Data processing method and apparatus, terminal device, and storage medium

Abstract

The application discloses a data processing method and device, terminal equipment and storage medium, and relates to the technical field of financial technology, and the method comprises the steps that the data of the customer is acquired; the data of the customer is input into a pre-constructed multi-level product preference probability value model to obtain the preference probability value of the customer for each product, and the multi-level product preference probability value model is obtained based on product classification and classification model training. The application can accurately capture the subtle differences between products.

Description

Technical Field

[0001] This invention relates to the field of financial technology, and in particular to a data processing method, apparatus, terminal equipment, and storage medium. Background Technology

[0002] With the continuous advancement of technology, businesses are paying increasing attention to customer needs. To better understand customer preferences for their various products, businesses use machine learning modeling methods to calculate customer preferences for each product.

[0003] The common practice is to model and score each product separately, using percentile gradations for the scores, and then compare customer preferences for different products based on these percentile gradations. However, this approach also has drawbacks: the granularity of the gradations is too large, making it impossible to fully account for subtle differences between products. Summary of the Invention

[0004] The main objective of this invention is to provide a data processing method, apparatus, terminal device, and storage medium, which aims to solve the problem that current methods for calculating customer preferences for each product cannot fully take into account the subtle differences between products.

[0005] To achieve the above objectives, the present invention provides a data processing method for calculating the probability values ​​of customer preferences for various products, the method comprising:

[0006] Obtain the customer's data;

[0007] The customer's data is input into a pre-built multi-level product preference probability value model to obtain the customer's preference probability value for each product. The multi-level product preference probability value model is trained based on product classification and classification model.

[0008] The step of inputting the customer's data into a pre-built multi-level product preference probability value model to obtain the customer's preference probability value for each product includes:

[0009] The customer's data is input into a pre-built multi-level product preference probability value model to obtain the single-category probability value and combination probability value of each product.

[0010] The individual probability values ​​and combined probability values ​​of each product are integrated to obtain the customer's preference probability value for each product.

[0011] Optionally, the step of integrating the individual probability values ​​and combined probability values ​​of each product to obtain the customer's preference probability value for each product includes:

[0012] The single-class probability value and the combined probability value are merged to obtain the model probability value of each product;

[0013] The model probability values ​​of each product are cascaded and multiplied to obtain the customer's preference probability value for each product.

[0014] Optionally, the step of obtaining the customer's data includes, prior to:

[0015] The multi-level product preference probability value model is constructed based on the product classification and the classification model.

[0016] Optionally, the step of constructing the multi-level product preference probability model based on the product classification and the classification model includes:

[0017] Select sample data at each level;

[0018] The corresponding classification model is selected based on the product classification at each level;

[0019] The corresponding classification model is trained based on the sample data at each level to obtain the multi-level product preference probability value model.

[0020] Optionally, the step of training the corresponding classification model based on the sample data at each level to obtain the multi-level product preference probability value model includes:

[0021] Select extended sample data;

[0022] The multi-level product preference probability value model is iterated based on the extended sample data to obtain the iterated multi-level product preference probability value model.

[0023] Optionally, the step of selecting sample data at each level includes:

[0024] Select primary sample data;

[0025] And / or, select secondary sample data;

[0026] And / or, select three levels of sample data;

[0027] The product classification at each level includes primary product classification, secondary product classification, and tertiary product classification. The step of selecting the corresponding classification model based on the product classification at each level includes:

[0028] A primary classification model is selected based on the aforementioned primary product classification.

[0029] And / or, a secondary classification model is selected based on the secondary product classification information;

[0030] And / or, a three-level classification model is selected based on the three-level product classification.

[0031] The step of training the corresponding classification model based on the sample data at each level to obtain the multi-level product preference probability value model includes:

[0032] The first-level classification model is trained based on the first-level sample data to obtain the first-level product preference probability value model;

[0033] And / or, the secondary classification model is trained based on the secondary sample data to obtain a secondary product preference probability value model;

[0034] And / or, the three-level classification model is trained based on the three-level sample data to obtain a three-level product preference probability value model.

[0035] This invention also proposes a data processing apparatus for calculating the probability values ​​of customer preferences for various products. The apparatus includes:

[0036] The data acquisition module acquires the customer's data;

[0037] The model prediction module inputs the customer's data into a pre-built multi-level product preference probability value model to obtain the customer's preference probability value for each product. The multi-level product preference probability value model is trained based on product classification and classification model.

[0038] This invention also proposes a terminal device, which includes a memory, a processor, and a data processing program stored in the memory and executable on the processor. When the data processing program is executed by the processor, it implements the data processing method described above.

[0039] This invention also proposes a computer-readable storage medium storing a data processing program, which, when executed by a processor, implements the data processing method described above.

[0040] The data processing method, apparatus, terminal device, and storage medium proposed in this invention acquire customer data; input the customer data into a pre-constructed multi-level product preference probability value model to obtain the customer's preference probability value for each product. The multi-level product preference probability value model is trained based on product classification and a classification model. This invention obtains a multi-level product preference probability value model based on product classification and a classification model. Since products at different levels may have different attributes and characteristics, these characteristics may play a key role in predicting customer preferences. By subdividing products into multiple levels and extracting corresponding features for each level, the model can more accurately capture subtle differences between products. Attached Figure Description

[0041] Figure 1 This is a schematic diagram of the functional modules of the terminal device to which the data processing device of the present invention belongs;

[0042] Figure 2 This is a flowchart illustrating an exemplary embodiment of the data processing method of the present invention;

[0043] Figure 3 This is a flowchart illustrating another exemplary embodiment of the data processing method of the present invention;

[0044] Figure 4 This is a schematic diagram of a three-level product system model in an embodiment of the present invention;

[0045] Figure 5 This is a flowchart illustrating another exemplary embodiment of the data processing method of the present invention;

[0046] Figure 6 This is a flowchart illustrating another exemplary embodiment of the data processing method of the present invention;

[0047] Figure 7 This is a flowchart illustrating another exemplary embodiment of the data processing method of the present invention;

[0048] Figure 8 This is a flowchart illustrating another exemplary embodiment of the data processing method of the present invention;

[0049] Figure 9 This is a flowchart illustrating another exemplary embodiment of the data processing method of the present invention;

[0050] Figure 10 This is a flowchart illustrating another exemplary embodiment of the data processing method of the present invention;

[0051] Figure 11 This is a flowchart illustrating another exemplary embodiment of the data processing method of the present invention.

