Training method and device of satisfaction prediction model, equipment and storage medium

Abstract

The embodiment of the invention provides a satisfaction prediction model training method and device, a user satisfaction prediction method and device, equipment and a storage medium, and belongs to the technical field of financial science and technology and artificial intelligence. The satisfaction degree prediction model training method comprises the following steps: obtaining feedback data of a user for a claim settlement service; constructing at least two training sample sets according to the feedback data; training each base classifier of a satisfaction prediction model based on Stacking integration by using each training sample set, and obtaining a predicted satisfaction attribution probability output by each base classifier; and training a meta-classifier of the satisfaction prediction model based on Stacking integration by utilizing the prediction satisfaction attribution probability to obtain a trained satisfaction prediction model based on Stacking integration. According to the embodiment of the invention, the accuracy, efficiency and convenience of user satisfaction prediction can be improved.

Description

Technical Field

[0001] The present application relates to the fields of financial technology and artificial intelligence technology, and in particular to a satisfaction prediction model training and user satisfaction prediction method, device, equipment and storage medium. Background Art

[0002] In the field of financial technology, such as financial insurance business scenarios, paying attention to and improving user satisfaction with claims services is one of the important means to improve the quality of financial services.

[0003] At present, users' satisfaction with claims services is usually predicted by collecting follow-up questionnaires from users and scoring their evaluation of each claim process in the follow-up questionnaires. This relies on manual reading and judgment and summary, which is time-consuming and labor-intensive, and is limited by the subjectivity of manual judgment and summary, which may deviate from the user's actual satisfaction.

[0004] Therefore, how to improve the accuracy, efficiency and convenience of user satisfaction prediction has become a technical problem that needs to be solved urgently. Summary of the invention

[0005] The main purpose of the embodiments of the present application is to propose a satisfaction prediction model training and user satisfaction prediction method, device, equipment and storage medium, aiming to improve the accuracy, efficiency and convenience of user satisfaction prediction.

[0006] To achieve the above-mentioned purpose, a first aspect of an embodiment of the present application proposes a training method for a satisfaction prediction model, and the training method for a satisfaction prediction model comprises:

[0007] Obtain user feedback data on claims services;

[0008] Constructing at least two training sample sets according to the feedback data;

[0009] Using each of the training sample sets, each base classifier of the satisfaction prediction model based on Stacking integration is trained to obtain the predicted satisfaction attribution probability output by each of the base classifiers;

[0010] The predicted satisfaction attribution probability is used to train the meta-classifier of the satisfaction prediction model based on Stacking ensemble to obtain a trained satisfaction prediction model based on Stacking ensemble.

[0011] To achieve the above purpose, a second aspect of an embodiment of the present application proposes a user satisfaction prediction method, the user satisfaction prediction method comprising:

[0012] Obtain target feedback data from target users on claims services;

[0013] The target feedback data is predicted for satisfaction by using a trained satisfaction prediction model based on Stacking integration to obtain the target user's satisfaction, wherein the trained satisfaction prediction model based on Stacking integration is obtained by training using the above-mentioned satisfaction prediction model training method.

[0014] To achieve the above-mentioned purpose, a third aspect of an embodiment of the present application proposes a training device for a satisfaction prediction model, the device comprising:

[0015] Feedback data acquisition module, used to obtain user feedback data on claims services;

[0016] A training sample set construction module, used to construct at least two training sample sets according to the feedback data;

[0017] A base classifier training module, used to train each base classifier of the satisfaction prediction model based on Stacking integration by using each of the training sample sets, and obtain the predicted satisfaction attribution probability output by each of the base classifiers;

[0018] The meta-classifier training module is used to train the meta-classifier of the satisfaction prediction model based on Stacking integration by using the predicted satisfaction attribution probability to obtain a trained satisfaction prediction model based on Stacking integration.

