Method for predicting purchase willingness of customer and related equipment thereof

By obtaining and analyzing marketing customers' historical purchasing intention and call text data, using preset speech analysis models to extract speech characteristics, combined with the intention prediction model, the problem of low accuracy in purchasing intention prediction in the existing technology is solved, and the more Accurate customer needs understanding and marketing strategy formulation.

CN119067709APending Publication Date: 2024-12-03WEBANK (CHINA)
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202411193731.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-28
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The existing technology is low in accuracy when predicting marketing customers' purchasing intentions and cannot effectively understand customers' deep needs and intentions.

Method used

By obtaining the historical purchasing intention data and call text data of marketing customers, using the preset speech analysis model to extract the target speech characteristics from the historical call text, and combining the target purchase intention characteristics, purchasing intention prediction is made through the preset intention prediction model. Characterize the relationship between purchasing intention characteristics, speech characteristics and willingness.

Benefits of technology

It improves the accuracy of forecasting marketing customers' purchasing intentions, can truly reflect customers' historical purchasing rules and potential needs, and improves the accuracy and effectiveness of forecasts.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119067709A_ABST
    Figure CN119067709A_ABST
Patent Text Reader

Abstract

The invention discloses a customer purchase willingness prediction method and related equipment thereof, and relates to the technical field of artificial intelligence. Target purchase willingness features of marketing customers are obtained based on obtained historical purchase willingness data; the target verbal skill characteristics are learned and extracted from the obtained historical conversation text data through a preset verbal skill analysis model, and the preset verbal skill analysis model can represent the association relationship between the conversation text data and the verbal skill characteristics, so that the target verbal skill characteristics of marketing customers can be accurately determined through the preset verbal skill analysis model; because the preset willingness prediction model can represent the incidence relation among the purchase willingness characteristics, the verbal skill characteristics and the purchase willingness, the target purchase willingness of the marketing customer can be accurately determined through the preset willingness prediction model; the target purchase willingness features can reflect historical purchase rules of marketing customers; therefore, the method can improve the prediction accuracy of the purchase willingness of the marketing customer.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method for predicting customer purchase intention and related equipment. Background Art

[0002] Currently, the purchasing intention of marketing customers is usually predicted based on their historical transaction records. However, although this method can reveal the purchasing habits of customers to a certain extent, it can only reflect the customers' past behavior patterns, and has limited understanding of the customers' deep-seated needs and intentions, resulting in low accuracy in predicting the purchasing intention of marketing customers. Summary of the invention

[0003] The main purpose of this application is to provide a method for predicting customer purchase intention and related equipment, aiming to solve the technical problem of low accuracy in predicting the purchase intention of marketing customers.

[0004] To achieve the above purpose, the present application proposes a method for predicting customer purchase intention, wherein the method comprises:

[0005] Obtain marketing customers’ historical purchase intention data and historical call text data;

[0006] Based on the historical purchase intention data, determine the target purchase intention characteristics of the marketing customer, and learn and extract the target speech characteristics from the historical call text data through a preset speech analysis model, wherein the preset speech analysis model is used to characterize the correlation between the call text data and the speech characteristics, the target purchase intention characteristics reflect the historical purchase rules of the marketing customer, and the target speech characteristics reflect the potential needs of the marketing customer;

[0007] Based on the target purchase intention characteristics and the target speech characteristics, the purchase intention of the marketing customer is predicted through a preset intention prediction model to obtain the target purchase intention, wherein the preset intention prediction model is used to characterize the correlation between the purchase intention characteristics, speech characteristics and purchase intention.

[0008] In one embodiment, the preset speech analysis model includes a first speech analysis model, and before the step of obtaining the marketing customer's historical purchase intention data and historical call text data, the following steps are included:

[0009] Acquire a training sample and a sample label of the training sample, wherein the training sample includes first call text data of a historical customer, and the sample label of the training sample includes a first speech label of the first call text data;

[0010] Based on the first call text data and the first speech tags, the preset first model to be trained is iteratively trained to obtain a first speech analysis model.

[0011] In one embodiment, the preset first model to be trained includes a preset large model, and the step of iteratively training the preset first model to be trained based on the first call text data and the first speech tag to obtain the first speech analysis model includes:

[0012] Based on the first call text data, pre-training a preset large model to obtain a content analysis model, wherein the content analysis model has the ability to understand the call content in the call text data;

[0013] Based on the first call text data, the first speech label, and a preset first prompt, the content analysis model is fine-tuned to obtain a first speech analysis model, wherein the preset first prompt is used to prompt the content analysis model to output speech features according to the call content in the call text data.

[0014] In one embodiment, the preset speech analysis model further includes a second speech analysis model. After the step of iteratively training the preset first model to be trained based on the first call text data and the first speech label to obtain the first speech analysis model, the method includes:

[0015] Acquire a test sample and a sample label of the test sample, wherein the test sample includes second call text data of a historical customer, and the sample label of the test sample includes a second speech label of the second call text data;

[0016] Based on the second call text data, the first speech analysis model is tested to obtain a test result;

[0017] Determining whether the test result is consistent with the second speech tag;

[0018] Use the test results that are inconsistent with the second speech label as rejection labels, and use the test results that are consistent with the second speech label as acceptance labels;

[0019] Based on the rejection labels and the acceptance labels, the first speech analysis model is optimized to obtain an optimized second speech analysis model.

[0020] In one embodiment, the step of optimizing the first speech analysis model based on the rejection label and the acceptance label to obtain an optimized second speech analysis model includes:

[0021] Based on the preset call text data and the preset first prompt, the rejection label and the acceptance label are scored respectively by the frozen first speech analysis model to obtain a first scoring result, and based on the preset call text data and the preset first prompt, the rejection label and the acceptance label are scored respectively by the third speech analysis model to obtain a second scoring result;

[0022] The third speech analysis model that has not been iteratively trained is the first speech analysis model, the first scoring result is used to characterize the recognition of the rejection label and the acceptance label by the frozen first speech analysis model, and the second scoring result is used to characterize the recognition of the rejection label and the acceptance label by the third speech analysis model;

[0023] Based on the first scoring result and the second scoring result, calculating a loss value by using a preset loss function;

[0024] Based on the loss value, the model parameters of the third speech analysis model are updated, and the step of scoring the rejection label and the acceptance label respectively based on the preset call text data and the preset first prompt is returned accordingly through the frozen first speech analysis model, and iterative training is continued until the second speech analysis model that meets the preset first accuracy condition is obtained.