[0052] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0053] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0054] The main solution of this invention is to: acquire customer data; input the customer data into a pre-constructed multi-level product preference probability value model to obtain the customer's preference probability value for each product. The multi-level product preference probability value model is trained based on product classification and a classification model. This invention obtains a multi-level product preference probability value model based on product classification and a classification model. Since products at different levels may have different attributes and characteristics, these characteristics may play a crucial role in predicting customer preferences. By subdividing products into multiple levels and extracting corresponding features for each level, the model can more accurately capture subtle differences between products.

[0055] Technical terms involved in the embodiments of this invention:

[0056] Organic response rate: In marketing or research, when a message or questionnaire is sent to a specific group, the proportion or probability of the group voluntarily responding or participating. It reflects the degree of willingness of respondents or target groups to participate spontaneously without external intervention or facilitation.

[0057] Binary classification models are used to classify samples into two different categories. Examples include determining whether an email is spam or legitimate, or identifying an image as a cat or a dog. Common binary classification models include logistic regression, support vector machines, and random forests. Training these models requires providing sample data with known labels and predicting which category a new unlabeled sample belongs to.

[0058] Multi-class classification models are used to classify samples into three or more distinct categories. For example, classifying an image as a cat, dog, bird, or car. Unlike binary classification models, multi-class models require distinguishing multiple categories during training. Common multi-class classification models include decision trees, K-nearest neighbors, multilayer perceptrons, and convolutional neural networks.

[0059] Stratified sampling: Stratified sampling is primarily used when there are significant differences in sample size between different subgroups. When the differences between subgroups are large and the sample sizes are unbalanced, stratified sampling ensures that a sufficient number of samples are obtained from each subgroup to guarantee the representativeness of the sample in the population. If the sample sizes of the two subgroups are not significantly different and their proportions are similar, non-stratified sampling methods can be used. This simplifies the sampling process, reduces computation, and still yields a sample set that is well-represented in terms of the characteristics of the population.

[0060] This invention takes into account that the commonly used method for calculating customer preference for each product is to model and score each product separately, using percentile gradations for the scores, and finally comparing customer preferences for different products based on these percentile gradations. However, this method also has drawbacks: the granularity of the gradations is too large, making it impossible to fully consider the subtle differences between products.

[0061] Therefore, this invention proposes a solution to acquire customer data; input the customer data into a pre-constructed multi-level product preference probability value model to obtain the customer's preference probability value for each product. The multi-level product preference probability value model is trained based on product classification and a classification model. This invention obtains a multi-level product preference probability value model based on product classification and a classification model. Since products at different levels may have different attributes and characteristics, these characteristics may play a crucial role in predicting customer preferences. By subdividing products into multiple levels and extracting corresponding features for each level, the model can more accurately capture subtle differences between products.

[0062] Specifically, refer to Figure 1 , Figure 1 This is a functional module diagram of the terminal device to which the data processing device of the present invention belongs. The data processing device can be a data processing device independent of the device itself, which can be carried on the device in hardware or software form. The device can be a smart mobile terminal with data processing capabilities, such as a mobile phone or tablet computer, or it can be a fixed device or server with data processing capabilities.

[0063] In this embodiment, the data processing device includes at least an output module 110, a processor 120, a memory 130, and a communication module 140.

[0064] The memory 130 stores the operating system and data processing programs; the output module 110 may be a display screen, etc. The communication module 140 may include a WIFI module and a Bluetooth module, etc., and communicates with external devices or servers through the communication module 140.

[0065] When the data processing program in memory 130 is executed by the processor, it performs the following steps:

[0066] Obtain the customer's data;

[0067] The customer's data is input into a pre-built multi-level product preference probability value model to obtain the customer's preference probability value for each product. The multi-level product preference probability value model is trained based on product classification and classification model.

[0068] Furthermore, when the data processing program in memory 130 is executed by the processor, it also performs the following steps:

[0069] The customer's data is input into a pre-built multi-level product preference probability value model to obtain the single-category probability value and combination probability value of each product.

[0070] The individual probability values ​​and combined probability values ​​of each product are integrated to obtain the customer's preference probability value for each product.

[0071] Furthermore, when the data processing program in memory 130 is executed by the processor, it also performs the following steps:

[0072] The single-class probability value and the combined probability value are merged to obtain the model probability value of each product;

[0073] The model probability values ​​of each product are cascaded and multiplied to obtain the customer's preference probability value for each product.

[0074] Furthermore, when the data processing program in memory 130 is executed by the processor, it also performs the following steps:

[0075] The multi-level product preference probability value model is constructed based on the product classification and the classification model.

[0076] Furthermore, when the data processing program in memory 130 is executed by the processor, it also performs the following steps:

[0077] Select sample data at each level;

[0078] The corresponding classification model is selected based on the product classification at each level;

[0079] The corresponding classification model is trained based on the sample data at each level to obtain the multi-level product preference probability value model.

[0080] Furthermore, when the data processing program in memory 130 is executed by the processor, it also performs the following steps:

[0081] Select extended sample data;

[0082] The multi-level product preference probability value model is iterated based on the extended sample data to obtain the iterated multi-level product preference probability value model.