[0019] To achieve the above-mentioned purpose, the fourth aspect of an embodiment of the present application proposes a computer device, which includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, it implements the training method of the satisfaction prediction model described in the first aspect or the user satisfaction prediction method described in the second aspect.

[0020] To achieve the above-mentioned purpose, the fifth aspect of an embodiment of the present application proposes a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the training method of the satisfaction prediction model described in the first aspect or the user satisfaction prediction method described in the second aspect.

[0021] The training method of the satisfaction prediction model proposed in the present application obtains user feedback data on the claims service; constructs at least two training sample sets based on the feedback data; uses each training sample set to train each base classifier of the satisfaction prediction model based on Stacking integration, and obtains the predicted satisfaction attribution probability output by each base classifier; uses the predicted satisfaction attribution probability to train the meta-classifier of the satisfaction prediction model based on Stacking integration, and obtains a trained satisfaction prediction model based on Stacking integration. In this way, based on multiple training sample sets, it can ensure that each base classifier can learn satisfaction prediction from different perspectives, and obtain the predicted satisfaction attribution probability of different perspectives output by different base classifiers. By using the predicted satisfaction attribution probability output by each base classifier as a new feature for the meta-classifier to learn the final satisfaction prediction, it can ensure that the meta-classifier can make full use of the outputs of multiple base classifiers, realize more effective integration of the advantages of different base classifiers, and effectively improve the overall performance of the trained satisfaction prediction model based on Stacking integration, so as to more accurately predict the user's satisfaction with the claims service, and improve the accuracy, efficiency and convenience of user satisfaction prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 is a flowchart of a method for training a satisfaction prediction model provided in an embodiment of the present application;

[0023] Figure 2 It is a schematic diagram of a training architecture for training a satisfaction prediction model based on Stacking integration provided in an embodiment of the present application;

[0024] Figure 3 is a flow chart of a user satisfaction prediction method provided by an embodiment of the present application;

[0025] Figure 4 is a schematic block diagram of a training device for a satisfaction prediction model provided in an embodiment of the present application;

[0026] Figure 5 It is a schematic block diagram of the structure of the computer device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0028] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0029] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0030] First, some nouns involved in this application are analyzed:

[0031] Artificial intelligence (AI) is a new technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence. AI is a branch of computer science that attempts to understand the essence of intelligence and produce a new type of intelligent machine that can respond in a similar way to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing and expert systems. AI can simulate the information process of human consciousness and thinking. AI is also a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0032] Stacking algorithm: It is an integrated learning method that enhances the prediction performance of the overall model by building a multi-layer model structure. Its success depends on the diversity of base models and the design of meta-models. It can learn how to best integrate the outputs of multiple base models by training a meta-model. The advantage of the Stacking algorithm is that it can make full use of the performance of different base models on different data features and improve the prediction performance of the overall model through the optimization of the meta-model.

[0033] Random Forest Classifier: is a powerful ensemble learning algorithm, widely used in classification and regression tasks. It is built on the basic model of "decision tree", but enhances the accuracy and robustness of the model by integrating multiple decision trees. The core idea of ​​random forest is to improve the generalization ability of the model and reduce overfitting by "introducing randomness" and "integrating multiple models".

[0034] XGBoost (Extreme Gradient Boosting) classifier: is an efficient gradient boosting framework, widely used in classification and regression tasks of machine learning. It performs well in processing large-scale data by introducing regularization and parallel computing, and is widely welcomed for its high accuracy and flexibility. XGBoost can be regarded as an optimized version of the gradient boosting algorithm, which solves various machine learning problems by improving the speed, performance and flexibility of the algorithm.

[0035] LightGBM (Light Gradient Boosting Machine) classifier: is an efficient gradient boosting framework designed to provide faster training speed, lower memory consumption and higher prediction accuracy. LightGBM is an ensemble learning method based on decision trees. Compared with other gradient boosting algorithms, it uses a unique tree growing algorithm that can process data and features more efficiently, which makes LightGBM particularly suitable for processing large-scale data sets and high-dimensional features.