[0025] In one embodiment, the preset first model to be trained also includes a general large model, and the general large model has a text analysis capability. The step of iteratively training the preset first model to be trained based on the first call text data and the first speech tag to obtain the first speech analysis model includes:

[0026] Based on the first call text data, the first speech tag, and a preset second prompt, the general big model is fine-tuned to obtain a first speech analysis model, wherein the preset second prompt is used to prompt the general big model to output speech features according to the call content in the call text data.

[0027] In one embodiment, before the step of obtaining the marketing customer's historical purchase intention data and historical call text data, the following steps are included:

[0028] Obtain historical customer purchase intention data and call text data;

[0029] Based on the purchase intention data of the historical customers, determine the historical purchase intention features of the historical customers, and learn and extract the historical speech features from the call text data of the historical customers through a preset speech analysis model;

[0030] Determine a purchase intention label based on the historical purchase intention feature and the historical speech feature, wherein the purchase intention label includes the purchase probability of the historical customer;

[0031] Based on the historical purchase intention features, the historical speech features and the purchase intention labels, the preset second model to be trained is iteratively trained to obtain a willingness prediction model.

[0032] In one embodiment, the preset second model to be trained includes a binary classification model to be trained, and the step of iteratively training the preset second model to be trained based on the historical purchase intention feature, the historical speech feature and the purchase intention label to obtain the intention prediction model includes:

[0033] Inputting the historical purchase intention feature, the historical speech feature and the purchase intention label into the binary classification model to be trained, predicting the purchase intention of the historical customer through the binary classification model to be trained, and obtaining the historical purchase probability of the historical customer;

[0034] Calculating the error between the historical purchase probability and the purchase intention label to obtain an error calculation result;

[0035] Based on the error calculation result, the model parameters of the binary classification model to be trained are updated, and the step of inputting the historical purchase intention features, the historical speech features and the purchase intention labels into the binary classification model to be trained is returned accordingly, until a willingness prediction model that meets the preset second precision condition is obtained.

[0036] In one embodiment, the step of predicting the marketing customer's purchase intention based on the target purchase intention feature and the target speech feature by using a preset intention prediction model to obtain the target purchase intention includes:

[0037] Based on the target purchase intention characteristics and the target speech characteristics, the purchase intention of the marketing customer is predicted by a preset intention prediction model to obtain a target purchase probability, and the target purchase probability is positively correlated with the purchase intention of the marketing customer;

[0038] After the step of predicting the marketing customer's purchase intention based on the target purchase intention feature and the target speech feature by using a preset intention prediction model to obtain the target purchase intention, the step includes:

[0039] The marketing customers are sorted from large to small according to the corresponding target purchase probability, and the marketing customers ranked in the front preset positions are determined as the customers to be marketed, so that the marketing personnel can market to the customers to be marketed.

[0040] In addition, to achieve the above-mentioned purpose, the present application also proposes a device for predicting customer purchase intention, the device for predicting customer purchase intention comprising:

[0041] The acquisition module is used to obtain the historical purchase intention data and historical call text data of marketing customers;

[0042] An extraction module, used to determine the target purchase intention characteristics of the marketing customer based on the historical purchase intention data, and learn to extract the target speech characteristics from the historical call text data through a preset speech analysis model, wherein the preset speech analysis model is used to characterize the correlation between the call text data and the speech characteristics, the target purchase intention characteristics reflect the historical purchase rules of the marketing customer, and the target speech characteristics reflect the potential needs of the marketing customer;

[0043] A prediction module is used to predict the purchasing intention of the marketing customer based on the target purchasing intention characteristics and the target speech characteristics through a preset intention prediction model to obtain the target purchasing intention, wherein the preset intention prediction model is used to characterize the correlation between the purchasing intention characteristics, the speech characteristics and the purchasing intention.

[0044] In addition, to achieve the above-mentioned purpose, the present application also proposes a device for predicting customer purchasing intention, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, and the computer program is configured to implement the steps of the method for predicting customer purchasing intention as described above.

[0045] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the method for predicting customer purchasing intention as described above are implemented.

[0046] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the steps of the method for predicting customer purchase intention as described above.

[0047] The above technical solution proposed in this application has the following technical effects:

[0048] The present application obtains the target purchase intention characteristics of the marketing customers based on the acquired historical purchase intention data, and learns to extract the target speech characteristics from the acquired historical call text data through a preset speech analysis model. Since the preset speech analysis model can characterize the correlation between call text data and speech characteristics, the target speech characteristics of the marketing customers can be accurately determined through the preset speech analysis model. It can be understood that the target speech characteristics can reflect the potential needs shown by the marketing customers in the historical calls. Since the preset intention prediction model can characterize the correlation between purchase intention characteristics, speech characteristics and purchase intention, the purchase intention of the marketing customers is predicted through the preset intention prediction model, and the target purchase intention of the marketing customers can be accurately determined. The target purchase intention characteristics can reflect the historical purchase rules of the marketing customers, such as purchase preferences or purchase habits. Therefore, while taking into account the historical purchase rules and potential needs of the customers, the present application can more truly and accurately reflect the purchase intention of the marketing customers, thereby improving the accuracy of predicting the purchase intention of the marketing customers. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0050] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0051] Figure 1 A first flow chart provided for the first embodiment of the method for predicting customer purchase intention of the present application;

[0052] Figure 2 A second flow chart provided for the second embodiment of the method for predicting customer purchase intention of the present application;

[0053] Figure 3 A third flow chart provided for the third embodiment of the method for predicting customer purchase intention of the present application;

[0054] Figure 4 A schematic diagram of the first scenario provided in Embodiment 3 of the method for predicting customer purchase intention of this application;

[0055] Figure 5 A fourth flow chart provided for the fourth embodiment of the method for predicting customer purchase intention of the present application;

[0056] Figure 6A second scenario diagram provided for the fourth embodiment of the method for predicting customer purchase intention of the present application;

[0057] Figure 7 A schematic diagram of the module structure of a device for predicting customer purchase intention according to an embodiment of the present application;

[0058] Figure 8 Schematic diagram of the device structure of the hardware operating environment involved in the method for predicting customer purchasing intention in the embodiment of the present application.

[0059] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0060] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0061] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0062] It should be noted that the execution subject of this embodiment may be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of realizing the above functions, a device for predicting customer purchase intention, etc. The following takes the device for predicting customer purchase intention as an example to illustrate this embodiment and the following embodiments.

[0063] Based on this, the present application embodiment provides a method for predicting customer purchase intention, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the method for predicting customer purchase intention of the present application.

[0064] In this embodiment, the method for predicting customer purchase intention includes steps S10 to S40:

[0065] Step S10, obtaining the marketing customer's historical purchase intention data and historical call text data;

[0066] It should be noted that marketing customers include existing customers and new customers; the historical purchase intention data of marketing customers may be data related to customer purchase intention collected from multiple channels, and the historical purchase intention data includes customer behavior data, transaction data, customer feedback and evaluation, and customer personal information, etc.