[0083] Furthermore, when the data processing program in memory 130 is executed by the processor, it also performs the following steps:

[0084] Select primary sample data;

[0085] And / or, select secondary sample data;

[0086] And / or, select three levels of sample data;

[0087] A primary classification model is selected based on the aforementioned primary product classification.

[0088] And / or, a secondary classification model is selected based on the secondary product classification information;

[0089] And / or, a three-level classification model is selected based on the three-level product classification.

[0090] The first-level classification model is trained based on the first-level sample data to obtain the first-level product preference probability value model;

[0091] And / or, the secondary classification model is trained based on the secondary sample data to obtain a secondary product preference probability value model;

[0092] And / or, the three-level classification model is trained based on the three-level sample data to obtain a three-level product preference probability value model.

[0093] This embodiment, through the above-described scheme, specifically obtains the customer's data; inputs the customer's data into a pre-constructed multi-level product preference probability value model to obtain the customer's preference probability value for each product. The multi-level product preference probability value model is trained based on product classification and a classification model. This embodiment of the invention obtains a multi-level product preference probability value model based on product classification and a classification model. Since products at different levels may have different attributes and characteristics, these characteristics may play a crucial role in predicting customer preferences. By subdividing products into multiple levels and extracting corresponding features for each level, the model can more accurately capture subtle differences between products.

[0094] Based on, but not limited to, the above-described device architecture, embodiments of the method of the present invention are proposed.

[0095] The execution subject of the method in this embodiment can be a data processing device. This data processing device can be a device that is independent of the device and capable of data processing. It can be carried on the device in the form of hardware or software.

[0096] Reference Figure 2 , Figure 2 This is a flowchart illustrating an exemplary embodiment of the data processing method of the present invention. The data processing method is used to calculate the probability values ​​of customer preferences for each product, and the data processing method includes:

[0097] Step S20: Obtain the customer's data.

[0098] The data processing method of this invention mainly uses model probabilities to solve the customer preference problem.

[0099] In banks, it is necessary to market products based on customer needs and preferences. The issues involved include (1) which customers have a preference for wealth management products; (2) which customers have a preference for major product types such as entrusted wealth management, funds, deposits, insurance, and precious metals; and (3) which customers have a preference for specific types of fund products such as equity funds, fixed income funds, bond funds, and money market funds.

[0100] Then, to solve the preference problem, machine learning methods are needed to predict the probability values ​​of customers' preferences for each product category. For example, the probability of a customer purchasing wealth management products can be used to solve the problem in (1) above; when a customer purchases wealth management products, the probability of purchasing a certain category of products can be used to solve the problem in (2) above; when a customer purchases funds, the probability of purchasing a specific type of fund can be used to solve the problem in (3) above. In summary, the models used are shown in Table (1) below:

[0101]

[0102] (1)

[0104] Step S20: Obtain the customer's data.

[0105] Specifically, as one implementation, the customer data used to train the multi-level product preference probability value model may include the following types of information:

[0106] 1. Personal basic information: such as age, gender, education background, occupation, etc. This information can provide clues about consumption habits and preferences to a certain extent.

[0107] 2. Financial status: This includes income level, asset status, and liability status. This information can reflect the client's financial capacity and risk tolerance.

[0108] 3. Consumer behavior data: This includes customers' purchase history, transaction records, and consumption preferences. For example, the types and frequency of financial products purchased, as well as the investment amounts for different products.

[0109] 4. Feedback and survey data: Collect customer feedback and satisfaction surveys to evaluate and understand customer needs and preferences.

[0110] It should be noted that customer data is not limited to the above implementation methods. The specific customer data used to train the product preference probability value model can vary depending on specific business needs and data availability.

[0111] Step S30: Input the customer's data into a pre-built multi-level product preference probability value model to obtain the customer's preference probability value for each product. The multi-level product preference probability value model is trained based on product classification and classification model.

[0112] The classification model used can be selected based on the classification of the products. For example, if a product at a certain level is classified into two categories, a binary classification model can be selected; if a product at a certain level is classified into multiple categories, a multi-class classification model can be selected.

[0113] As one implementation method, the product classification can be referred to Figure 4 , Figure 4 This is a schematic diagram of a three-level product system model in an embodiment of the present invention.

[0114] The Lv0 layer calculates whether customers have the intention to purchase wealth management products. The Lv1 layer constructs a multi-classification model based on major product categories to calculate customers' preferences for major product types such as entrusted wealth management, funds, deposits, insurance, and precious metals. The Lv2 layer constructs a binary / multi-classification model based on the subcategories corresponding to each major category to calculate customers' preferences for each subcategory, such as customers' preferences for equity, fixed income+, bond, and money market fund products.

[0115] Since the Lv0 layer only has two categories: customers' willingness to purchase wealth products, a binary classification model is used for the Lv0 layer. The difference in sample size between those with and without a preference for wealth products is not too large, so stratified sampling is not required.

[0116] Then, since there are multiple categories in the Lv1 layer, multi-class classification is used for modeling in the Lv1 layer. Although there are differences in sample size between the major categories, the sample size meets the modeling requirements and stratified sampling is not required.

[0117] Then, for layer Lv2, the implementation method is the same as layer Lv1. Multi-classification models are constructed for each sub-category under the major category. For example, a multi-classification model is constructed for the fund preference sub-category such as equity funds / fixed income+ funds / bond funds / money market funds; a multi-classification model is constructed for the wealth management product preference sub-category such as current / short-term / medium-term / medium-to-long-term / long-term wealth management.

[0118] Among them, the sample size of different subclasses of the same major category is not significantly different, so there is no need to use stratification to model them. The modeling method is the same as that of Lv1 layer.