[0036] Logistic Regression: It is a statistical model used for classification problems, mainly used to deal with binary and multi-classification problems. Logistic regression is based on the idea of ​​linear regression and realizes the prediction of categories by transforming the output.

[0037] Bidirectional Encoder Representations from Transformers (BERT) model: It is a deep learning language model. The BERT model can calculate the relationship between each word and other words in the input sequence based on the self-attention mechanism, thereby capturing contextual information and effectively identifying and extracting key semantic information in the text. This makes the BERT model a powerful tool for keyword extraction tasks in the field of natural language processing.

[0038] Based on this, the embodiments of the present application provide a satisfaction prediction model training and user satisfaction prediction method, device, equipment and storage medium, aiming to improve the accuracy, efficiency and convenience of user satisfaction prediction.

[0039] The training of the satisfaction prediction model and the user satisfaction prediction method, device, equipment and storage medium provided in the embodiments of the present application are specifically illustrated through the following embodiments. First, the training of the satisfaction prediction model and the user satisfaction prediction method in the embodiments of the present application are described.

[0040] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0041] The basic technologies of artificial intelligence generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. Artificial intelligence software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0042] The training of the satisfaction prediction model and the user satisfaction prediction method provided in the embodiments of the present application relate to the fields of financial technology and artificial intelligence technology. The training of the satisfaction prediction model and the user satisfaction prediction method provided in the embodiments of the present application can be applied to the terminal, can also be applied to the server side, and can also be software running in the terminal or the server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or can be configured as a server cluster or a distributed system composed of multiple physical servers, and can also be configured as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be the application of the training of the satisfaction prediction model and the user satisfaction prediction method, but is not limited to the above forms.

[0043] It should be noted that in each specific implementation of the present application, when it comes to the need to perform relevant processing based on data related to user identity or characteristics such as user information, user behavior data, user historical data, and user location information, the user's permission or consent will be obtained first, and the collection, use, and processing of these data will comply with relevant laws, regulations, and standards. In addition, when the embodiment of the present application needs to obtain the user's sensitive personal information, the user's separate permission or consent will be obtained through a pop-up window or by jumping to a confirmation page. After clearly obtaining the user's separate permission or consent, the necessary user-related data for the normal operation of the embodiment of the present application will be obtained.

[0044] Figure 1 is an optional flowchart of the training method of the satisfaction prediction model provided in the embodiment of the present application. Figure 1The method may include but is not limited to steps S101 to S104.

[0045] Step S101, obtaining user feedback data on claim settlement services.

[0046] Taking the application in financial insurance business scenario as an example, the training method of the satisfaction prediction model provided in the embodiment of the present application is described in detail.

[0047] In order to improve the efficiency and convenience of user satisfaction prediction, a satisfaction prediction model based on Stacking ensemble is constructed based on Stacking algorithm. The satisfaction prediction model based on Stacking ensemble includes multiple base classifiers and a meta classifier.

[0048] In some embodiments, in order to reduce computational complexity while ensuring prediction accuracy, a satisfaction prediction model based on Stacking integration may include three base classifiers and one meta classifier.

[0049] For example, see Figure 2 , Figure 2 A diagram of the training architecture for training a satisfaction prediction model based on Stacking ensemble. Figure 2 In the illustrated satisfaction prediction model based on Stacking integration, the base classifiers include random forest classifier, XGBoost classifier and LightGBM classifier; the role of the meta classifier is to integrate the outputs of all base classifiers, and the meta classifier can be a logistic regression model or a support vector machine.

[0050] In order to improve the training effect of the satisfaction prediction model based on Stacking integration, a large amount of user feedback data on claims services can be obtained from the database of the financial business system to more effectively train the satisfaction prediction model based on Stacking integration.

[0051] User feedback data on claims services can be user comments, evaluations, and reading data on insurance product claims services. Feedback data can include user ID, claim application ID, comment content, rating (such as 1-5 stars), comment time, and other related metadata (such as product type, region), etc.

