a fourth calculation module, used to calculate the outlier value of each sample point in each trained isolated tree based on the path length of each sample point in the isolated tree using an outlier function, and to classify the dimensionality-reduced customer dataset using the outlier value of each sample point in each trained isolated tree and the preset threshold, thereby obtaining a first-class customer dataset and a second-class customer dataset.

[0015] Furthermore, the training module includes: a first selection submodule, used to randomly select no more than N customer samples from the customer dataset as the root node of the isolated tree; a specification submodule, used to randomly specify a data feature dimension and a random cut point of the data feature dimension; a configuration submodule, used to configure sample points smaller than the random cut point in the current node's left branch and sample points larger than the random cut point in the current node's right branch under the current data feature dimension; and a training submodule, used to repeat the steps of randomly specifying a data feature dimension and a random cut point in the data feature dimension, configuring sample points smaller than the random cut point in the current node's left branch and sample points larger than the random cut point in the current node's right branch, constructing new leaf nodes, until there is only one sample point on the leaf node of the isolated tree or the current isolated tree has reached a preset tree depth, confirming that the training of the current isolated tree is complete.

[0016] Further, the clustering unit includes: a first selection module, used to randomly select a customer sample from the first type of customer dataset as an initial cluster center based on the K-means clustering algorithm; a second selection module, used to select K cluster centers based on the shortest distance between each customer sample in the first type of customer dataset and the existing cluster centers; a first allocation module, used to calculate the distance from each customer sample in the first type of customer dataset to the K cluster centers and allocate each customer sample in the first type of customer dataset to the nearest cluster; a third determination module, used to use the mean of the customer samples in each cluster as the cluster center of each cluster; a second allocation module, used to repeatedly execute the steps of calculating the distance from each customer sample in the first type of customer dataset to the K cluster centers, allocating each customer sample in the first type of customer dataset to the nearest cluster, and using the mean of the customer samples in each cluster as the cluster center of each cluster, until a preset condition is met; a third allocation module, used to calculate the distance from each customer sample in the second type of customer dataset to each cluster and allocate each customer sample in the second type of customer dataset to the nearest cluster; and a classification module, used to treat the customer samples in each cluster as a category to obtain customer group classification information.

[0017] Furthermore, the second selection module includes: a calculation submodule, used to calculate the shortest distance between each customer sample in the first type of customer dataset and all cluster centers, and to calculate the probability of each customer sample in the first type of customer dataset being a cluster center; a second selection submodule, used to select the customer sample with the highest probability as the next cluster center; and a third selection submodule, used to repeatedly execute the steps of calculating the shortest distance between each customer sample in the first type of customer dataset and all cluster centers, calculating the probability of each customer sample in the first type of customer dataset being a cluster center, and selecting the customer sample with the highest probability as the next cluster center, until K cluster centers are selected.

[0018] Furthermore, the device also includes: a conversion unit for converting textual information into numerical information to obtain second numerical information; and a processing unit for processing the first and second numerical information to obtain a processed customer dataset.

[0019] According to another aspect of this application, a computer-readable storage medium is provided, which includes a stored program, wherein, when the program is executed, it controls the device on which the computer-readable storage medium is located to perform a service method for any financial institution's customer base.

[0020] According to another aspect of this application, an electronic device is provided, comprising: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include a service method for performing any type of financial institution's customer base.

[0021] According to another aspect of this application, a computer program product is provided, including computer instructions that, when executed by a processor, implement the steps of a service method for a customer group of a financial institution as described above.

[0022] In this application, a customer dataset to be classified is obtained, the customer dataset is dimensionality reduced, and the dimensionality-reduced customer dataset is classified. Based on the outliers in the customer transaction information in the customer dataset, a first type of customer dataset and a second type of customer dataset are obtained. Cluster analysis is performed on the first type of customers corresponding to the first type of customer dataset to obtain the customer group classification information corresponding to the first type of customers. Based on the more accurate classification results, the service process corresponding to each type of customer group is determined and the corresponding financial services are provided. This solves the technical problem that the classification accuracy of customer classification based on manual classification is low, resulting in a poor service experience for customer groups, and achieves the technical effect of improving the customer experience. Attached Figure Description

[0023] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0024] Figure 1 A hardware block diagram of a computer terminal for implementing a service method for a financial institution's customer base is shown.

[0025] Figure 2 This is a flowchart illustrating a service method for a financial institution's customer base provided in an embodiment of this application;

[0026] Figure 3This is a flowchart illustrating the classification of a financial institution's customer base according to an embodiment of this application;

[0027] Figure 4 This is a schematic diagram of a service device for a financial institution's customer base provided according to an embodiment of this application;

[0028] Figure 5 This is a structural block diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0029] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0030] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0031] It should be noted that all relevant information (including but not limited to customer personal information, customer group information, etc.) and data (including but not limited to data used for classification, data used for analysis, etc.) involved in this application are information and data authorized by the customer or fully authorized by all parties. For example, this system has an interface with relevant users or organizations. Before obtaining relevant information, it needs to send an acquisition request to the aforementioned users or organizations through the interface, and obtain the relevant information after receiving consent from the aforementioned users or organizations.

[0032] Example 1

[0033] According to an embodiment of this application, a method embodiment for serving a customer base of a financial institution is also provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0034] The method embodiment provided in Embodiment 1 of this application can be executed on a mobile terminal, computer terminal, or similar computing device. Figure 1 A hardware block diagram of a computer terminal (or mobile device) for implementing a service method for a financial institution's customer base is shown. Figure 1 As shown, the computer terminal 10 (or mobile device) may include one or more processors 102 (shown as 102a, 102b, ..., 102n in the figure) 102 (processor 102 may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of a BUS bus), a network interface, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0035] It should be noted that the aforementioned one or more processors 102 and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element within the computer terminal 10 (or mobile device). As involved in the embodiments of this application, the data processing circuits serve as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0036] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the service method for the customer groups of financial institutions in this embodiment of the application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby realizing the aforementioned service method for the customer groups of financial institutions. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the computer terminal 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0037] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computer terminal 10. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0038] The display can be, for example, a touchscreen liquid crystal display (LCD) that allows the user to interact with the user interface of the computer terminal 10 (or mobile device).

[0039] Under the aforementioned operating environment, this application provides the following: Figure 2 The service methods for the customer groups of the financial institutions shown. Figure 2 This is a flowchart of a service method for a financial institution's customer base according to Embodiment 1 of this application.

[0040] Step S201: With the customer's authorization, obtain the customer dataset to be classified.

[0041] As an optional implementation, with the customer's authorization, the customer dataset to be classified may include basic information such as the customer's gender and age, as well as information such as deposit, loan, and credit card usage records.

[0042] Step S202: Perform dimensionality reduction on the customer dataset to obtain the dimensionality-reduced customer dataset.