[0119] It should be noted that the proposed multi-level product preference probability value model is not limited to constructing a three-level product preference probability value model based on the above three-level product system model. The multi-level product preference probability value model proposed in this invention can construct an adaptive multi-level product preference probability value model based on the product system model proposed by different businesses.

[0120] This embodiment, through the above-described scheme, acquires customer data; inputs the customer data into a pre-constructed multi-level product preference probability value model to obtain the customer's preference probability value for each product. The multi-level product preference probability value model is trained based on product classification and a classification model. This embodiment of the invention obtains a multi-level product preference probability value model based on product classification and a classification model. Since products at different levels may have different attributes and characteristics, these characteristics may play a crucial role in predicting customer preferences. By subdividing products into multiple levels and extracting corresponding features for each level, the model can more accurately capture subtle differences between products.

[0121] Reference Figure 3 , Figure 3 This is a flowchart illustrating an exemplary embodiment of the data processing method of the present invention. Based on the above... Figure 2 In the illustrated embodiment, step S30, which involves inputting the customer's data into a pre-built multi-level product preference probability value model to obtain the customer's preference probability value for each product, includes the following steps:

[0122] Step S31: Input the customer's data into a pre-built multi-level product preference probability value model to obtain the single-class probability value and combination probability value of each product. The multi-level product preference probability value model is trained based on product classification and classification model.

[0123] Since a customer may have preferences for multiple product categories, this multi-category approach is essentially an implementation of the multi-label model. Multiple categories are combined into a new category; when a customer purchases from all of these categories, the new category is matched. For example, if a customer has purchases of both large-denomination certificates of deposit (CDs) and structured deposits, the category "large-denomination CDs + structured deposits" is matched. Therefore, the prediction results of the multi-label model...

[0124] The result will not only yield the individual probability values ​​for each product, but also the combined probability values ​​for multiple products. See Table 5 (2) below:

[0125] (2)

[0127] Therefore, after obtaining the prediction results of the product preference probability value model at each level, it is necessary to integrate the prediction results in order to obtain the customer's preference probability value for each product.

[0128] Step S32: Integrate the single-category probability values ​​and combined probability values ​​of each product to obtain the customer's preference probability values ​​for each product.

[0129] One approach is to first perform a weighted summation of the individual and combined probability values ​​for each product. Based on the importance or weight of different products, the individual and combined probability values ​​are weighted separately to obtain a weighted score.

[0130] Then, the weighted scores are normalized. The normalized probability value is obtained by dividing the weighted score by the sum of all weighted scores.

[0131] Finally, using the normalized probability values, the customer's preference for each product is calculated. The customer's normalized probability value can be compared with the normalized probability values ​​of all customers to determine the strength of their preference for that product.

[0132] Specifically, as another implementation method, firstly, the single-class probability value and the combined probability value can be merged to obtain the model probability value of each product.

[0133] Then, the model probability values ​​of each product can be cascaded and multiplied to obtain the customer's preference probability value for each product.

[0134] It should be noted that the integration of the single-class probability values ​​and combined probability values ​​of each product is not limited to the two implementation methods listed above, and other processing can also be performed on the single-class probability values ​​and combined probability values ​​of each product.

[0135] This embodiment, through the above-described scheme, obtains customer data; inputs the customer data into a pre-constructed multi-level product preference probability value model to obtain the customer's preference probability value for each product. The multi-level product preference probability value model is trained based on product classification and a classification model. Specifically, inputting the customer data into the pre-constructed multi-level product preference probability value model yields individual and combined probability values ​​for each product; integrating the individual and combined probability values ​​of each product, the customer's preference probability value for each product is obtained.

[0136] This invention provides a multi-level product preference probability model trained based on product classification and a classification model. Since products at different levels may have different attributes and characteristics, these characteristics can play a crucial role in predicting customer preferences. By subdividing products into multiple levels and extracting corresponding features for each level, the model can more accurately capture subtle differences between products.

[0137] Reference Figure 5 , Figure 5 This is a flowchart illustrating another exemplary embodiment of the data processing method of the present invention.

[0138] Based on the above Figure 3 In the embodiment shown, step S32, which involves inputting the customer's data into a pre-built multi-level product preference probability value model to obtain the individual probability value and combination probability value of each product, includes:

[0139] Step S321: Combine the single-class probability value and the combined probability value to obtain the model probability value of each product.

[0140] The merging of the single-class probability value and the combined probability value can be shown in Table (3) below:

[0141] (3)

[0143] In this embodiment of the invention, by merging single-class probability values ​​and combined probability values, the influence of multiple factors on customer preferences can be comprehensively considered. Single-class probability values ​​represent the relative preference for each product, while combined probability values ​​represent the overall preference for combinations of multiple products. Merging them allows for a more comprehensive measurement of customers' overall preferences for different products.

[0144] Step S322: The model probability values ​​of each product are cascaded and multiplied to obtain the customer's preference probability value for each product.

[0145] Specifically, the following examples illustrate the cascaded multiplication of the model probability values ​​for each product:

[0146] In this process, after the Lv0, Lv1, and Lv2 layers of the model are trained and the model prediction is completed, the preference probability values ​​of each type of product are obtained. Finally, the model probability values ​​of each product are concatenated and multiplied. The Lv0 layer's model probability value is equal to the Lv0 layer's model probability value. The preference value of each category in the Lv1 layer is equal to the probability value of Lv0 multiplied by Lv1. The preference value of each category in the Lv2 layer is equal to the probability value of Lv0 multiplied by Lv1 multiplied by Lv2. As shown in Table (4) below:

[0147] (4)

[0149] Therefore, the preference for a major category is equal to the sum of its sub-categories. For example, fund preference = equity fund preference + fixed income fund preference + money market fund preference.