[0067] Among them, behavioral data can be browsing behavior: the time customers spend browsing pages on the website and the types of products they browse, etc.; it can be search records: customers' search keywords, search frequency, and post-search behavior (for example, clicking on product details and purchasing products, etc.); transaction data can be purchase records: the number of purchases and the frequency of logging into the purchasing platform, etc.; customer feedback and evaluations can be customers' evaluations, questions, likes, or shares of products, etc.; the customer's personal information can be the customer's age, and different age groups have different needs and interests; it can be the city they belong to, and the market demand and consumption habits in different regions are different; it can be professional identity, and the needs and purchasing power of different professional groups are different.

[0068] It should also be noted that the historical call text data can be determined based on the historical call data of marketing customers. It can be understood that since the historical call data of marketing customers are voice data, the historical call data can be converted into historical call text data through voice recognition technology (Automatic Speech Recognition, ASR), which facilitates the subsequent processing of the historical call text data.

[0069] Among them, historical call data includes call records between customers and marketing personnel, as well as call records between customers and other non-marketing personnel.

[0070] Step S20, based on the historical purchase intention data, determine the target purchase intention feature of the marketing customer, and learn and extract the target speech feature from the historical call text data through a preset speech analysis model, wherein the preset speech analysis model is used to characterize the correlation between the call text data and the speech feature, the target purchase intention feature reflects the historical purchase rules of the marketing customer, and the target speech feature reflects the potential needs of the marketing customer;

[0071] Currently, the purchasing intention of marketing customers is usually predicted based on their historical transaction records. However, although this method can reveal the purchasing habits of customers to a certain extent, it can only reflect the customers' past behavior patterns and has limited understanding of the customers' deep-seated needs and intentions, resulting in low accuracy in predicting the purchasing intention of marketing customers.

[0072] This embodiment obtains target purchase intention features that can reflect the historical purchase patterns of the marketing customers based on the historical purchase intention data, and learns and extracts target speech features that can reflect the potential needs of the marketing customers from the historical call text data through a preset speech analysis model, taking into account not only the historical purchase patterns of the marketing customers (for example, purchase preferences or purchase habits, etc.), but also the potential needs of the marketing customers (for example, potential interests or immediate interests, etc.).

[0073] Among them, the preset speech analysis model can be obtained through iterative training based on historical customer call text data. It can be understood that the trained preset speech analysis model can characterize the correlation between call text data and speech features; therefore, by learning to extract target speech features from the historical call text data through the preset speech analysis model, the accuracy of the target speech features can be guaranteed.

[0074] Specifically, the target speech features include opening introduction, product introduction, purchase intention inquiry, introduction to product transaction costs, appointment for next call, closing remarks, thank you remarks, product consultation and price consultation, company introduction, interest rate introduction, customer retention, etc.

[0075] It should be noted that the specific implementation method of determining the target purchase intention characteristics of the marketing customers based on the historical purchase intention data may be:

[0076] Based on preset keywords, the target purchase intention features of the marketing customers are screened from the historical purchase intention data; and / or the target purchase intention features of the marketing customers are extracted by learning from the historical purchase intention data through a preset feature extraction model.

[0077] Among them, the target purchase intention characteristics include behavioral characteristics, transaction characteristics, characteristics of customer feedback and evaluation, and personal characteristics of customers, etc.; the preset feature extraction model can be based on historical customer purchase intention data and obtained through iterative training. It can be understood that the trained preset feature extraction model can characterize the correlation between purchase intention data and purchase intention characteristics; therefore, by learning to extract the target purchase intention characteristics from the historical purchase intention data through the preset feature extraction model, the accuracy of the target purchase intention characteristics can be guaranteed.

[0078] Step S30, based on the target purchase intention characteristics and the target speech characteristics, the purchase intention of the marketing customer is predicted by a preset intention prediction model to obtain the target purchase intention, wherein the preset intention prediction model is used to characterize the correlation between the purchase intention characteristics, speech characteristics and purchase intention.

[0079] Since the target purchase intention feature can reflect the historical purchase pattern of the marketing customer, and the target speech feature can reflect the potential demand of the marketing customer, this embodiment predicts the purchase intention of the marketing customer through a preset intention prediction model based on the target purchase intention feature and the target speech feature to obtain the target purchase intention. The target purchase intention is obtained while taking into account the potential demand of the marketing customer's historical purchase pattern. Therefore, the target purchase intention can more truly and accurately reflect the marketing customer's purchase intention.

[0080] Among them, the preset intention prediction model can be obtained through iterative training based on historical customer purchase intention data. It can be understood that the trained preset intention prediction model can characterize the relationship between purchase intention characteristics, speech characteristics and purchase intention; therefore, by predicting the purchase intention of the marketing customer through the preset intention prediction model, the accuracy of the predicted target purchase intention can be guaranteed.

[0081] Specifically, the specific implementation method of predicting the marketing customer's purchase intention based on the target purchase intention feature and the target speech feature through a preset intention prediction model to obtain the target purchase intention may be:

[0082] Based on the target purchase intention characteristics and the target speech characteristics, the purchase intention of the marketing customer is predicted by a preset intention prediction model to obtain a target purchase probability, and the target purchase probability is positively correlated with the purchase intention of the marketing customer;

[0083] It should be noted that the target purchase intention can be determined based on the target purchase probability. The target purchase probability is positively correlated with the purchase intention of the marketing customer. The greater the target purchase probability, the higher the target purchase intention. Conversely, the smaller the target purchase probability, the lower the target purchase intention. It can be understood that the target purchase intention of marketing customers can be quantified by means of purchase probability, which is convenient for subsequent screening of marketing customers.

[0084] Specifically, based on the target purchase intention feature and the target speech feature, the purchase intention of the marketing customer is predicted by a preset intention prediction model. Before obtaining the target purchase intention, the target purchase intention feature and the target speech feature can also be format converted. For example, the target purchase intention feature and the target speech feature can be converted into One-Hot Encoding, etc., to obtain the converted purchase intention feature and the converted target speech feature.

[0085] Furthermore, after the step of predicting the marketing customer's purchase intention based on the target purchase intention feature and the target speech feature by using a preset intention prediction model to obtain the target purchase intention, the following steps may be performed:

[0086] The marketing customers are sorted from large to small according to the corresponding target purchase probability, and the marketing customers ranked in the front preset positions are determined as the customers to be marketed, so that the marketing personnel can market to the customers to be marketed.