[0043] As an optional implementation, principal component analysis (PCA) can be used to reduce the dimensionality of the collected multidimensional customer dataset. The purpose of dimensionality reduction is to remove redundant information between features, reduce the complexity of data processing, and thus accelerate the efficiency of subsequent processing. Dimensionality reduction of the customer dataset using PCA may include centering the customer dataset, calculating the covariance matrix, solving for the eigenvalues ​​of the covariance matrix and the corresponding eigenvectors, thereby obtaining the dimensionality-reduced customer dataset.

[0044] For example, the original customer dataset contains 100 features, including several highly correlated financial indicators (e.g., deposit balance, total deposits, annual deposit growth rate, etc.). Using principal component analysis, these highly correlated indicators are merged into a few principal components, reducing the dimensionality of the data while preserving its inherent structure.

[0045] Step S203: Classify the dimensionality-reduced customer dataset to obtain a first type of customer dataset and a second type of customer dataset. The outlier values ​​of the transaction information of the first type of customers in the first type of customer dataset are lower than the outlier values ​​of the transaction information of the second type of customers in the second type of customer dataset.

[0046] As an optional implementation, the Isolation Forest algorithm can be used to detect anomalies in the dimensionality-reduced customer data, dividing the detection results into two categories: normal datasets (corresponding to the first type of customer dataset mentioned above) and anomaly datasets (corresponding to the second type of customer dataset mentioned above). Specifically, an anomaly threshold can be set. Customers whose transaction information anomalies are less than the threshold are classified into the first type of customer dataset, while customers whose transaction information anomalies are greater than the threshold are classified into the second type of customer dataset. In other words, the anomalies in the transaction information of the first type of customers in the first type of customer dataset are lower than the anomalies in the transaction information of the second type of customers in the second type of customer dataset. This effectively eliminates the negative impact of outliers on clustering analysis in subsequent customer group classification, ensuring the accuracy and representativeness of the clustering results.

[0047] For example, if a small number of customers in the customer dataset frequently make large or rapid fund transfers, which is significantly different from the trading habits of other customers, the information of these customers will be identified as outliers by the Isolation Forest algorithm and classified into the second type of customer dataset.

[0048] Step S204: Perform cluster analysis on the first type of customers corresponding to the first type of customer dataset to obtain the customer group classification information corresponding to the first type of customers.

[0049] As an optional implementation, K-means clustering can be used to perform clustering analysis on the first type of customer dataset to obtain classification information for each customer group. After removing outliers, K-means clustering is used to cluster the first type of customers in the first type of customer dataset, achieving a refined classification of customer groups and providing accurate basis for subsequent personalized services and risk management.

[0050] Step S205: Determine the service process corresponding to each customer group in the customer group classification information, and use the service process to provide financial services to the corresponding customer group.

[0051] As an optional implementation, based on the clustering results, specific service processes and risk control strategies can be developed for each customer group. Based on the attributes and needs of each customer group, targeted service solutions can be designed to adapt to the risk management needs of different types of customers, thereby improving the customer experience.

[0052] For example, "frequent transaction users" can enjoy a faster transaction confirmation process; while "high-risk loan users" need to undergo additional identity verification or periodic credit checks.

[0053] The service method for financial institutions' customer groups provided in this application embodiment involves obtaining a customer dataset to be classified with customer authorization; performing dimensionality reduction on the customer dataset to obtain a dimensionality-reduced customer dataset; classifying the dimensionality-reduced customer dataset to obtain a first type of customer dataset and a second type of customer dataset, wherein the outlier values ​​of the transaction information of the first type of customers in the first type of customer dataset are lower than the outlier values ​​of the transaction information of the second type of customers in the second type of customer dataset; performing cluster analysis on the first type of customers in the first type of customer dataset to obtain customer group classification information corresponding to the first type of customers; determining the service process corresponding to each type of customer group in the customer group classification information, and using the service process to provide financial services to the corresponding customer group. This solves the technical problem of low classification accuracy based on manual customer classification, which leads to a poor service experience for customer groups, and achieves the technical effect of improving the customer group's experience.

[0054] To reduce the complexity of data processing and improve the efficiency of subsequent customer group classification, optionally, in the customer group service method for financial institutions provided in this application embodiment, the customer dataset is subjected to dimensionality reduction processing to obtain the dimensionality-reduced customer dataset, which includes:

[0055] The first step is to centralize the customer dataset to be classified based on the principal component analysis algorithm, resulting in a centralized customer dataset.

[0056] As an optional implementation, centering is an important preprocessing step in principal component analysis (PCA) algorithms. By subtracting the mean of the dataset from the data, the data bias is eliminated, and the data is distributed in a space centered on the origin, allowing PCA algorithms to correctly identify the direction of the data variance.

[0057] For example, a customer dataset consists of m n-dimensional data points: X n×m =[x1,x2,...,x m The customer dataset obtained after centralized data processing is as follows:

[0058]

[0059] The second step is to calculate the covariance matrix of the centralized customer dataset.

[0060] As an optional implementation, the covariance matrix reflects the strength of relationships between different features in a customer dataset. The covariance matrix is ​​a symmetric matrix where each element represents the covariance between any two features. A larger covariance value indicates a stronger relationship between the two features. Principal component analysis (PCA) algorithms use the covariance matrix to determine the direction of dimensionality reduction. The formula for calculating the covariance matrix is:

[0061]

[0062] The third step is to calculate the eigenvalues ​​of the covariance matrix and the corresponding eigenvectors.

[0063] As an optional implementation, eigenvalues ​​and corresponding eigenvectors can be extracted from the covariance matrix. The eigenvalues ​​represent the variance of the data along the corresponding eigenvector direction. The eigenvalues ​​and eigenvectors are obtained by eigenvalue decomposition of the covariance matrix. The eigenvectors indicate the optimal direction of data projection, while the eigenvalues ​​quantify the degree of data dispersion along these directions. For example, the eigenvalues ​​of the covariance matrix obtained in the second step can be calculated as w1≥w2≥...≥w n The eigenvectors corresponding to the eigenvalues ​​are λ1, λ2, ..., λ n .

[0064] The fourth step is to obtain the dimensionality-reduced customer dataset based on the eigenvalues ​​and their corresponding eigenvectors.

[0065] As an optional implementation, the eigenvectors λ1,...,λ corresponding to the first p largest eigenvalues ​​can be selected. p The dimensionality-reduced customer dataset is as follows:

[0066]

[0067] In summary, the above steps use principal component analysis to reduce the dimensionality of the customer dataset, resulting in a dimensionality-reduced customer dataset. This enables effective management and analysis of the vast customer data of financial institutions, reduces the complexity of data processing, and thus improves the efficiency of customer group classification.

[0068] To more accurately classify the dimensionality-reduced customer dataset, optionally, in the service method for financial institutions' customer groups provided in this application embodiment, classifying the dimensionality-reduced customer dataset to obtain a first type of customer dataset and a second type of customer dataset includes:

[0069] The first step is to determine the number of isolated trees, the maximum number of customer samples N that each isolated tree can sample, and a preset threshold using the isolated forest algorithm, where each customer sample corresponds to a sample point.