[0150] In this embodiment of the invention, the correlation between products can be taken into account by cascading the model probability values ​​of each product. For example, since a customer's preference for a product is related to its direct and indirect parent products, when calculating the preference probability value of that product, it is necessary to cascade the model probability value of that product with all its parent products (direct and indirect parent products). This correlation will be reflected in the final preference probability value, making the result more accurately reflect the customer's preference for the product.

[0151] Furthermore, by merging and cascading the individual and combined probability values ​​of each product, the need to compare the preference probabilities of different customers within the same product category and the preference probabilities of the same customer across different subcategories can be met at a micro level.

[0152] Furthermore, by merging and cascading the individual and combined probability values ​​of each product, we can obtain the customer's preference probability values ​​for different products. This comprehensive consideration and cascading multiplication method helps to provide more accurate preference prediction results. It combines the relative preference of each product with the correlation between products, and can more accurately characterize the customer's preference for product combinations, providing more valuable reference for decision-making.

[0153] This embodiment, through the above-described scheme, obtains customer data; inputs the customer data into a pre-constructed multi-level product preference probability value model to obtain the customer's preference probability value for each product. The multi-level product preference probability value model is trained based on product classification and a classification model. Specifically, inputting the customer data into the pre-constructed multi-level product preference probability value model yields individual and combined probability values ​​for each product; integrating the individual and combined probability values ​​for each product yields the customer's preference probability value for each product. Furthermore, merging the individual and combined probability values ​​yields the model probability value for each product; and then cascading and multiplying the model probability values ​​for each product yields the customer's preference probability value for each product.

[0154] This invention provides a multi-level product preference probability model trained based on product classification and a classification model. Since products at different levels may have different attributes and characteristics, these characteristics can play a crucial role in predicting customer preferences. By subdividing products into multiple levels and extracting corresponding features for each level, the model can more accurately capture subtle differences between products. Specifically, this invention combines single-class probability values ​​and combined probability values ​​to comprehensively consider the impact of multiple factors on customer preferences. Single-class probability values ​​represent the relative preference for each product, while combined probability values ​​represent the overall preference for combinations of multiple products. Combining them provides a more comprehensive measure of the overall customer preference for different products. Furthermore, this invention considers the correlation between products by cascading and multiplying the model probability values ​​of each product. For example, since a customer's preference for a product is related to its direct and indirect parent products, calculating the product's preference probability value requires cascading and multiplying the product's model probability value with all its parent products (direct and indirect). This correlation is reflected in the final preference probability value, making the result more accurately reflect the customer's product preference needs. Furthermore, merging and cascading the individual and combined probability values ​​of each product allows for comparison of preference probabilities among different customers within the same product category, and between the same customer across different subcategories. Moreover, merging and cascading the individual and combined probability values ​​of each product yields the customer's preference probability values ​​for different products. This comprehensive approach and cascading multiplication helps provide more accurate preference prediction results, combining the relative preference levels of each product with the correlations between products, thus more accurately characterizing the customer's preference for product combinations and providing more valuable references for decision-making.

[0155] Currently, for multiple products, the modeling method using machine learning to calculate the customer preference level for each product category can also be: model all products together, establish a multi-classification or multi-label model, and obtain the customer preference value for each product by model scoring, which can be compared with each other.

[0156] However, the above method has a drawback: while building a multi-classification or multi-label model for all products allows for comparison of product ratings and reflects customer preferences for each product at a micro level, the large differences in sample size between products necessitate the use of sampling methods for modeling. However, sampling methods may introduce biases or fail to fully represent the characteristics of the overall population, significantly disrupting the natural response rates of different products and resulting in inaccurate model ratings. To address these issues, the following implementation method can be proposed.

[0157] Reference Figure 6 , Figure 6 This is a flowchart illustrating another exemplary embodiment of the data processing method of the present invention.

[0158] Based on the above Figure 2 In the illustrated embodiment, step S20, prior to obtaining the customer's data, includes:

[0159] Step S10: Construct the multi-level product preference probability value model based on the product classification and the classification model.

[0160] Specifically, as one implementation method, sample data at various levels can be selected first.

[0161] Then, the corresponding classification model is selected based on the product classification at each level.

[0162] Finally, the corresponding classification model is trained based on the sample data at each level to obtain the multi-level product preference probability value model.

[0163] In this embodiment of the invention, the multi-level product preference probability value model is constructed based on the product classification and the classification model. This allows each level of the multi-level product preference probability value model to reduce the difference in the number of positive and negative samples or the number of samples in each category. In this way, the modeling does not require the use of sampling methods, the natural response rate does not change, and the obtained preference probability value is accurate.

[0164] After constructing the multi-level product preference probability value model, it can be applied to different marketing scenarios. Two examples are given below:

[0165] 1. When operations staff want to market a product, they need to find customers through the product. For example, if they want to market an equity fund, they need to acquire customers who are highly likely to prefer that equity fund. At this point, it is necessary to be able to compare the preference probabilities of different customers for the same product. The specific steps are as follows:

[0166] S10001, Obtain data on potential customer groups;

[0167] S10002, input the data of the potential customer group into the pre-built multi-level product preference probability value model to obtain the single probability value and the combination probability value of the potential customer group for the product to be marketed;

[0168] S10003, integrate the individual probability values ​​and the combination probability values ​​of the product to be marketed to obtain the preference probability values ​​of the potential customer group for the product to be marketed;

[0169] S10004, based on a preset probability threshold for product preference, the probability values ​​of each preference are filtered, and customers whose probability values ​​for the product preference are greater than the preset probability threshold for product preference are identified as high-quality customers.