[0087] Specifically, the marketing customers are sorted from large to small according to the corresponding target purchase probability, so that the marketing customers with high purchase intention can be screened quickly and accurately, thereby improving the screening efficiency; the marketing customers ranked in the first preset number of digits are customers with a greater target purchase probability, that is, the marketing customers ranked in the first preset number of digits are customers with a higher target purchase intention. Therefore, the marketing customers ranked in the first preset number of digits are regarded as customers to be marketed, and the marketing personnel can carry out targeted marketing to the customers to be marketed, thereby improving customer conversion rate and customer satisfaction, thereby improving marketing efficiency; reducing marketing to customers with low purchase intention, reducing marketing costs, reducing excessive marketing to customers with low purchase intention, improving marketing effectiveness and improving customer experience at the same time.

[0088] Furthermore, the score of the marketing customer can be determined based on the correlation between the target purchase probability and the score. For example, when the target purchase probability of the marketing customer is 0.9, the corresponding marketing customer score is 90 points; when the target purchase probability of the marketing customer is 0.4, the corresponding marketing customer score is 40 points, and so on. It can be understood that the higher the score, the higher the purchasing intention of the marketing customer, and the lower the score, the lower the purchasing intention of the marketing customer. Therefore, when screening influential customers, screening can be based on the score.

[0089] In this embodiment, based on the acquired historical purchase intention data, the target purchase intention characteristics of the marketing customer are obtained, and the target speech characteristics are learned and extracted from the acquired historical call text data through a preset speech analysis model. Since the preset speech analysis model can characterize the correlation between call text data and speech characteristics, the target speech characteristics of the marketing customer can be accurately determined through the preset speech analysis model. It can be understood that the target speech characteristics can reflect the potential needs shown by the marketing customer in the historical call process; since the preset intention prediction model can characterize the correlation between purchase intention characteristics, speech characteristics and purchase intention, the purchase intention of the marketing customer is predicted through the preset intention prediction model, and the target purchase intention of the marketing customer can be accurately determined; the target purchase intention characteristics can reflect the historical purchase rules of the marketing customer, such as purchase preferences or purchase habits; therefore, this embodiment can more realistically and accurately reflect the purchase intention of the marketing customer while taking into account the customer's historical purchase rules and potential needs, thereby improving the accuracy of predicting the purchase intention of the marketing customer.

[0090] Furthermore, by sorting the marketing customers from large to small according to the corresponding target purchase probability, and determining the marketing customers ranked in the front preset number as the customers to be marketed, marketing is conducted on the customers to be marketed with higher purchasing intention, thereby improving customer conversion rate and customer satisfaction, thereby improving marketing efficiency.

[0091] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above embodiment can be referred to the above introduction, and will not be repeated in the following. Figure 2 Before step S10, the method for predicting customer purchase intention further includes steps S01 to S02:

[0092] Step S01, obtaining a training sample and a sample label of the training sample, wherein the training sample includes first call text data of a historical customer, and the sample label of the training sample includes a first speech label of the first call text data;

[0093] It should be noted that since the text data of the first call of historical customers can reflect the language used by marketing personnel and / or historical customers during the call, the language can indirectly reflect the potential needs of marketing customers; for example, when the language of "use retention introduction" appears during a product call, it can be considered that the marketing customer's willingness to buy is low, and when the language of "product consultation" appears during a product call, it can be considered that the marketing customer's willingness to buy is high.

[0094] Therefore, the first call text data of the historical customer is used as a training sample, and a first speech tag is marked for the training sample. The first speech tag is used to characterize the speech used by the marketing personnel and / or the historical customer in the first call text data.

[0095] Specifically, the first speech tag can be determined based on a specific marketing scenario. For example, for a marketing loan scenario, the first speech tag can be an introduction to the loan, an inquiry about the loan amount, etc. For a marketing real estate scenario, the first speech tag can be an introduction to the geographical location, an inquiry about the real estate price, etc.

[0096] Step S02, based on the first call text data and the first speech tag, iteratively train the preset first model to be trained to obtain a first speech analysis model.

[0097] It should be noted that the preset speech analysis model includes a first speech analysis model; based on the above-mentioned first call text data and the first speech label, the preset first model to be trained is iteratively trained to obtain the first speech analysis model.

[0098] Among them, the preset first model to be trained can be a preset large model (Large Model), and the preset large model refers to a machine learning model containing a large number of parameters and computing resources.

[0099] Specifically, the specific implementation of iteratively training the preset first to-be-trained model based on the first call text data and the first speech tag to obtain the first speech analysis model may be:

[0100] Based on the first call text data, a preset large model is pre-trained to obtain a content analysis model, and the content analysis model has the ability to understand the call content in the call text data; based on the first call text data, the first speech label, and the preset first prompt, the content analysis model is fine-tuned to obtain the first speech analysis model, wherein the preset first prompt is used to prompt the content analysis model to output speech features according to the call content in the call text data.

[0101] In order to fully fit the intention of outputting speech features in this embodiment, the preset large model can be pre-trained based on the first call text data to obtain a content analysis model that has the ability to understand the call content in the call text data; then based on the first call text data, the first speech label, and the preset first prompt (prompt), the content analysis model is fine-tuned to obtain a first speech analysis model, which can accurately output speech features according to the call content in the call text data.

[0102] Specifically, this approach utilizes the powerful learning and expression capabilities of the preset large model to improve the accuracy of the speech features output by the first speech analysis model.

[0103] It should be noted that the amount of first call text data used to train the preset large model is huge, for example, the order can be more than one million; the amount of first call text data used to fine-tune the content analysis model is relatively small, for example, it can be 1,000 copies of first call text data or 10,000 copies of first call text data, etc.

[0104] Among them, pre-training enables the preset large model to learn general, cross-domain feature representations about call content without labels.

[0105] It can be understood that after pre-training the preset large model based on the first call text data, a content analysis model capable of understanding the call content in the call text data can be obtained.

[0106] Furthermore, fine-tuning can be a process in which, based on pre-training, for specific tasks (for example, the task of outputting speech features) the content analysis model is further trained on training samples with sample labels to adapt the content analysis model to the specific task requirements. In this process, some or all parameters of the content analysis model will be adjusted to achieve the purpose of fine-tuning, so that the first speech analysis model obtained after fine-tuning can accurately output speech features based on the call content in the call text data.

[0107] Specifically, the fine-tuning task can be described in the preset first prompt as: outputting speech features according to the call content in the call text data, thereby prompting the content analysis model to output speech features according to the call content in the call text data; inputting the first call text data, the first speech label, and the preset first prompt into the content analysis model, performing content analysis processing based on the content analysis model, outputting speech features, adjusting the model parameters of the content analysis model based on the error between the speech features and the first speech label, and returning to the step of inputting the first call text data, the first speech label, and the preset first prompt into the content analysis model, until a first speech analysis model that meets the preset requirements is obtained.