[0070] As an alternative implementation, the Isolation Forest algorithm is a tree-based unsupervised learning method for outlier detection. Its core idea is to build decision trees on randomly selected features and feature values ​​to "isolate" outlier data points. Before classification begins, it's necessary to determine the number of isolated trees in the Isolation Forest, the maximum number of samples N sampled per isolated tree during training, and the preset threshold for judging outliers. The number of isolated trees determines the algorithm's complexity and detection accuracy; more isolated trees mean more detailed detection. The maximum number of customer samples N sampled per isolated tree affects the diversity of each tree; a smaller N is better at detecting local anomalies, while a larger N is more conducive to detecting all anomalies. The preset threshold is the boundary for determining whether a data point is an outlier.

[0071] For example, we can determine to build 100 isolated trees, and each isolated tree can be used to extract a maximum of 1,000 sample points for training. The preset threshold is set to 0.7.

[0072] The second step is to train each isolated tree.

[0073] As an alternative implementation, the core of the Isolation Forest algorithm is to quickly locate outliers in a dataset using a tree structure. During the construction of each isolation tree, the algorithm randomly selects a feature and a split point to divide the dataset into two parts. This process is then recursively repeated on each subset until each subset contains only one sample or the maximum depth is reached. Essentially, this process involves finding paths within the dataset that isolate outliers.

[0074] The third step is to calculate the path length of each sample point in each trained isolated tree.

[0075] As an optional implementation, after each isolation tree is trained, the path length from the root node to each sample point in the isolation tree is calculated. Path length is a key indicator for determining whether a sample point is an outlier. In an isolation tree, normal sample points typically require a longer path to be isolated, while outliers, due to their unique behavior patterns, are often isolated on shorter paths. The shorter the path length, the easier it is for the sample point to be isolated, and the more likely it is to be an outlier.

[0076] For example, a customer who frequently trades late at night may quickly reach a leaf node in most isolated trees, indicating that their behavior pattern is different from that of most customers, and thus they are marked as an outlier.

[0077] The fourth step involves calculating the outlier value of each sample point in each trained isolated tree based on the path length of each sample point in the isolated tree using an outlier function. Then, using the outlier value of each sample point in each trained isolated tree and a preset threshold, the dimensionality-reduced customer dataset is classified to obtain the first type of customer dataset and the second type of customer dataset.

[0078] As an optional implementation, the outlier score of each sample point in all trees is calculated using an outlier function. Based on the outlier of each sample point and a preset threshold, the dimensionality-reduced customer dataset can be divided into two categories: the first category of customer dataset (scores below the threshold, considered normal) and the second category of customer dataset (scores above the threshold, considered abnormal).

[0079] In summary, by following the steps outlined above, we can effectively identify and classify abnormal customers when processing dimensionality-reduced customer datasets, improve the accuracy of customer group classification, and thus provide more precise personalized services for each customer category, thereby enhancing customer satisfaction.

[0080] To improve the accuracy of customer group classification, optionally, in the customer group service method for financial institutions provided in this application embodiment, training each isolated tree includes:

[0081] The first step is to randomly select no more than N customer samples from the customer dataset as the root node of the isolated tree.

[0082] As an optional implementation, the Isolation Forest algorithm first randomly selects no more than N samples from the customer dataset when constructing each tree, and selects one of these samples as the root node to start tree construction. The root node is the starting point for building the tree, ensuring that the construction of each tree starts from a random perspective, increasing the diversity of the model.

[0083] The second step is to randomly specify a data feature dimension and a random cut point for that data feature dimension.

[0084] As an alternative implementation, after selecting the root node, a data feature dimension is randomly chosen, and a cut point is randomly selected within that dimension to split the dataset into two parts. By randomly selecting features and cut points, the Isolation Forest algorithm can segment the data across different dimensions, thereby discovering the behavior of outliers in different dimensions. This randomness helps to build trees that are not entirely identical, allowing for analysis from different perspectives even when using the same sample set.

[0085] The third step is to configure sample points smaller than the random cut point in the left branch of the current node and sample points larger than the random cut point in the right branch of the current node, based on the current data feature dimensions.

[0086] As an optional implementation, after each segmentation, all sample points can be configured to either the left or right branch based on whether they are less than or greater than the cut point, thus constructing an isolation tree structure. Through this configuration process, the dataset on each branch gradually shrinks, making it easier to identify anomalous customers. As the depth of the isolation tree increases, the probability of outliers being isolated in leaf nodes increases, improving the efficiency and accuracy of outlier detection.

[0087] The fourth step is to repeatedly specify a data feature dimension and a random cut point for the data feature dimension, and then configure sample points smaller than the random cut point in the left branch of the current node and sample points larger than the random cut point in the right branch of the current node. This process is repeated until there is only one sample point on a leaf node of the isolated tree or the current isolated tree has reached the preset tree depth, confirming that the training of the current isolated tree is complete.

[0088] As an optional implementation, the fourth step recursively repeats the second and third steps to construct the structure of an isolated tree until each leaf node contains only one sample point or the tree reaches a preset maximum depth. This process, by constructing trees of varying depths through a repeated segmentation strategy, can increase the ability to detect outliers. The preset tree depth limits the maximum complexity of the tree, preventing overfitting.

[0089] In summary, the above steps can effectively detect outliers in the customer dataset, reduce noise interference in customer classification, and improve the accuracy and efficiency of classification.

[0090] To improve the accuracy of customer group classification, optionally, in the customer group service method for financial institutions provided in this application embodiment, cluster analysis is performed on the first type of customers corresponding to the first type of customer dataset to obtain customer group classification information corresponding to the first type of customers, including:

[0091] The first step is to randomly select a customer sample from the first type of customer dataset as the initial cluster center based on the K-means clustering algorithm.

[0092] As an alternative implementation, the K-means clustering algorithm is an unsupervised learning method that can be used for data clustering, which divides a set of data into K disjoint subsets (or clusters), where data points within each cluster are more similar to each other, while the differences between clusters are larger.

[0093] The second step is to select K cluster centers based on the shortest distance between each customer sample in the first type of customer dataset and the existing cluster centers.

[0094] As an optional implementation, K cluster centers can be selected by calculating the shortest distance from each customer sample in the first type of customer dataset to the existing cluster centers. The K-means clustering algorithm selects new cluster centers based on the sum of squared distances, where the probability of each data point being selected as a new cluster center is proportional to the square of its distance to the nearest cluster center. In other words, the farther a data point is from the current cluster center, the greater its probability of being selected as the next cluster center, thus ensuring that the K cluster centers are evenly distributed in the data space.

[0095] The third step is to calculate the distance from each customer sample in the first type of customer dataset to the K cluster centers, and then assign each customer sample in the first type of customer dataset to the nearest cluster.

[0096] As an optional implementation, the cluster to which each customer sample belongs can be determined by calculating the distance from each customer sample to the K cluster centers. Calculating the distance from each customer sample in the first type of customer dataset to the K cluster centers allows us to calculate the Euclidean distance, which is the length of the line segment between two points.