[0170] 2. When operations staff need to manage a customer, they need to understand the customer's preferences to select suitable products. That is, operations staff need to understand the customer's preferences for different products, such as whether they prefer investment funds or wealth management products, bond funds or a longer-term investment approach. This requires comparing the probability values ​​of the same customer's preferences across different broad categories and different subcategories. The specific steps are as follows:

[0171] S20001, Obtain target customer data;

[0172] S20002, Input the target customer's data into the pre-built multi-level product preference probability value model to obtain the target customer's single-category probability value and combination probability value for each product;

[0173] S20003, integrate the single-category probability values ​​and combination probability values ​​of each product to obtain the target customer's preference probability value for each product;

[0174] S20004, Based on a preset target customer preference probability threshold, the preference probability values ​​of each product are filtered, and products whose preference probability values ​​are greater than the preset target customer preference probability threshold are recommended products.

[0175] This embodiment, through the above-described scheme, obtains customer data; inputs the customer data into a pre-constructed multi-level product preference probability value model to obtain the customer's preference probability value for each product. The multi-level product preference probability value model is trained based on product classification and a classification model. Specifically, the multi-level product preference probability value model is constructed based on the product classification and the classification model.

[0176] This invention provides a multi-level product preference probability model trained based on product classification and a classification model. Since products at different levels may have different attributes and characteristics, these characteristics can play a crucial role in predicting customer preferences. By subdividing products into multiple levels and extracting corresponding features for each level, the model can more accurately capture subtle differences between products. Specifically, this invention constructs the multi-level product preference probability model based on the product classification and the classification model, ensuring that each level of the model reduces the difference in positive and negative sample sizes or sample sizes across categories. This eliminates the need for sampling methods, maintains the natural response rate, and yields accurate preference probability values.

[0177] Reference Figure 7 , Figure 7 This is a flowchart illustrating another exemplary embodiment of the data processing method of the present invention.

[0178] Based on the above Figure 6 In the embodiment shown, step S10, constructing the multi-level product preference probability value model based on the product classification and the classification model, includes:

[0179] Step S11: Select sample data at each level.

[0180] The selected sample data at each level are independent of each other; that is, the selected sample data at each level can be the same or different.

[0181] Step S12: Select the corresponding classification model based on the product classification at each level.

[0182] If a product category at a certain level has only two classifications, a binary classification model can be selected; if a product category at a certain level has multiple classifications, a multi-classification model can be selected.

[0183] Step S13: Train the corresponding classification model based on the sample data at each level to obtain the multi-level product preference probability value model.

[0184] In this embodiment of the invention, the training order of the product preference probability models at each level in the multi-level product preference probability value model is not limited, based on the sample data at each level for training the corresponding classification model.

[0185] This embodiment, through the above-described scheme, obtains customer data; inputs the customer data into a pre-constructed multi-level product preference probability value model to obtain the customer's preference probability value for each product. The multi-level product preference probability value model is trained based on product classification and a classification model. Specifically, the multi-level product preference probability value model is constructed based on the product classification and the classification model. This involves selecting sample data at each level; selecting corresponding classification models based on the product classification at each level; and training the corresponding classification models based on the sample data at each level to obtain the multi-level product preference probability value model.

[0186] This invention provides a multi-level product preference probability model trained based on product classification and a classification model. Since products at different levels may have different attributes and characteristics, these characteristics can play a crucial role in predicting customer preferences. By subdividing products into multiple levels and extracting corresponding features for each level, the model can more accurately capture subtle differences between products. Specifically, this invention constructs the multi-level product preference probability model based on the product classification and the classification model, ensuring that each level of the model reduces the difference in positive and negative sample sizes or sample sizes across categories. This eliminates the need for sampling methods, maintains the natural response rate, and yields accurate preference probability values.

[0187] Reference Figure 8 , Figure 8 This is a flowchart illustrating another exemplary embodiment of the data processing method of the present invention.

[0188] Based on the above Figure 7 In the embodiment shown, step S13, after training the corresponding classification model based on the sample data at each level to obtain the multi-level product preference probability value model, includes:

[0189] Step S14: Select extended sample data.

[0190] To address the needs of business expansion or changes, after training the multi-level product preference probability value model, it is necessary to modify the product categories at one level, i.e., add, remove, or change product categories. The following example of adding product categories illustrates how to adjust the multi-level product preference probability value model accordingly.

[0191] First, obtain expanded sample data.

[0192] The expanded sample data includes relevant data on the need to add product categories.

[0193] Step S15: Iterate the multi-level product preference probability value model based on the extended sample data to obtain the iterated multi-level product preference probability value model.

[0194] In one implementation method, it is assumed that product categories need to be added to the Lv1 layer. Since the Lv1 layer corresponds to the second-level product preference probability value model, the second-level product preference probability value model can be iterated based on the extended sample data to obtain the iterated second-level product preference probability value model, and thus also obtain the iterated multi-level product preference probability value model.

[0195] In this case, it is only necessary to iterate the secondary product preference probability value model corresponding to the Lv1 layer, without needing to iterate the other product preference probability value models again.

[0196] In addition, when it is necessary to add subcategories under a major category, such as adding a subcategory of insurance preferences to the Lv2 layer, a new Lv2 layer model can be built without iterating or modeling the corresponding models of other layers, thus making it easier to modify the model.

[0197] It should be noted that the operations of deleting or changing product categories are the same as those described above.

[0198] In this embodiment of the invention, by iterating the multi-level product preference probability value model through expanded sample data, it can continuously adapt to changes in business needs and provide more accurate personalized recommendations for customers.

[0199] This embodiment, through the above-described scheme, obtains customer data; inputs the customer data into a pre-constructed multi-level product preference probability value model to obtain the customer's preference probability value for each product. The multi-level product preference probability value model is trained based on product classification and a classification model. Specifically, the multi-level product preference probability value model is constructed based on the product classification and the classification model. This involves selecting sample data at each level; selecting corresponding classification models based on each level of product classification; and training the corresponding classification models based on the sample data at each level to obtain the multi-level product preference probability value model. Additionally, extended sample data is selected; the multi-level product preference probability value model is iterated based on the extended sample data to obtain an iterated multi-level product preference probability value model.