[0108] It should be noted that, based on the first speech analysis model, the historical call text data is subjected to speech analysis processing to extract the target speech features of the marketing customers.

[0109] Furthermore, the specific implementation of iteratively training the preset first to-be-trained model based on the first call text data and the first speech tag to obtain the first speech analysis model may also be:

[0110] Based on the first call text data, the first speech tag, and a preset second prompt, the general big model is fine-tuned to obtain a first speech analysis model, wherein the preset second prompt is used to prompt the general big model to output speech features according to the call content in the call text data.

[0111] It should be noted that the preset first model to be trained can also be a general big model, and the general big model has text analysis capabilities; that is, the general big model can understand the call content in the first call text data; it only needs to be fine-tuned on the basis of the general big model based on the first call text data, the first speech label, and the preset second prompt, to obtain the first speech analysis model that can output speech features according to the call content in the call text data.

[0112] Specifically, the fine-tuning task can be described in the preset second prompt as: outputting speech features according to the call content in the call text data, thereby prompting the general large model to output speech features according to the call content in the call text data; inputting the first call text data, the first speech label, and the preset first prompt into the general large model, performing content analysis processing based on the general large model, outputting speech features, adjusting the model parameters of the general large model based on the error between the speech features and the first speech label, and returning to the step of inputting the first call text data, the first speech label, and the preset first prompt into the general large model until a first speech analysis model that meets the preset requirements is obtained.

[0113] In this embodiment, based on the first call text data, the preset big model is pre-trained to obtain a content analysis model, and based on the first call text data, the first speech label, and the preset first prompt, the content analysis model is fine-tuned; or, based on the first call text data, the first speech label, and the preset second prompt, the general big model is fine-tuned; both can obtain a first speech analysis model that can output speech features according to the call content in the call text data; due to the characteristics of many parameters of the preset big model and the general big model, this embodiment adopts the above method to improve the accuracy of the target speech features of marketing customers extracted based on the first speech analysis model.

[0114] Based on the first and second embodiments of the present application, in the third embodiment of the present application, the same or similar contents as those in the above embodiments can be referred to the above introduction, and will not be described in detail later. Figure 3 After step S02, the method for predicting customer purchase intention further includes steps S021 to S025:

[0115] Step S021, obtaining a test sample and a sample label of the test sample, wherein the test sample includes the second call text data of the historical customer, and the sample label of the test sample includes the second speech label of the second call text data;

[0116] In order to test the performance of the first speech analysis model mentioned above, the second call text data of historical customers can be used as a test sample, and a second speech label can be marked for the test sample. The second speech label is used to characterize the speech used by marketing personnel and / or historical customers in the second call text data.

[0117] Step S022, testing the first speech analysis model based on the second call text data to obtain a test result;

[0118] Specifically, the second call text data can be input into the first speech analysis model, and the second call text data can be subjected to speech analysis by the first speech analysis model to obtain the test result. It can be understood that the test result includes the speech features of the test.

[0119] Step S023, determining whether the test result is consistent with the second speech tag;

[0120] Specifically, it can be determined whether the test result is consistent with the speech represented by the second speech label. If consistent, it can reflect to a certain extent that the second call text data has a high accuracy in determining the speech features, that is, the model performance is high.

[0121] Among them, the method of determining whether the test result is consistent with the second speech tag can be by determining whether the keywords are consistent, or determining whether the meanings of the two are the same or similar, etc.

[0122] Step S024, taking the test result inconsistent with the second speech tag as a rejection tag, and taking the test result consistent with the second speech tag as an acceptance tag;

[0123] If the test result is inconsistent with the second speech label, the test result inconsistent with the second speech label can be used as a rejection label, that is, the speech label that the model does not want to output; the test result consistent with the second speech label can be used as an acceptance label, that is, the speech label that the model wants to output; so as to facilitate the subsequent optimization of the first speech analysis model based on the rejection label and the acceptance label.

[0124] Step S025, based on the rejection label and the acceptance label, the first speech analysis model is optimized to obtain an optimized second speech analysis model.

[0125] It should be noted that the preset speech analysis model includes a second speech analysis model; since the rejection label is used to characterize the speech label that the model does not want to output, and the acceptance label is used to characterize the speech label that the model wants to output, the first speech analysis model is optimized based on the rejection label and the acceptance label to obtain an optimized and more accurate second speech analysis model.

[0126] Specifically, the specific implementation of optimizing the first speech analysis model based on the rejection label and the acceptance label to obtain the optimized second speech analysis model may be:

[0127] Based on the preset call text data and the preset first prompt, the rejection label and the acceptance label are scored respectively by the frozen first speech analysis model to obtain a first scoring result, and based on the preset call text data and the preset first prompt, the rejection label and the acceptance label are scored respectively by the third speech analysis model to obtain a second scoring result;

[0128] The third speech analysis model that has not been iteratively trained is the first speech analysis model, the first scoring result is used to characterize the recognition of the rejection label and the acceptance label by the frozen first speech analysis model, and the second scoring result is used to characterize the recognition of the rejection label and the acceptance label by the third speech analysis model;

[0129] Based on the first scoring result and the second scoring result, a loss value is calculated by a preset loss function; based on the loss value, the model parameters of the third speech analysis model are updated, and the step of scoring the rejection label and the acceptance label respectively based on the preset call text data and the preset first prompt is returned accordingly by the frozen first speech analysis model, and iterative training is continued until a second speech analysis model that meets the preset first accuracy condition is obtained.

[0130] It should be noted that although the first speech analysis model can perform the task well, there may still be some decisions that are not accurate enough or do not meet expectations; therefore, the DPO (Direct Preference Optimization) in RLHF (Reinforcement Learning from Human Feedback) can be used to further optimize the first speech analysis model, that is, to perform preference alignment so that the first speech analysis model can be more in line with expectations, improve the accuracy of the first speech analysis model, and make more reasonable and practical decisions in uncertain or complex situations.

[0131] Specifically, the DPO algorithm can directly obtain preference information from the labeled information, thereby adjusting the model to better conform to actual preferences; it can be understood that a rejection label is a label that does not reflect the desired result, that is, it is less preferred to generate such a rejection label; an acceptance label is a label that reflects the desired result, that is, it is more preferred to generate such an acceptance label.

[0132] Since preferences are constantly changing and vary in different environments, the use of the DPO algorithm allows the model to dynamically adapt to these changing preferences and maintain the relevance and effectiveness between the model and the preferences.