[0097] The fourth step is to use the mean of the customer samples in each cluster as the cluster center of each cluster.

[0098] As an optional implementation, this step is an iterative step in the K-means clustering algorithm. It adjusts the cluster positions by recalculating the cluster centers to minimize the sum of squared total distances within each cluster. By updating the cluster centers of each cluster, the cluster positions can be dynamically adjusted, gradually converging to the optimal clustering result and improving clustering accuracy.

[0099] The fifth step is to repeatedly calculate the distance from each customer sample in the first type of customer dataset to the K cluster centers, assign each customer sample in the first type of customer dataset to the nearest cluster, and use the mean of the customer samples in each cluster as the cluster center of each cluster, until the preset conditions are met.

[0100] As an optional implementation, this step repeats steps three and four. The preset conditions can be reaching a preset number of iterations or the change in cluster centroids being less than a set threshold. Iteration can stop when the preset conditions are met. Setting preset conditions avoids inaccurate clustering caused by premature stopping of iterations and also controls the computational complexity of the algorithm. Through each round of iteration, the cluster assignments and centroids are updated until the clustering results stabilize and no longer show significant changes.

[0101] Step 6: Calculate the distance from each customer sample in the second type of customer dataset to each cluster, and assign each customer sample in the second type of customer dataset to the nearest cluster.

[0102] As an optional implementation, the isolated forest algorithm in the above steps has already separated outliers. This step can assign these outliers to the most similar normal customer clusters, ensuring that all customer samples are classified. Even unconventional or abnormal behavior patterns can be reasonably categorized, which facilitates unified service and risk assessment in the later stages.

[0103] The seventh step involves treating the customer samples in each cluster as a category to obtain customer group classification information.

[0104] As an optional implementation, after all iterations are complete, the customer samples within each cluster represent a customer group. These groups are classified based on their common behavioral characteristics or attributes, providing a basis for customer segmentation.

[0105] For example, five clusters are obtained, representing "high-frequency traders," "corporate clients," "high-risk borrowers," "long-term stable depositors," and "young clients who frequently use online services." This classification information provides a basis for subsequent personalized services and risk management. Based on the characteristics of each customer group, customized service strategies can be designed and implemented for different customer groups.

[0106] High-frequency traders: Provide these customers with convenient mobile payment solutions, optimize the processing speed of high-frequency trading, and offer benefits such as points rewards to increase customer loyalty;

[0107] Corporate clients: We provide professional corporate services, including inter-company transfers, corporate loans, and financial planning consulting, while optimizing the security and transaction efficiency of corporate accounts to ensure that their financial needs are met;

[0108] High-risk borrowers: For these customers, credit approval processes should be strengthened or additional collateral should be required, and credit assessments should be conducted regularly to reduce potential credit risks;

[0109] Long-term stable depositors: Offer higher-interest savings products, priority services, and recommendations for bank wealth management products to customers with long-term stable deposits, further encouraging deposit behavior;

[0110] For young customers who frequently use online services: Invest in optimizing and innovating digital banking platforms, providing fintech products suitable for young customers, such as digital wallets, while strengthening online security education to enhance customer trust in digital banking.

[0111] In summary, by performing detailed cluster analysis on the first type of customer dataset through the above steps, and by properly processing the second type of customer dataset, we can provide refined management and personalized services for all customers, thereby improving the user experience.

[0112] To more accurately classify customer groups, optionally, in the service method for financial institutions' customer groups provided in this application embodiment, K cluster centers are selected based on the shortest distance between each customer sample in the first type of customer dataset and the existing cluster centers, including:

[0113] The first step is to calculate the shortest distance between each customer sample in the first type of customer dataset and all cluster centers, and then calculate the probability that each customer sample in the first type of customer dataset is a cluster center.

[0114] As an optional implementation, for the existing cluster centers (initially only one randomly selected center), the shortest distance from each customer sample to all selected cluster centers is calculated. Then, based on these distances, the probability of each sample point being the next cluster center is calculated. The probability of a sample point being selected as the next cluster center is the ratio of the square of its distance to the nearest cluster center to the sum of the squares of the distances from all sample points to the nearest cluster center.

[0115] For example, having already selected the first cluster center, we then calculate the shortest distance from this center to every customer sample in the entire dataset of customers in the first category. Customers C and D are 5 and 10 distances from the center, respectively, while the average distance to the center for other customers is 2. Therefore, the probability of customers C and D being the next cluster center will be based on the square of their distances to the center, with customer D having a higher probability because it is farther from the current cluster center.

[0116] The second step is to select the customer sample with the highest probability as the next cluster center.

[0117] As an optional implementation, after calculating the probability of each sample point as the next cluster center in the first step, the customer sample with the highest probability is selected as the new cluster center.

[0118] Continuing with the previous example, if customer D has the highest probability, then customer D will be selected as the second cluster center. The shortest distance from each customer sample to the existing two cluster centers will be calculated to determine the next center point.

[0119] The third step is to repeatedly calculate the shortest distance between each customer sample in the first type of customer dataset and all cluster centers, calculate the probability of each customer sample in the first type of customer dataset being a cluster center, and select the customer sample with the highest probability as the next cluster center, until K cluster centers are selected.

[0120] As an optional implementation, this step will repeat steps one and two until K cluster centers are selected. This is an iterative process; by continuously selecting the points with the highest probabilities as new cluster centers, K cluster centers covering the entire dataset can be gradually constructed, ensuring the optimization of clustering results.

[0121] For example, the goal is to obtain 5 cluster centers. After selecting the first two centers, the algorithm recalculates the shortest distance from each customer sample to these two centers, updates the probability of each sample point being the next cluster center, and then selects the sample point with the highest probability as the third cluster center. This process is repeated until all five centers are selected.

[0122] In summary, obtaining K cluster centers through the above steps improves the efficiency and quality of the clustering algorithm, enabling more accurate identification and analysis of different types of customer groups, thereby providing more suitable services.

[0123] To improve the accuracy of customer group classification, optionally, in the customer group service method for financial institutions provided in this application embodiment, the customer dataset to be classified includes textual information and first numerical information. After obtaining the customer dataset to be classified with customer authorization, the method further includes:

[0124] The first step is to convert the text information into numerical information to obtain the second numerical information.

[0125] As an optional implementation, before transforming textual information in the customer dataset, to avoid missing data, operations such as mode imputation, mean imputation, and deletion of missing values ​​can be performed. The specific method used depends on the data type of the column containing the missing values ​​and the handling of missing data in accordance with business requirements. For textual information, text preprocessing can be performed first to transform the raw text into clean and structured data. Operations such as word segmentation and feature extraction can be used to transform textual information into numerical information.

[0126] The second step is to process the first and second numerical information to obtain the processed customer dataset.