[0200] This invention provides a multi-level product preference probability model trained based on product classification and a classification model. Since products at different levels may have different attributes and characteristics, these characteristics can play a crucial role in predicting customer preferences. By subdividing products into multiple levels and extracting corresponding features for each level, the model can more accurately capture subtle differences between products. Specifically, this invention constructs the multi-level product preference probability model based on the product classification and the classification model, ensuring that each level of the model reduces the difference in positive and negative sample sizes or sample sizes across categories. This eliminates the need for sampling methods, maintains the response rate, and yields accurate preference probability values. Furthermore, this invention iterates the multi-level product preference probability model by expanding sample data, continuously adapting to changing business needs and enabling more accurate personalized recommendations for customers.

[0201] Reference Figure 9 , Figure 9 This is a flowchart illustrating another exemplary embodiment of the data processing method of the present invention.

[0202] Based on the above Figure 7 In the embodiment shown, step S11, selecting sample data at each level, includes:

[0203] Step S111: Select primary sample data.

[0204] Step S112, and / or, select secondary sample data.

[0205] Step S113, and / or, select three levels of sample data.

[0206] The first, second, and third level sample data are selected independently and do not affect each other.

[0207] In this embodiment of the invention, the order in which first-, second-, and third-level sample data are selected is not limited.

[0208] Reference Figure 10 , Figure 10 This is a flowchart illustrating another exemplary embodiment of the data processing method of the present invention.

[0209] Based on the above Figure 7 The illustrated embodiment includes primary product classification, secondary product classification, and tertiary product classification. Step S12, selecting the corresponding classification model based on the product classification at each level, includes:

[0210] Step S121: Select a primary classification model based on the primary product classification.

[0211] Step S122, and / or, select a secondary classification model based on the secondary product classification information.

[0212] Step S123, and / or, select a three-level classification model based on the three-level product classification.

[0213] Specifically, as one implementation method, the primary product classification includes preferences for wealth management products and preferences for those without. The secondary product classification includes preferences for funds, insurance, precious metals, wealth management products, and deposits. The tertiary product classification is further categorized based on the secondary product classification. For example, fund preferences can be classified by asset type into equity funds / fixed income+ funds / bond funds / money market funds. Wealth management product preferences can be classified by asset type into equity wealth management / fixed income+ wealth management / fixed income wealth management / cash wealth management. Wealth management product preferences can also be classified by product term into current / short-term / medium-term / medium-to-long-term / long-term wealth management.

[0214] Reference Figure 11 , Figure 11 This is a flowchart illustrating another exemplary embodiment of the data processing method of the present invention.

[0215] Based on the above Figure 7 In the embodiment shown, step S13, training the corresponding classification model based on the sample data at each level to obtain the multi-level product preference probability value model, includes:

[0216] Step S131: Train the first-level classification model based on the first-level sample data to obtain the first-level product preference probability value model.

[0217] Step S132, and / or, train the secondary classification model based on the secondary sample data to obtain a secondary product preference probability value model.

[0218] Step S133, and / or, train the three-level classification model based on the three-level sample data to obtain a three-level product preference probability value model.

[0219] It should be noted that the embodiments of the present invention do not limit the order of training the above-mentioned models.

[0220] In this embodiment of the invention, steps are proposed to create a three-level product preference probability value model. By selecting the first, second, and third level product classification, a more detailed and accurate division of product classification can be achieved. This enables the model to better understand and identify the characteristics of different products at the same level, thereby improving the model's ability to predict user preferences.

[0221] This embodiment, through the above-described scheme, obtains customer data; inputs the customer data into a pre-constructed multi-level product preference probability value model to obtain the customer's preference probability value for each product. The multi-level product preference probability value model is trained based on product classification and a classification model. Specifically, the multi-level product preference probability value model is constructed based on the product classification and the classification model. This involves selecting sample data at each level; selecting corresponding classification models based on the product classification at each level; and training the corresponding classification models based on the sample data at each level to obtain the multi-level product preference probability value model. Further, select primary sample data; and / or, select secondary sample data; and / or, select tertiary sample data; obtain a primary classification model based on the primary product classification; and / or, obtain a secondary classification model based on the secondary product classification; and / or, obtain a tertiary classification model based on the tertiary product classification; train the primary classification model based on the primary sample data to obtain a primary product preference probability value model; and / or, train the secondary classification model based on the secondary sample data to obtain a secondary product preference probability value model; and / or, train the tertiary classification model based on the tertiary sample data to obtain a tertiary product preference probability value model.

[0222] This invention provides a multi-level product preference probability model trained based on product classification and a classification model. Since products at different levels may have different attributes and characteristics, these characteristics can play a crucial role in predicting customer preferences. By subdividing products into multiple levels and extracting corresponding features for each level, the model can more accurately capture subtle differences between products. Specifically, this invention constructs the multi-level product preference probability model based on the product classification and the classification model, enabling each level of the model to reduce the difference in positive and negative sample sizes or sample sizes across categories. This eliminates the need for sampling methods, maintains the natural response rate, and ensures accurate preference probability values. Furthermore, this invention proposes steps for creating a three-level product preference probability model. By selecting first, second, and third-level product classifications, a more detailed and accurate product classification can be achieved. This allows the model to better understand and identify the characteristics of different products at the same level, thereby improving the model's ability to predict user preferences.