[0052] Step S102: construct at least two training sample sets according to the feedback data.

[0053] Based on the feedback data, construct at least two training sample sets.

[0054] The number of training sample sets may be the same as the number of base classifiers of the satisfaction prediction model based on Stacking integration. For example, if the satisfaction prediction model based on Stacking integration has three base classifiers, three training sample sets may be constructed according to the feedback data.

[0055] Since different base classifiers perform well on different training sample sets, each base classifier can fully learn the characteristics of different training sample sets. The advantages of different base classifiers can be fully utilized through Stacking integration, thereby enhancing the diversity of the satisfaction prediction model based on Stacking integration. This diversity helps to improve the robustness of the satisfaction prediction model based on Stacking integration.

[0056] Step S103, using each training sample set, train each base classifier of the satisfaction prediction model based on Stacking integration, and obtain the predicted satisfaction attribution probability output by each base classifier.

[0057] Satisfaction prediction is regarded as a classification problem. Each base classifier of the satisfaction prediction model based on Stacking ensemble is trained with each training set to obtain the predicted satisfaction attribution probability output by each base classifier for the input training sample set, that is, the probability that each base classifier can predict which satisfaction category the input training sample set belongs to.

[0058] For example, please refer to Figure 2 For example, the satisfaction is divided into five categories: very satisfied, satisfied, average, dissatisfied and very dissatisfied. Assume that three training sample sets are constructed, and the three training sample sets are defined as the first training sample set, the second training sample set and the third training sample set respectively. The first training sample set is input into the random forest classifier for satisfaction prediction learning and training, and the first probability Y1 of which satisfaction category the first training sample set predicted by the random forest classifier belongs to is obtained; the second training sample set is input into the XGBoost classifier for satisfaction prediction learning and training, and the second probability Y2 of which satisfaction category the second training sample set predicted by the XGBoost classifier belongs to is obtained; the third training sample set is input into the LightGBM classifier for satisfaction prediction learning and training, and the third probability Y3 of which satisfaction category the third training sample set predicted by the LightGBM classifier belongs to is obtained.

[0059] Step S104, using the predicted satisfaction attribution probability, train the meta-classifier of the satisfaction prediction model based on Stacking ensemble to obtain a trained satisfaction prediction model based on Stacking ensemble.

[0060] By using the predicted satisfaction attribution probability, the meta-classifier of the satisfaction prediction model based on Stacking ensemble is optimized and trained to obtain the trained satisfaction prediction model based on Stacking ensemble.

[0061] For example, please refer to Figure 2 , the first probability of the first training sample set predicted by the random forest classifier belonging to a certain satisfaction category, the second probability of the second training sample set predicted by the XGBoost classifier belonging to a certain satisfaction category, and the third probability of the third training sample set predicted by the LightGBM classifier belonging to a certain satisfaction category can be combined to form a new training sample set, and the meta-classifier can be optimized for satisfaction prediction learning.

[0062] In this way, the meta-classifier is used to fully learn the outputs of different base classifiers. If a base classifier performs poorly on a specific sample, the predictions of other base classifiers can make up for this deficiency. Through integration, the meta-classifier can learn how to make the final satisfaction prediction based on the outputs of different base classifiers, enhance the overall model's tolerance to noise and outliers, and effectively improve the overall performance of the trained satisfaction prediction model based on Stacking integration.