[0127] As an optional embodiment, the first and second numerical information can be processed using a standard score method to obtain the processed customer dataset. For the first and second numerical information x1, x2, ..., x... n The original data is standardized using the standard score method.

[0128]

[0129] Where, x i For the original data (i.e., the first numerical information and the second numerical information), y i The processed customer dataset is denoted as s, where s is the standard deviation of the original data.

[0130] In summary, by processing the customer dataset through the above steps, a more standardized customer dataset can be obtained, thereby improving the accuracy of subsequent classification based on the customer dataset.

[0131] For example, a financial institution's customer base can be categorized through the following steps: Figure 3 The flowchart illustrates the process of classifying a financial institution's customer base. First, a customer dataset of the financial institution is acquired. Next, the acquired customer dataset is preprocessed to obtain a preprocessed customer dataset. Then, principal component analysis (PCA) is used to reduce the dimensionality of the preprocessed customer dataset. Specifically, after centering the customer dataset, the covariance matrix is ​​calculated, and the eigenvalues ​​and corresponding eigenvectors of the covariance matrix are solved to obtain the dimensionality-reduced data. Next, an isolation algorithm is used to determine whether the dimensionality-reduced data represents normal values. Finally, if the dimensionality-reduced data is determined to be normal, cluster analysis is performed on the normal dataset to complete the clustering of samples, thus completing the classification of the financial institution's customer dataset.

[0132] In summary, classifying the customer groups of financial institutions through the above examples can improve the accuracy of customer group classification, thereby enabling the provision of corresponding services to each customer group and ultimately improving customer satisfaction.

[0133] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0134] Example 2

[0135] This application also provides a service device for a financial institution's customer base. It should be noted that this service device can be used to execute the service method for a financial institution's customer base provided in this application. The following describes the service device for a financial institution's customer base provided in this application.

[0136] According to embodiments of this application, an apparatus for implementing the aforementioned service method for the customer groups of financial institutions is also provided, such as... Figure 4 As shown, the device includes: an acquisition unit 401, a dimensionality reduction unit 402, a classification unit 403, a clustering unit 404, and a determination unit 405.

[0137] Specifically, the acquisition unit 401 is used to acquire the customer dataset to be classified with the customer's authorization.

[0138] Dimensionality reduction unit 402 is used to reduce the dimensionality of the customer dataset to obtain the dimensionality-reduced customer dataset;

[0139] Classification unit 403 is used to classify the dimensionality-reduced customer dataset to obtain a first type of customer dataset and a second type of customer dataset, wherein the outlier value of the transaction information of the first type of customer in the first type of customer dataset is lower than the outlier value of the transaction information of the second type of customer in the second type of customer dataset.

[0140] Clustering unit 404 is used to perform cluster analysis on the first type of customers corresponding to the first type of customer dataset to obtain the customer group classification information corresponding to the first type of customers;

[0141] The determining unit 405 is used to determine the service process corresponding to each customer group in the customer group classification information, and to provide financial services to the corresponding customer group using the service process.

[0142] The customer service device for financial institutions provided in this application embodiment acquires a customer dataset to be classified by an acquisition unit 401 with customer authorization; a dimensionality reduction unit 402 performs dimensionality reduction processing on the customer dataset to obtain a dimensionality-reduced customer dataset; a classification unit 403 classifies the dimensionality-reduced customer dataset to obtain a first type of customer dataset and a second type of customer dataset, wherein the outlier values ​​of the transaction information of the first type of customers in the first type of customer dataset are lower than the outlier values ​​of the transaction information of the second type of customers in the second type of customer dataset; a clustering unit 404 performs cluster analysis on the first type of customers in the first type of customer dataset to obtain customer group classification information corresponding to the first type of customers; and a determination unit 405 determines the service process corresponding to each type of customer group in the customer group classification information and provides financial services to the corresponding customer group using the service process. This solves the technical problem in related technologies where the classification accuracy of customer classification based on manual methods is low, resulting in a poor service experience for customer groups, thereby improving the customer experience.

[0143] Optionally, in the service device for the customer groups of financial institutions provided in this application embodiment, the dimensionality reduction unit 402 includes: a centralization module, used to centralize the customer dataset to be classified based on the principal component analysis algorithm to obtain a centralized customer dataset; a first calculation module, used to calculate the covariance matrix of the centralized customer dataset; a second calculation module, used to calculate the eigenvalues ​​of the covariance matrix and the eigenvectors corresponding to the eigenvalues; and a first determination module, used to obtain the dimensionality-reduced customer dataset based on the eigenvalues ​​and the eigenvectors corresponding to the eigenvalues.

[0144] Optionally, in the service device for the customer groups of financial institutions provided in this application embodiment, the classification unit 403 includes: a second determining module, used to determine the number of isolated trees, the maximum value N of customer samples sampled in each isolated tree, and a preset threshold through the isolated forest algorithm, wherein each customer sample corresponds to a sample point; a training module, used to train each isolated tree; a third calculation module, used to calculate the path length of each sample point in each trained isolated tree in the isolated tree; and a fourth calculation module, used to calculate the outlier value of each sample point in each trained isolated tree through an outlier function based on the path length of each sample point in each trained isolated tree, and to classify the dimensionality-reduced customer dataset by using the outlier value of each sample point in each trained isolated tree and the preset threshold to obtain a first type of customer dataset and a second type of customer dataset.

[0145] Optionally, in the service device for the customer group of financial institutions provided in this application embodiment, the training module includes: a first selection submodule, used to randomly select no more than N customer samples from the customer dataset as the root node of the isolated tree; a designation submodule, used to randomly designate a data feature dimension and a random cut point of the data feature dimension; a configuration submodule, used to configure sample points smaller than the random cut point in the left branch of the current node and sample points larger than the random cut point in the right branch of the current node under the current data feature dimension; and a training submodule, used to repeat the steps of randomly designating a data feature dimension and a random cut point of the data feature dimension, configuring sample points smaller than the random cut point in the left branch of the current node and sample points larger than the random cut point in the right branch of the current node, constructing new leaf nodes, until there is only one sample point on the leaf node of the isolated tree or the current isolated tree has reached a preset tree depth, confirming that the training of the current isolated tree is complete.

[0146] Optionally, in the service device for the customer groups of financial institutions provided in this application embodiment, the clustering unit 404 includes: a first selection module, used to randomly select a customer sample from the first type of customer dataset as an initial cluster center based on the K-means clustering algorithm; a second selection module, used to select K cluster centers based on the shortest distance between each customer sample in the first type of customer dataset and the existing cluster centers; a first allocation module, used to calculate the distance from each customer sample in the first type of customer dataset to the K cluster centers, and allocate each customer sample in the first type of customer dataset to the nearest cluster; and a third determination module, used to assign each... The first module uses the mean of customer samples in a cluster as the cluster center of each cluster. The second module repeatedly calculates the distance from each customer sample in the first customer dataset to the K cluster centers, assigns each customer sample in the first customer dataset to the nearest cluster, and uses the mean of customer samples in each cluster as the cluster center of each cluster, until a preset condition is met. The third module calculates the distance from each customer sample in the second customer dataset to each cluster and assigns each customer sample in the second customer dataset to the nearest cluster. The classification module treats each customer sample in a cluster as a category to obtain customer group classification information.