[0133] Specifically, Figure 4 As shown, based on the preset call text data and the preset first prompt, the rejection label (rejected) and the acceptance label (chosen) are scored respectively by the frozen first speech analysis model (Frozen LM) to obtain a first scoring result; it can be understood that the first scoring result includes a first scoring result (rejected score) corresponding to the rejection label and a first scoring result (chosen score) corresponding to the acceptance label.

[0134] It should be noted that the first scoring result is used to characterize the recognition of the frozen first speech analysis model for the rejection label and the acceptance label respectively. The higher the first scoring result, the higher the recognition of the frozen first speech analysis model for the rejection label or the acceptance label, and vice versa.

[0135] Specifically, Figure 4 As shown, based on the preset call text data and the preset first prompt, the rejection label and the acceptance label are scored respectively through the third speech analysis model (Trained LM) to obtain a second scoring result; it can be understood that the second scoring result includes a second scoring result corresponding to the rejection label and a second scoring result corresponding to the acceptance label.

[0136] It should be noted that the second scoring result is used to characterize the recognition of the third speech analysis model for the rejection label and the acceptance label respectively. The higher the second scoring result is, the higher the recognition of the third speech analysis model for the rejection label or the acceptance label is, and vice versa.

[0137] Among them, if the third speech analysis model has not undergone iterative training, then the third speech analysis model that has not undergone iterative training is the first speech analysis model; if it has undergone iterative training, then the third speech analysis model is the trained first speech analysis model.

[0138] Furthermore, if Figure 4 As shown, based on the first scoring result and the second scoring result, the implementation method of calculating the loss value by using a preset loss function may be:

[0139] The first scoring result corresponding to the accepted label is divided by the first scoring result corresponding to the rejected label to calculate the reference model response ratio R reference , divide the second scoring result corresponding to the accepted label by the second scoring result corresponding to the rejected label to calculate the actual model response ratio R policy ; Based on the reference model response ratio and the model response ratio, calculate the loss value loss.

[0140] in,

[0141]

[0142] Wherein, both σ and β are known parameters.

[0143] It can be understood that the higher the loss value, the worse the model performance is, and the lower the loss value, the better the model performance is. Therefore, based on the loss value and the preset loss value threshold, it is determined whether to update the model parameters of the third speech analysis model. If the loss value is higher than the preset loss value threshold, the model parameters of the third speech analysis model are updated, and the step of scoring the rejection label and the acceptance label respectively based on the preset call text data and the preset first prompt by the frozen first speech analysis model is returned accordingly, and iterative training is continued until a second speech analysis model that meets the preset first precision condition is obtained; if the loss value is not higher than the preset loss value threshold, the second speech analysis model that meets the preset first precision condition is directly obtained; the second speech analysis model can achieve preference consistency.

[0144] In this embodiment, by testing the first speech analysis model, rejection labels and acceptance labels are obtained, and based on the rejection labels and the acceptance labels, the first speech analysis model is optimized by the DPO algorithm to obtain a second speech analysis model that achieves preference consistency; this can not only improve the accuracy of the first speech analysis model in the task of generating speech features, but also avoid the first speech analysis model from producing outputs that are inconsistent with reality. The output speech features can meet expectations and preferences, thereby improving the output accuracy of speech features.

[0145] Based on the first, second and third embodiments of the present application, in the fourth embodiment of the present application, the same or similar contents as those in the above embodiments can be referred to the above description and will not be described in detail later. Figure 5 Before step S10, the method for predicting customer purchase intention further includes steps S001 to S004:

[0146] Step S001, obtaining historical customer purchase intention data and call text data;

[0147] It should be noted that the purchase intention data of historical customers may be data related to customer purchase intention collected from multiple channels, and the call text data of historical customers may be converted based on the call data of historical customers.

[0148] Step S002, based on the purchase intention data of the historical customers, determining the historical purchase intention features of the historical customers, and learning and extracting the historical speech features from the call text data of the historical customers through a preset speech analysis model;

[0149] This embodiment obtains historical purchase intention features that can reflect the historical purchase patterns of historical customers based on the purchase intention data of historical customers, and learns and extracts historical speech features that can reflect the potential needs of historical customers from the call text data of historical customers through a preset speech analysis model, which not only considers the historical purchase patterns of historical customers, but also the potential needs of historical customers.

[0150] Step S003, determining a purchase intention label based on the historical purchase intention feature and the historical speech feature, wherein the purchase intention label includes the purchase probability of the historical customer;

[0151] In order to inform the model of the content that should be output based on the historical purchase intention features and the historical speech features, purchase intention labels corresponding to the historical purchase intention features and the historical speech features can be marked. Specifically, the purchase intention labels include the purchase probability of the historical customers.

[0152] The specific implementation method of determining the purchase intention label based on the historical purchase intention characteristics and the historical speech characteristics may be: based on the historical purchase intention characteristics and the historical speech characteristics, a logistic regression algorithm or a decision tree algorithm is used to predict the purchase probability of the historical customer, the purchase probability of the historical customer is between 0 and 1, and the purchase probability of the historical customer is used as the purchase intention label, such as Figure 6 shown.

[0153] Step S004, based on the historical purchase intention features, the historical speech features and the purchase intention labels, iteratively train the preset second model to be trained to obtain a willingness prediction model.

[0154] Based on the historical purchase intention characteristics, the historical speech characteristics and the purchase intention labels, the preset second model to be trained is iteratively trained to obtain a willingness prediction model that can characterize the correlation between the purchase intention characteristics, the speech characteristics and the purchase intention labels.

[0155] Specifically, before iteratively training the preset second model to be trained based on the historical purchase intention features, the historical speech features and the purchase intention labels, the historical purchase intention features and the historical speech features may also be format converted.

[0156] The preset second model to be trained may be a binary classification model to be trained or a multi-classification model to be trained, and the specific implementation method of iteratively training the preset second model to be trained based on the historical purchase intention features, the historical speech features and the purchase intention labels to obtain the intention prediction model may be:

[0157] The historical purchase intention feature, the historical speech feature and the purchase intention label are input into the binary classification model to be trained, and the purchase intention of the historical customer is predicted by the binary classification model to be trained to obtain the historical purchase probability of the historical customer; the error between the historical purchase probability and the purchase intention label is calculated to obtain the error calculation result; based on the error calculation result, the model parameters of the binary classification model to be trained are updated, and the step of inputting the historical purchase intention feature, the historical speech feature and the purchase intention label into the binary classification model to be trained is returned accordingly, until a willingness prediction model that meets the preset second accuracy condition is obtained.

[0158] Specifically, the binary classification model to be trained processes the above-mentioned historical purchase intention features and the historical speech features in the following manner: analyze the corresponding relationship between the historical purchase intention features and the historical speech features and the purchase intention, and obtain the predicted purchase intention of the historical customers, that is, obtain the historical purchase probability of the historical customers.