[0063] The training method of the satisfaction prediction model provided in the above embodiment obtains user feedback data on the claims service; constructs at least two training sample sets based on the feedback data; uses each training sample set to train each base classifier of the satisfaction prediction model based on Stacking integration, and obtains the predicted satisfaction attribution probability output by each base classifier; uses the predicted satisfaction attribution probability to train the meta-classifier of the satisfaction prediction model based on Stacking integration, and obtains a trained satisfaction prediction model based on Stacking integration. In this way, based on multiple training sample sets, it can ensure that each base classifier can learn satisfaction prediction from different perspectives, and obtain the predicted satisfaction attribution probability of different perspectives output by different base classifiers. By using the predicted satisfaction attribution probability output by each base classifier as a new feature for the meta-classifier to learn the final satisfaction prediction, it can ensure that the meta-classifier can make full use of the outputs of multiple base classifiers, realize more effective integration of the advantages of different base classifiers, and effectively improve the overall performance of the trained satisfaction prediction model based on Stacking integration, so as to more accurately predict the user's satisfaction with the claims service, and improve the accuracy, efficiency and convenience of user satisfaction prediction.

[0064] In step S102 of some embodiments, the feedback data can be cleaned to obtain clean feedback data; the preset language model is used to perform topic extraction processing on the clean feedback data to obtain the business topics of the clean feedback data; according to the business topics, the clean feedback data is divided into three training sample sets.

[0065] To improve the quality of the training sample sets, the feedback data can be cleaned first. Exemplarily, for example, removing irrelevant stop words (such as "de", "le", etc.), special characters, and noise data, associating comments with relevant metadata (such as product type, region, etc.), and so on, so as to obtain clean feedback data.

[0066] Then, through the preset language model, topic extraction processing is performed on the clean feedback data to obtain the business topics of the clean feedback data. Exemplarily, the business topics are, for example, business systems (such as the source business system of the feedback data), regions (such as users in different provinces or cities), insurance product types (such as different types of insurance products like health insurance, auto insurance, property insurance, etc.), purchase methods (such as users who purchase through online or offline channels), purchase time of insurance products (such as the time period when users purchase insurance products), expiration period (such as the expiration period of the insurance products purchased by users), insurance premiums (such as feedback from users with different premium levels), and so on.

[0067] Exemplarily, when the satisfaction prediction model based on Stacking integration includes three base classifiers, according to the business topics, the clean feedback data can be divided into three different training sample sets, where each training sample set contains the same number of clean feedback data.

[0068] Exemplarily, a business topic (such as region) can be selected, and by means of random sampling, the same number of clean feedback data is randomly selected to form three different training sample sets. For example, assuming there are 1200 pieces of clean feedback data, and "region" is selected as the business topic for division. If the 1200 pieces of clean feedback data cover 3 regions, 400 pieces of clean feedback data can be randomly selected from each region, so that 3 equal training sample sets are obtained, and each training sample set contains 400 pieces of clean feedback data.

[0069] Through such division, it can be ensured that different base classifiers can respectively fully learn the features of different training sample sets, so as to give play to the advantages of different base classifiers in satisfaction prediction for the meta-classifier to integrate and learn.

[0070] In some embodiments, the preset language model includes a pre-trained BERT model, and the preset language model is used to perform topic extraction processing on the clean feedback data to obtain the business theme of the clean feedback data. The clean feedback data can be subjected to keyword extraction processing through the trained BERT model to obtain the keywords of the clean feedback data; the keywords are subjected to topic classification processing to obtain the themes of the keywords; and the themes of the keywords are used as the business themes.

[0071] In order to obtain the business themes of clean feedback data more efficiently and accurately, the pre-trained BERT model can be used to perform keyword extraction on the clean feedback data to obtain the keywords of the clean feedback data, and then the topic categories to which the keywords belong are analyzed to obtain the themes of the keywords, thereby using the themes of the keywords as business themes. For example, if the keywords extracted by the trained BERT model are "health insurance", "auto insurance", etc., the themes of the keywords can be classified as insurance product types.

[0072] In this way, since the pre-trained BERT model can capture the rich semantic information of clean feedback data, it can improve the accuracy of keyword extraction, thereby improving the accuracy of business topics, so that the clean feedback data can be divided into at least two training sample sets based on business topics with higher accuracy, thereby improving the rationality of the training sample sets and ensuring that the training process is more efficient and stable.