[0147] Optionally, in the service device for the customer groups of financial institutions provided in the embodiments of this application, the second selection module includes: a calculation submodule, used to calculate the shortest distance between each customer sample in the first type of customer dataset and all cluster centers, and to calculate the probability that each customer sample in the first type of customer dataset is a cluster center; a second selection submodule, used to select the customer sample with the highest probability as the next cluster center; and a third selection submodule, used to repeatedly execute the steps of calculating the shortest distance between each customer sample in the first type of customer dataset and all cluster centers, calculating the probability that each customer sample in the first type of customer dataset is a cluster center, and selecting the customer sample with the highest probability as the next cluster center, until K cluster centers are selected.

[0148] Optionally, in the service device for the customer groups of financial institutions provided in the embodiments of this application, the device further includes: a conversion unit for converting textual information into numerical information to obtain second numerical information; and a processing unit for processing the first numerical information and the second numerical information to obtain a processed customer dataset.

[0149] It should be noted that the acquisition unit 401, dimensionality reduction unit 402, classification unit 403, clustering unit 404, and determination unit 405 mentioned above correspond to steps S201 to S205 in Embodiment 1. The five modules and their corresponding steps implement the same instances and application scenarios, but are not limited to the content disclosed in Embodiment 1. It should be noted that the above modules or units can be hardware or software components stored in memory (e.g., memory 104) and processed by one or more processors (e.g., processors 102a, 102b, ..., 102n). The above modules can also be part of a device and run in the computer terminal 10 provided in Embodiment 1.

[0150] Example 3

[0151] Embodiments of this application may provide a computer terminal, which may be any computer terminal device in a group of computer terminals. Optionally, in this embodiment, the aforementioned computer terminal may also be replaced with a mobile terminal or an electronic device, etc.

[0152] Optionally, in this embodiment, the computer terminal may be located in at least one of a plurality of network devices in a computer network.

[0153] In this embodiment, the computer terminal described above can execute the program code for the following steps in the service method for customer groups of financial institutions: obtaining a customer dataset to be classified upon obtaining customer authorization; performing dimensionality reduction on the customer dataset to obtain a dimensionality-reduced customer dataset; classifying the dimensionality-reduced customer dataset to obtain a first type of customer dataset and a second type of customer dataset, wherein the outlier values ​​of the transaction information of the first type of customer in the first type of customer dataset are lower than the outlier values ​​of the transaction information of the second type of customer in the second type of customer dataset; performing cluster analysis on the first type of customers in the first type of customer dataset to obtain customer group classification information corresponding to the first type of customers; determining the service process corresponding to each customer group in the customer group classification information, and using the service process to provide financial services to the corresponding customer group.

[0154] As an optional embodiment, the computer terminal described above can execute program code for the following steps in the service method for a customer group of a financial institution: dimensionality reduction of the customer dataset to obtain a dimensionality-reduced customer dataset includes: centering the customer dataset to be classified based on principal component analysis algorithm to obtain a centralized customer dataset; calculating the covariance matrix of the centralized customer dataset; calculating the eigenvalues ​​of the covariance matrix and the eigenvectors corresponding to the eigenvalues; and obtaining the dimensionality-reduced customer dataset based on the eigenvalues ​​and the eigenvectors corresponding to the eigenvalues.

[0155] As an optional embodiment, the aforementioned computer terminal can execute program code for the following steps in the service method for a financial institution's customer group: classifying the dimensionality-reduced customer dataset to obtain a first type of customer dataset and a second type of customer dataset includes: determining the number of isolated trees, the maximum value N of customer samples sampled in each isolated tree, and a preset threshold using the isolated forest algorithm, wherein each customer sample corresponds to a sample point; training each isolated tree; calculating the path length of each sample point in each trained isolated tree; calculating the outlier value of each sample point in each trained isolated tree using an outlier function based on the path length of each sample point in each trained isolated tree; and classifying the dimensionality-reduced customer dataset using the outlier value of each sample point in each trained isolated tree and the preset threshold to obtain a first type of customer dataset and a second type of customer dataset.

[0156] As an optional embodiment, the computer terminal described above can execute the program code for the following steps in the service method for a customer group of a financial institution: training each isolated tree includes: randomly selecting no more than N customer samples from the customer dataset as the root node of the isolated tree; randomly specifying a data feature dimension and a random cut point of the data feature dimension; configuring sample points smaller than the random cut point in the current data feature dimension on the left branch of the current node and sample points larger than the random cut point on the right branch of the current node; repeating the steps of randomly specifying a data feature dimension and a random cut point of the data feature dimension, configuring sample points smaller than the random cut point in the current data feature dimension on the left branch of the current node and sample points larger than the random cut point on the right branch of the current node, constructing new leaf nodes, until there is only one sample point on the leaf node of the isolated tree or the current isolated tree has reached a preset tree depth, confirming that the training of the current isolated tree is complete.

[0157] As an optional embodiment, the aforementioned computer terminal can execute program code for the following steps in the service method for a financial institution's customer group: performing cluster analysis on the first type of customers corresponding to the first type of customer dataset to obtain customer group classification information corresponding to the first type of customers, including: randomly selecting a customer sample from the first type of customer dataset as an initial cluster center based on the K-means clustering algorithm; selecting K cluster centers based on the shortest distance between each customer sample in the first type of customer dataset and the existing cluster centers; calculating the distance from each customer sample in the first type of customer dataset to the K cluster centers, and then... The first customer dataset assigns each customer sample to the nearest cluster; the mean of the customer samples in each cluster is used as the cluster center of each cluster; the steps of calculating the distance from each customer sample in the first customer dataset to the K cluster centers, assigning each customer sample in the first customer dataset to the nearest cluster, and using the mean of the customer samples in each cluster as the cluster center of each cluster are repeated until a preset condition is met; the second customer dataset calculates the distance from each customer sample in the second customer dataset to each cluster, and assigns each customer sample in the second customer dataset to the nearest cluster; the customer samples in each cluster are considered as a category, thus obtaining customer group classification information.

[0158] As an optional embodiment, the computer terminal described above can execute program code for the following steps in the service method for a financial institution's customer group: selecting K cluster centers based on the shortest distance between each customer sample in the first type of customer dataset and the existing cluster centers includes: calculating the shortest distance between each customer sample in the first type of customer dataset and all cluster centers, and calculating the probability that each customer sample in the first type of customer dataset is a cluster center; selecting the customer sample with the highest probability as the next cluster center; repeatedly executing the steps of calculating the shortest distance between each customer sample in the first type of customer dataset and all cluster centers, calculating the probability that each customer sample in the first type of customer dataset is a cluster center, and selecting the customer sample with the highest probability as the next cluster center, until K cluster centers are selected.