[0159] Furthermore, the above-mentioned iterative training process also includes: calculating the error between the historical purchase probability and the purchase intention label to obtain an error calculation result; judging whether the error calculation result meets the error standard indicated by a preset error threshold range; wherein the preset error threshold includes a preset mean square error threshold, and the smaller the mean square error threshold is, the more accurate the model is considered to be, therefore, judging whether the mean square error calculation result is less than the preset mean square error threshold, and if so, it is considered that the error calculation result meets the error standard indicated by the preset error threshold range.

[0160] Furthermore, the above-mentioned iterative training process also includes: if the above-mentioned error calculation result does not meet the error standard indicated by the preset error threshold range, then the model parameters of the binary classification model to be trained are updated, and the step of inputting the historical purchase intention feature, the historical speech feature and the purchase intention label into the binary classification model to be trained is returned, and the training is stopped after the error calculation result meets the error standard indicated by the preset error threshold range, so as to obtain the intention prediction model that meets the preset second accuracy condition.

[0161] Based on the above, through iterative training, the binary classification model to be trained is converged until the error calculation result meets the error standard indicated by the preset error threshold range, and the iterative training is completed to obtain the intention prediction model that meets the preset second precision condition, thereby improving the accuracy of the intention prediction model in predicting customer purchase intention.

[0162] This application also provides a device for predicting customer purchase intentions. Figure 7 , the customer purchase intention prediction device comprises:

[0163] An acquisition module 10 is used to acquire the historical purchase intention data and historical call text data of marketing customers;

[0164] An extraction module 20 is used to determine the target purchase intention characteristics of the marketing customer based on the historical purchase intention data, and learn to extract the target speech characteristics from the historical call text data through a preset speech analysis model, wherein the preset speech analysis model is used to characterize the correlation between the call text data and the speech characteristics, the target purchase intention characteristics reflect the historical purchase rules of the marketing customer, and the target speech characteristics reflect the potential needs of the marketing customer;

[0165] The prediction module 30 is used to predict the purchasing intention of the marketing customer based on the target purchasing intention characteristics and the target speech characteristics through a preset intention prediction model to obtain the target purchasing intention, wherein the preset intention prediction model is used to characterize the correlation between the purchasing intention characteristics, the speech characteristics and the purchasing intention.

[0166] The device for predicting customer purchase intention provided by the present application adopts the method for predicting customer purchase intention in the above-mentioned embodiment, and can solve the technical problem of predicting customer purchase intention. Compared with the prior art, the beneficial effect of the device for predicting customer purchase intention provided by the present application is the same as the beneficial effect of the method for predicting customer purchase intention provided by the above-mentioned embodiment, and other technical features in the device for predicting customer purchase intention are the same as the features disclosed in the above-mentioned embodiment method, which will not be described in detail here.

[0167] The present application provides a device for predicting customer purchase intention, and the device for predicting customer purchase intention includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method for predicting customer purchase intention in the above-mentioned embodiment one.

[0168] Reference below Figure 8 , which shows a schematic diagram of the structure of a device for predicting customer purchase intentions suitable for implementing the embodiment of the present application. The device for predicting customer purchase intentions in the embodiment of the present application may include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 8The customer purchase intention prediction device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0169] like Figure 8 As shown, the prediction device of customer purchase intention may include a processing device 1001 (such as a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. In RAM1004, various programs and data required for the operation of the prediction device of customer purchase intention are also stored. The processing device 1001, ROM1002 and RAM1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems may be connected to the I / O interface 1006: input devices 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 1003 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 1009. The communication device 1009 may allow the prediction device of the customer's purchase intention to communicate wirelessly or wired with other devices to exchange data. Although the figure shows the prediction device of the customer's purchase intention with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems may be implemented or have instead.

[0170] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0171] The device for predicting customer purchase intention provided by the present application adopts the method for predicting customer purchase intention in the above embodiment, and can solve the technical problem of predicting customer purchase intention. Compared with the prior art, the beneficial effect of the device for predicting customer purchase intention provided by the present application is the same as the beneficial effect of the method for predicting customer purchase intention provided by the above embodiment, and the other technical features in the device for predicting customer purchase intention are the same as the features disclosed in the method of the previous embodiment, and will not be repeated here.

[0172] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0173] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0174] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, wherein the computer-readable program instructions are used to execute the method for predicting customer purchase intention in the above-mentioned embodiment.

[0175] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0176] The computer-readable storage medium may be included in the device for predicting customer purchase intention; or may exist independently without being incorporated into the device for predicting customer purchase intention.

[0177] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the device for predicting customer purchase intention, the device for predicting customer purchase intention executes the method for predicting customer purchase intention.

[0178] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the client computer, partially on the client computer, as a separate software package, partially on the client computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the client computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0179] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0180] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0181] The readable storage medium provided by the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned method for predicting customer purchase intention, and can solve the technical problem of predicting customer purchase intention. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided by the present application are the same as the beneficial effects of the method for predicting customer purchase intention provided by the above-mentioned embodiment, and will not be repeated here.

[0182] The present application also provides a computer program product, including a computer program, which implements the steps of the above-mentioned method for predicting customer purchase intention when executed by a processor.

[0183] The computer program product provided by this application can solve the technical problem of predicting customer purchase intention. Compared with the prior art, the beneficial effects of the computer program product provided by this application are the same as the beneficial effects of the method for predicting customer purchase intention provided by the above embodiment, which will not be repeated here.

[0184] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A method for predicting customer purchase intention, characterized in that: The method includes: Obtain marketing customers’ historical purchase intention data and historical call text data; Based on the historical purchase intention data, determine the target purchase intention characteristics of the marketing customer, and learn and extract the target speech characteristics from the historical call text data through a preset speech analysis model, wherein the preset speech analysis model is used to characterize the correlation between the call text data and the speech characteristics, the target purchase intention characteristics reflect the historical purchase rules of the marketing customer, and the target speech characteristics reflect the potential needs of the marketing customer; Based on the target purchase intention characteristics and the target speech characteristics, the purchase intention of the marketing customer is predicted through a preset intention prediction model to obtain the target purchase intention, wherein the preset intention prediction model is used to characterize the correlation between the purchase intention characteristics, speech characteristics and purchase intention.

2. The method according to claim 1, characterized in that The preset speech analysis model includes a first speech analysis model. Before the step of obtaining the historical purchase intention data and historical call text data of the marketing customers, the method includes: Acquire a training sample and a sample label of the training sample, wherein the training sample includes first call text data of a historical customer, and the sample label of the training sample includes a first speech label of the first call text data; Based on the first call text data and the first speech tags, the preset first model to be trained is iteratively trained to obtain a first speech analysis model.