[0073] In step S104 of some embodiments, the predicted satisfaction attribution probabilities can be combined to obtain a new training sample set; the new training sample set is input into the meta-classifier for training to update the hyperparameters of the meta-classifier until the hyperparameters are less than a preset hyperparameter threshold, thereby obtaining a trained satisfaction prediction model based on Stacking integration.

[0074] For example, the predicted satisfaction attribution probability output by the random forest classifier is Y1, the predicted satisfaction attribution probability output by the XGBoost classifier is Y2, and the predicted satisfaction attribution probability output by the LightGBM classifier is Y3. Y1, Y2 and Y3 are combined to form a new training sample set [Y1, Y2, Y3], and then a meta-classifier (such as a logistic regression model) is trained on the new training sample set [Y1, Y2, Y3] to update the hyperparameters of the meta-classifier until the hyperparameters are less than the preset hyperparameter threshold, thereby obtaining a trained satisfaction prediction model based on Stacking integration.

[0075] In this way, by combining different base classifiers and using meta-classifiers for satisfaction prediction learning, Stacking ensemble can effectively improve the overall prediction accuracy of the trained satisfaction prediction model.

[0076] Based on the above examples, please refer to Figure 3 , Figure 3 is an optional flowchart of the user satisfaction prediction method provided in the embodiment of the present application. Figure 3 The method may include but is not limited to steps S105 to S106.

[0077] Step S105, obtaining target feedback data of target users on the claims service.

[0078] After the satisfaction prediction model based on Stacking integration is trained, the trained satisfaction prediction model based on Stacking integration can be applied to the satisfaction prediction task, and the target feedback data of the target users of the satisfaction prediction task on the claims service can be obtained.

[0079] Step S106, using the trained satisfaction prediction model based on Stacking integration to perform satisfaction prediction on the target feedback data to obtain the satisfaction of the target user, wherein the trained satisfaction prediction model based on Stacking integration is trained by a training method of a satisfaction prediction model.

[0080] Through the trained Stacking-based ensemble satisfaction prediction model, satisfaction prediction is performed on the target feedback data to obtain the target user's satisfaction.

[0081] In step S106 of some embodiments, the trained satisfaction prediction model based on Stacking integration includes a trained random forest classifier, a trained XGBoost classifier, a trained LightGBM classifier and a trained meta-classifier. The target feedback data can be input into the trained random forest classifier for satisfaction prediction to obtain a first satisfaction attribution probability predicted by the trained random forest classifier; the target feedback data can be input into the trained XGBoost classifier for satisfaction prediction to obtain a second satisfaction attribution probability predicted by the trained XGBoost classifier; the target feedback data can be input into the trained LightGBM classifier for satisfaction prediction to obtain a third satisfaction attribution probability predicted by the trained LightGBM classifier; the first satisfaction attribution probability, the second satisfaction attribution probability and the third satisfaction attribution probability can be input into the trained meta-classifier for satisfaction prediction to obtain the satisfaction of the target user.

[0082] Specifically, the target feedback data can be respectively input into the trained random forest classifier, the trained XGBoost classifier, the trained LightGBM classifier and the trained meta-classifier for satisfaction prediction, so as to obtain the first satisfaction attribution probability predicted by the trained random forest classifier, the second satisfaction attribution probability predicted by the trained XGBoost classifier and the third satisfaction attribution probability predicted by the trained LightGBM classifier; then the first satisfaction attribution probability, the second satisfaction attribution probability and the third satisfaction attribution probability are input into the trained meta-classifier for satisfaction prediction, and the satisfaction of the target user finally output by the meta-classifier is obtained.

[0083] In this way, by integrating the prediction results of the trained random forest classifier, the trained XGBoost classifier, the trained LightGBM classifier and the trained meta classifier, the accuracy and stability of the satisfaction prediction are effectively improved.