[0159] As an optional embodiment, the computer terminal described above can execute program code for the following steps in the service method for a customer group of a financial institution: the customer dataset to be classified includes textual information and first numerical information. After obtaining the customer dataset to be classified with the customer's authorization, the method further includes: converting the textual information into numerical information to obtain second numerical information; and processing the first numerical information and the second numerical information to obtain a processed customer dataset.

[0160] Optionally, Figure 5 This is a structural block diagram of an electronic device according to an embodiment of this application. Figure 5 As shown, the electronic device may include: one or more ( Figure 5 Only one of the components is shown: processor 502, memory 504, memory controller, and peripheral interface, wherein the peripheral interface is connected to the radio frequency module, audio module, and display.

[0161] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the service method and apparatus for the customer groups of financial institutions in this application embodiment. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby realizing the aforementioned service method for the customer groups of financial institutions. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0162] The processor can access the information and application programs stored in the memory via a transmission device to execute the aforementioned steps in the service methods for the customer groups of the aforementioned financial institutions.

[0163] This application provides a service scheme for a financial institution's customer groups. By obtaining customer authorization, a customer dataset to be classified is acquired; the customer dataset is then dimensionality-reduced to obtain a dimensionality-reduced customer dataset; the dimensionality-reduced customer dataset is then classified to obtain a first type of customer dataset and a second type of customer dataset, wherein the outlier values ​​of transaction information for the first type of customers in the first type of customer dataset are lower than the outlier values ​​of transaction information for the second type of customers in the second type of customer dataset; cluster analysis is performed on the first type of customers in the first type of customer dataset to obtain customer group classification information corresponding to the first type of customers; a service process corresponding to each customer group in the customer group classification information is determined, and financial services are provided to the corresponding customer groups using the service process. This solves the technical problem of low classification accuracy due to manual customer classification, leading to a poor service experience for customer groups, and achieves the technical effect of improving the customer experience.

[0164] Those skilled in the art will understand that Figure 5 The structure shown is for illustrative purposes only. Electronic devices can also be smartphones (such as Android phones, iOS phones, etc.), tablets, PDAs, mobile internet devices (MIDs), PADs, and other terminal devices. Figure 5 This does not limit the structure of the aforementioned electronic device. For example, electronic devices may also include components that are more... Figure 5 The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 5 The different configurations shown.

[0165] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0166] Example 4

[0167] Embodiments of this application also provide a storage medium. Optionally, in this embodiment, the storage medium can be used to store the program code executed by the service method for the customer groups of financial institutions provided in Embodiment 1.

[0168] Optionally, in this embodiment, the storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0169] Optionally, in this embodiment, the storage medium is configured to store program code for performing the following steps: obtaining a customer dataset to be classified upon obtaining customer authorization; performing dimensionality reduction on the customer dataset to obtain a dimensionality-reduced customer dataset; classifying the dimensionality-reduced customer dataset to obtain a first-class customer dataset and a second-class customer dataset, wherein the outlier values ​​of the transaction information of the first-class customers corresponding to the first-class customer dataset are lower than the outlier values ​​of the transaction information of the second-class customers corresponding to the second-class customer dataset; performing cluster analysis on the first-class customers corresponding to the first-class customer dataset to obtain customer group classification information corresponding to the first-class customers; determining the service process corresponding to each customer group in the customer group classification information, and using the service process to provide financial services to the corresponding customer group.

[0170] As an optional embodiment, the storage medium is also configured to store program code for performing the following steps: dimensionality reduction of the customer dataset to obtain a dimensionality-reduced customer dataset includes: centering the customer dataset to be classified based on principal component analysis algorithm to obtain a centralized customer dataset; calculating the covariance matrix of the centralized customer dataset; calculating the eigenvalues ​​of the covariance matrix and the eigenvectors corresponding to the eigenvalues; and obtaining the dimensionality-reduced customer dataset based on the eigenvalues ​​and the eigenvectors corresponding to the eigenvalues.

[0171] As an optional embodiment, the storage medium is also configured to store program code for performing the following steps: classifying the dimensionality-reduced customer dataset to obtain a first-class customer dataset and a second-class customer dataset, including: determining the number of isolated trees, the maximum value N of customer samples sampled in each isolated tree, and a preset threshold using the isolated forest algorithm, wherein each customer sample corresponds to a sample point; training each isolated tree; calculating the path length of each sample point in each trained isolated tree within the isolated tree; calculating the outlier value of each sample point in each trained isolated tree using an outlier function based on the path length of each sample point in each trained isolated tree; and classifying the dimensionality-reduced customer dataset using the outlier value of each sample point in each trained isolated tree and the preset threshold to obtain a first-class customer dataset and a second-class customer dataset.

[0172] As an optional embodiment, the storage medium is also configured to store program code for performing the following steps: training each isolated tree includes: randomly selecting no more than N customer samples from the customer dataset as the root node of the isolated tree; randomly specifying a data feature dimension and a random cut point for the data feature dimension; configuring sample points smaller than the random cut point in the current data feature dimension on the left branch of the current node and sample points larger than the random cut point on the right branch of the current node; repeating the steps of randomly specifying a data feature dimension and a random cut point for the data feature dimension, configuring sample points smaller than the random cut point in the current data feature dimension on the left branch of the current node and sample points larger than the random cut point on the right branch of the current node, constructing new leaf nodes, until there is only one sample point on the leaf node of the isolated tree or the current isolated tree has reached a preset tree depth, confirming that the training of the current isolated tree is complete.

[0173] As an optional embodiment, the storage medium is also configured to store program code for performing the following steps: performing cluster analysis on the first type of customers corresponding to the first type of customer dataset to obtain customer group classification information corresponding to the first type of customers, including: randomly selecting a customer sample from the first type of customer dataset as an initial cluster center based on the K-means clustering algorithm; selecting K cluster centers based on the shortest distance between each customer sample in the first type of customer dataset and the existing cluster centers; calculating the distance from each customer sample in the first type of customer dataset to the K cluster centers, and assigning each customer sample in the first type of customer dataset to the nearest cluster; using the mean of the customer samples in each cluster as the cluster center of each cluster; repeatedly executing the steps of calculating the distance from each customer sample in the first type of customer dataset to the K cluster centers, assigning each customer sample in the first type of customer dataset to the nearest cluster, and using the mean of the customer samples in each cluster as the cluster center of each cluster, until a preset condition is met; calculating the distance from each customer sample in the second type of customer dataset to each cluster, and assigning each customer sample in the second type of customer dataset to the nearest cluster; and treating the customer samples in each cluster as a category to obtain customer group classification information.

[0174] As an optional embodiment, the storage medium is also configured to store program code for performing the following steps: selecting K cluster centers based on the shortest distance between each customer sample in the first type of customer dataset and the existing cluster centers, including: calculating the shortest distance between each customer sample in the first type of customer dataset and all cluster centers, and calculating the probability that each customer sample in the first type of customer dataset is a cluster center; selecting the customer sample with the highest probability as the next cluster center; repeating the steps of calculating the shortest distance between each customer sample in the first type of customer dataset and all cluster centers, calculating the probability that each customer sample in the first type of customer dataset is a cluster center, and selecting the customer sample with the highest probability as the next cluster center, until K cluster centers are selected.