3. The method according to claim 2, characterized in that The preset first model to be trained includes a preset large model, and the step of iteratively training the preset first model to be trained based on the first call text data and the first speech tag to obtain a first speech analysis model includes: Based on the first call text data, pre-training a preset large model to obtain a content analysis model, wherein the content analysis model has the ability to understand the call content in the call text data; Based on the first call text data, the first speech label, and a preset first prompt, the content analysis model is fine-tuned to obtain a first speech analysis model, wherein the preset first prompt is used to prompt the content analysis model to output speech features according to the call content in the call text data.

4. The method according to any one of claims 2 to 3, characterized in that The preset speech analysis model also includes a second speech analysis model. After the step of iteratively training the preset first model to be trained based on the first call text data and the first speech label to obtain the first speech analysis model, the method includes: Acquire a test sample and a sample label of the test sample, wherein the test sample includes second call text data of a historical customer, and the sample label of the test sample includes a second speech label of the second call text data; Based on the second call text data, the first speech analysis model is tested to obtain a test result; Determining whether the test result is consistent with the second speech tag; Use the test results that are inconsistent with the second speech label as rejection labels, and use the test results that are consistent with the second speech label as acceptance labels; Based on the rejection labels and the acceptance labels, the first speech analysis model is optimized to obtain an optimized second speech analysis model.

5. The method according to claim 4, characterized in that The step of optimizing the first speech analysis model based on the rejection label and the acceptance label to obtain an optimized second speech analysis model includes: Based on the preset call text data and the preset first prompt, the rejection label and the acceptance label are scored respectively by the frozen first speech analysis model to obtain a first scoring result, and based on the preset call text data and the preset first prompt, the rejection label and the acceptance label are scored respectively by the third speech analysis model to obtain a second scoring result; The third speech analysis model that has not been iteratively trained is the first speech analysis model, the first scoring result is used to characterize the recognition of the rejection label and the acceptance label by the frozen first speech analysis model, and the second scoring result is used to characterize the recognition of the rejection label and the acceptance label by the third speech analysis model; Based on the first scoring result and the second scoring result, calculating a loss value by using a preset loss function; Based on the loss value, the model parameters of the third speech analysis model are updated, and the step of scoring the rejection label and the acceptance label respectively based on the preset call text data and the preset first prompt is returned accordingly through the frozen first speech analysis model, and iterative training is continued until the second speech analysis model that meets the preset first accuracy condition is obtained.

6. The method according to claim 2, characterized in that The preset first model to be trained also includes a general large model, and the general large model has a text analysis capability. The step of iteratively training the preset first model to be trained based on the first call text data and the first speech tag to obtain a first speech analysis model includes: Based on the first call text data, the first speech tag, and a preset second prompt, the general big model is fine-tuned to obtain a first speech analysis model, wherein the preset second prompt is used to prompt the general big model to output speech features according to the call content in the call text data.

7. The method according to claim 1, characterized in that Before the step of obtaining the marketing customer's historical purchase intention data and historical call text data, the method includes: Obtain historical customer purchase intention data and call text data; Based on the purchase intention data of the historical customers, determine the historical purchase intention features of the historical customers, and learn and extract the historical speech features from the call text data of the historical customers through a preset speech analysis model; Determine a purchase intention label based on the historical purchase intention feature and the historical speech feature, wherein the purchase intention label includes the purchase probability of the historical customer; Based on the historical purchase intention features, the historical speech features and the purchase intention labels, the preset second model to be trained is iteratively trained to obtain a willingness prediction model.

8. The method according to claim 7, characterized in that The preset second model to be trained includes a binary classification model to be trained, and the step of iteratively training the preset second model to be trained based on the historical purchase intention feature, the historical speech feature and the purchase intention label to obtain the intention prediction model includes: Inputting the historical purchase intention feature, the historical speech feature and the purchase intention label into the binary classification model to be trained, predicting the purchase intention of the historical customer through the binary classification model to be trained, and obtaining the historical purchase probability of the historical customer; Calculating the error between the historical purchase probability and the purchase intention label to obtain an error calculation result; Based on the error calculation result, the model parameters of the binary classification model to be trained are updated, and the step of inputting the historical purchase intention features, the historical speech features and the purchase intention labels into the binary classification model to be trained is returned accordingly, until a willingness prediction model that meets the preset second precision condition is obtained.

9. The method according to claim 1, characterized in that The step of predicting the marketing customer's purchase intention based on the target purchase intention feature and the target speech feature by using a preset intention prediction model to obtain the target purchase intention includes: Based on the target purchase intention characteristics and the target speech characteristics, the purchase intention of the marketing customer is predicted by a preset intention prediction model to obtain a target purchase probability, and the target purchase probability is positively correlated with the purchase intention of the marketing customer; After the step of predicting the marketing customer's purchase intention based on the target purchase intention feature and the target speech feature by using a preset intention prediction model to obtain the target purchase intention, the step includes: The marketing customers are sorted from large to small according to the corresponding target purchase probability, and the marketing customers ranked in the front preset positions are determined as the customers to be marketed, so that the marketing personnel can market to the customers to be marketed.

10. A device for predicting customer purchase intention, characterized in that: The device comprises: The acquisition module is used to obtain the historical purchase intention data and historical call text data of marketing customers; An extraction module, used to determine the target purchase intention characteristics of the marketing customer based on the historical purchase intention data, and learn to extract the target speech characteristics from the historical call text data through a preset speech analysis model, wherein the preset speech analysis model is used to characterize the correlation between the call text data and the speech characteristics, the target purchase intention characteristics reflect the historical purchase rules of the marketing customer, and the target speech characteristics reflect the potential needs of the marketing customer; A prediction module is used to predict the purchasing intention of the marketing customer based on the target purchasing intention characteristics and the target speech characteristics through a preset intention prediction model to obtain the target purchasing intention, wherein the preset intention prediction model is used to characterize the correlation between the purchasing intention characteristics, the speech characteristics and the purchasing intention.

11. A device for predicting customer purchase intention, characterized in that: The device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the method for predicting customer purchase intention according to any one of claims 1 to 9.

12. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the method for predicting customer purchase intention according to any one of claims 1 to 9 are implemented.

13. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the method for predicting customer purchase intention according to any one of claims 1 to 9 are implemented.

Citation Information

Cited By

  • Intelligent voice recognition method and system, computer equipment and storage medium

    CN120015027A

  • Method and system for dynamically grading transaction prediction probability based on user behavior data

    CN121504532A