[0223] Furthermore, embodiments of this application also propose a data processing apparatus for calculating the probability values ​​of customer preferences for each product. The data processing apparatus includes:

[0224] The data acquisition module acquires the customer's data;

[0225] The model prediction module inputs the customer's data into a pre-built multi-level product preference probability value model to obtain the customer's preference probability value for each product. The multi-level product preference probability value model is trained based on product classification and classification model.

[0226] The principle and implementation process of data processing in this embodiment are explained in the above embodiments and will not be repeated here.

[0227] Furthermore, this application also proposes a terminal device, which includes a memory, a processor, and a data processing program stored in the memory and executable on the processor. When the data processing program is executed by the processor, it implements the steps of the data processing method described above.

[0228] Since this data processing program employs all the technical solutions of all the foregoing embodiments when executed by the processor, it has at least all the beneficial effects brought about by all the technical solutions of all the foregoing embodiments, which will not be elaborated here.

[0229] Furthermore, embodiments of this application also propose a computer-readable storage medium storing a data processing program, which, when executed by a processor, implements the steps of the data processing method described above.

[0230] Since this data processing program employs all the technical solutions of all the foregoing embodiments when executed by the processor, it has at least all the beneficial effects brought about by all the technical solutions of all the foregoing embodiments, which will not be elaborated here.

[0231] This embodiment, through the above-described scheme, specifically obtains the customer's data; inputs the customer's data into a pre-constructed multi-level product preference probability value model to obtain the customer's preference probability value for each product. The multi-level product preference probability value model is trained based on product classification and a classification model. This embodiment of the invention obtains a multi-level product preference probability value model based on product classification and a classification model. Since products at different levels may have different attributes and characteristics, these characteristics may play a crucial role in predicting customer preferences. By subdividing products into multiple levels and extracting corresponding features for each level, the model can more accurately capture subtle differences between products.

[0232] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or approach that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or approach. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or approach that includes that element.

[0233] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0234] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, controlled terminal, or network device, etc.) to execute the methods of each embodiment of the present invention.

[0235] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A data processing method, characterized in that, The method is used to calculate the probability value of customer preference for each product, and the method includes the following steps: Obtain the customer's data; The customer's data is input into a pre-built multi-level product preference probability value model to obtain the customer's preference probability value for each product. The multi-level product preference probability value model is trained based on product classification and classification model. The step of inputting the customer's data into a pre-built multi-level product preference probability value model to obtain the customer's preference probability value for each product includes: The customer's data is input into a pre-built multi-level product preference probability value model to obtain the single-category probability value and combination probability value of each product. By integrating the individual probability values ​​and combined probability values ​​of each product, the customer's preference probability value for each product is obtained. The step of integrating the individual probability values ​​and combined probability values ​​of each product to obtain the customer's preference probability value for each product includes: The single-class probability value and the combined probability value are merged to obtain the model probability value of each product; The model probability values ​​of each product are cascaded and multiplied to obtain the customer's preference probability value for each product.

2. The method according to claim 1, characterized in that, Prior to the step of obtaining the customer's data, the following are included: The multi-level product preference probability value model is constructed based on the product classification and the classification model.

3. The method according to claim 2, characterized in that, The steps of constructing the multi-level product preference probability value model based on the product classification and the classification model include: Select sample data at each level; The corresponding classification model is selected based on the product classification at each level; The corresponding classification model is trained based on the sample data at each level to obtain the multi-level product preference probability value model.

4. The method according to claim 3, characterized in that, The step of training the corresponding classification model based on the sample data at each level to obtain the multi-level product preference probability value model includes: Select extended sample data; The multi-level product preference probability value model is iterated based on the extended sample data to obtain the iterated multi-level product preference probability value model.

5. The method according to claim 3, characterized in that, The steps for selecting sample data at each level include: Select primary sample data; And / or, select secondary sample data; And / or, select three levels of sample data; The product classification at each level includes primary product classification, secondary product classification, and tertiary product classification. The step of selecting the corresponding classification model based on the product classification at each level includes: A primary classification model is selected based on the aforementioned primary product classification. And / or, a secondary classification model is selected based on the secondary product classification information; And / or, a three-level classification model is selected based on the three-level product classification. The step of training the corresponding classification model based on the sample data at each level to obtain the multi-level product preference probability value model includes: The first-level classification model is trained based on the first-level sample data to obtain the first-level product preference probability value model; And / or, the secondary classification model is trained based on the secondary sample data to obtain a secondary product preference probability value model; And / or, the three-level classification model is trained based on the three-level sample data to obtain a three-level product preference probability value model.

6. The method according to claim 5, characterized in that, The primary product classification includes preferences for wealth management products and preferences for those without. The secondary product classification includes preferences for funds, insurance, precious metals, wealth management products, and deposits. The tertiary product classification is further categorized based on the secondary product classification.

7. A data processing apparatus, characterized in that, The device is used to calculate the probability value of a customer's preference for each product, and the device includes: The data acquisition module acquires the customer's data; The model prediction module inputs the customer's data into a pre-built multi-level product preference probability value model to obtain the customer's preference probability value for each product. The multi-level product preference probability value model is trained based on product classification and classification model. The step of inputting the customer's data into a pre-built multi-level product preference probability value model to obtain the customer's preference probability value for each product includes: The customer's data is input into a pre-built multi-level product preference probability value model to obtain the single-category probability value and combination probability value of each product. By integrating the individual probability values ​​and combined probability values ​​of each product, the customer's preference probability value for each product is obtained. The step of integrating the individual probability values ​​and combined probability values ​​of each product to obtain the customer's preference probability value for each product includes: The single-class probability value and the combined probability value are merged to obtain the model probability value of each product; The model probability values ​​of each product are cascaded and multiplied to obtain the customer's preference probability value for each product.

8. A data processing terminal device, characterized in that, The data processing terminal device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the data processing method as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the data processing method as described in any one of claims 1-6.

Images