[0175] As an optional embodiment, the storage medium is also configured to store program code for performing the following steps: the customer dataset to be classified includes textual information and first numerical information; after obtaining the customer dataset to be classified with customer authorization, the method further includes: converting the textual information into numerical information to obtain second numerical information; and processing the first numerical information and the second numerical information to obtain a processed customer dataset.

[0176] This application also provides a computer program product that, when executed on a data processing device, is suitable for performing service method steps for a financial institution's customer base.

[0177] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0178] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0179] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of units or modules may be electrical or other forms.

[0180] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0181] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0182] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0183] The above are merely preferred embodiments of this application. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A service method for a financial institution's customer base, characterized in that, include: With the customer's authorization, obtain the customer dataset to be classified; The customer dataset is then subjected to dimensionality reduction processing to obtain a dimensionality-reduced customer dataset. The dimensionality-reduced customer dataset is classified to obtain a first type of customer dataset and a second type of customer dataset, wherein the outlier value of the transaction information of the first type of customer in the first type of customer dataset is lower than the outlier value of the transaction information of the second type of customer in the second type of customer dataset. Cluster analysis is performed on the first type of customers corresponding to the first type of customer dataset to obtain the customer group classification information corresponding to the first type of customers. Determine the service process corresponding to each customer group in the customer group classification information, and use the service process to provide financial services to the corresponding customer group.

2. The method according to claim 1, characterized in that, The customer dataset is then subjected to dimensionality reduction processing to obtain the dimensionality-reduced customer dataset, which includes: The customer dataset to be classified is centered based on the principal component analysis algorithm to obtain the centered customer dataset. Calculate the covariance matrix of the centralized customer dataset; Calculate the eigenvalues ​​of the covariance matrix and the corresponding eigenvectors; Based on the feature values ​​and the corresponding feature vectors, the dimensionality-reduced customer dataset is obtained.

3. The method according to claim 1, characterized in that, The dimensionality-reduced customer dataset is classified to obtain a first-class customer dataset and a second-class customer dataset, which include: The number of isolated trees, the maximum value N of customer samples collected by each isolated tree, and the preset threshold are determined by the isolated forest algorithm, where each customer sample corresponds to one sample point. Train each isolated tree; Calculate the path length of each sample point in each trained isolated tree; Based on the path length of each sample point in each trained isolated tree, the outlier value of each sample point in each trained isolated tree is calculated using an outlier function. The dimensionality-reduced customer dataset is then classified using the outlier value of each sample point in each trained isolated tree and the preset threshold to obtain a first type of customer dataset and a second type of customer dataset.

4. The method according to claim 3, characterized in that, Training each isolated tree includes: No more than N customer samples are randomly selected from the customer dataset to serve as the root node of the isolated tree; Randomly specify a data feature dimension and a random cut point for that data feature dimension; Under the current data feature dimension, sample points smaller than the random cut point are configured in the left branch of the current node, and sample points larger than the random cut point are configured in the right branch of the current node. Repeat the steps of randomly specifying a data feature dimension and a random cut point for the data feature dimension, configuring sample points smaller than the random cut point in the left branch of the current node and configuring sample points larger than the random cut point in the right branch of the current node, constructing new leaf nodes, until there is only one sample point on the leaf node of the isolated tree or the current isolated tree has reached the preset tree depth, confirming that the training of the current isolated tree is complete.

5. The method according to claim 3, characterized in that, Cluster analysis was performed on the first type of customers corresponding to the first type of customer dataset to obtain the customer group classification information corresponding to the first type of customers, including: Based on the K-means clustering algorithm, a customer sample is randomly selected from the first type of customer dataset as the initial cluster center; Based on the shortest distance between each customer sample in the first type of customer dataset and the existing cluster centers, K cluster centers are selected. Calculate the distance from each customer sample in the first type of customer dataset to the K cluster centers, and assign each customer sample in the first type of customer dataset to the nearest cluster; The mean of the customer samples in each cluster is used as the cluster center of each cluster; Repeat the steps of calculating the distance from each customer sample in the first type of customer dataset to the K cluster centers, assigning each customer sample in the first type of customer dataset to the nearest cluster, and using the mean of the customer samples in each cluster as the cluster center of each cluster, until the preset condition is met; Calculate the distance from each customer sample in the second type of customer dataset to each cluster, and assign each customer sample in the second type of customer dataset to the nearest cluster; Customer samples in each cluster are treated as a category, thus obtaining customer group classification information.

6. The method according to claim 5, characterized in that, Based on the shortest distance between each customer sample in the first type of customer dataset and the existing cluster centers, K cluster centers are selected, including: Calculate the shortest distance between each customer sample in the first type of customer dataset and all cluster centers, and calculate the probability that each customer sample in the first type of customer dataset is a cluster center; Select the customer sample with the highest probability as the next cluster center; Repeatedly calculate the shortest distance between each customer sample in the first type of customer dataset and all cluster centers, calculate the probability of each customer sample in the first type of customer dataset being a cluster center, and select the customer sample with the highest probability as the next cluster center, until K cluster centers are selected.

7. The method according to claim 1, characterized in that, The customer dataset to be classified includes textual information and first numerical information. After obtaining the customer dataset with customer authorization, the method further includes: The textual information is converted into numerical information to obtain the second numerical information; The first and second numerical information are processed to obtain the processed customer dataset.

8. A service device for a financial institution's customer base, characterized in that, include: The acquisition unit is used to acquire the customer dataset to be classified, with the customer's authorization. The dimensionality reduction unit is used to reduce the dimensionality of the customer dataset to obtain the dimensionality-reduced customer dataset. A classification unit is used to classify the dimensionality-reduced customer dataset to obtain a first type of customer dataset and a second type of customer dataset, wherein the outlier value of the transaction information of the first type of customer in the first type of customer dataset is lower than the outlier value of the transaction information of the second type of customer in the second type of customer dataset. Clustering unit, used to perform cluster analysis on the first type of customers corresponding to the first type of customer dataset, to obtain customer group classification information corresponding to the first type of customers; The determining unit is used to determine the service process corresponding to each customer group in the customer group classification information, and to provide financial services to the corresponding customer group using the service process.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored executable program, wherein, when the executable program is executed, it controls the device on which the computer-readable storage medium is located to perform the service method for a customer group of a financial institution as described in any one of claims 1 to 7.

10. An electronic device, characterized in that, include: Memory, which stores executable programs; A processor for running the program, wherein the program, when running, performs the service method for a customer group of a financial institution as described in any one of claims 1 to 7.

11. A computer program product comprising computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the steps of the service method for the customer group of a financial institution as described in any one of claims 1 to 7.

Images