[0084] See also Figure 4 , Figure 4 It is a schematic block diagram of a training device for a satisfaction prediction model provided in an embodiment of the present application. The training device for the satisfaction prediction model can be configured in a computer device to execute the aforementioned training method for the satisfaction prediction model.

[0085] like Figure 4 As shown, the training device 400 of the satisfaction prediction model includes: a feedback data acquisition module 401, a training sample set construction module 402, a base classifier training module 403 and a meta classifier training module 404.

[0086] Feedback data acquisition module 401, used to acquire user feedback data on claim settlement services;

[0087] A training sample set construction module 402, configured to construct at least two training sample sets according to the feedback data;

[0088] The base classifier training module 403 is used to train each base classifier of the satisfaction prediction model based on Stacking integration by using each training sample set, and obtain the predicted satisfaction attribution probability output by each base classifier;

[0089] The meta-classifier training module 404 is used to train the meta-classifier of the satisfaction prediction model based on Stacking integration by using the predicted satisfaction attribution probability to obtain a trained satisfaction prediction model based on Stacking integration.

[0090] In one embodiment, the meta-classifier training module 404 is further configured to:

[0091] Combining the predicted satisfaction attribution probabilities to obtain a new training sample set;

[0092] The new training sample set is input into the meta-classifier for training to update the hyperparameters of the meta-classifier until the hyperparameters are less than a preset hyperparameter threshold, thereby obtaining a trained satisfaction prediction model based on Stacking integration.

[0093] In one embodiment, the base classifiers include a random forest classifier, an XGBoost classifier, and a LightGBM classifier.

[0094] In one embodiment, the training sample set construction module 402 is further used to:

[0095] Cleaning the feedback data to obtain clean feedback data;

[0096] Performing topic extraction processing on the clean feedback data by using a preset language model to obtain a business topic of the clean feedback data;

[0097] According to the business topic, the cleaning feedback data is divided into at least two training sample sets.

[0098] In one embodiment, the preset language model includes a pre-trained BERT model, and the training sample set construction module 402 is further used to:

[0099] Using the trained BERT model, performing keyword extraction processing on the clean feedback data to obtain keywords of the clean feedback data;

[0100] Performing subject classification processing on the keywords to obtain the subjects of the keywords;

[0101] The subject of the keyword is used as the business subject.

[0102] Among them, each module in the training device 400 of the satisfaction prediction model corresponds to each step in the above-mentioned user satisfaction prediction method embodiment, and its functions and implementation processes are no longer described one by one here.

[0103] The training device 400 of the satisfaction prediction model can execute the training method of the satisfaction prediction model provided in the embodiment of the present application. Therefore, it can achieve the beneficial effects that can be achieved by the training method of the satisfaction prediction model provided in the embodiment of the present application. Please see the previous embodiment for details and will not be repeated here.

[0104] An embodiment of the present application also provides a user satisfaction prediction device, which can be configured in a computer device to execute the aforementioned user satisfaction prediction method.

[0105] The user satisfaction prediction device comprises: a target feedback data acquisition module and a satisfaction prediction module.

[0106] A target feedback data acquisition module is used to acquire target feedback data of target users on claims services;

[0107] The satisfaction prediction module is used to perform satisfaction prediction on the target feedback data through a trained satisfaction prediction model based on Stacking integration to obtain the satisfaction of the target user, wherein the trained satisfaction prediction model based on Stacking integration is trained by the training method of the satisfaction prediction model as described above.

[0108] In one embodiment, the satisfaction prediction module is further used to:

[0109] Inputting the target feedback data into the trained random forest classifier to perform satisfaction prediction, and obtaining the first satisfaction attribution probability predicted by the trained random forest classifier;

[0110] Inputting the target feedback data into the trained XGBoost classifier to perform satisfaction prediction, and obtaining the second satisfaction attribution probability predicted by the trained XGBoost classifier;

[0111] Input the target feedback data into the trained LightGBM classifier for satisfaction prediction, and obtain the third satisfaction attribution probability predicted by the trained LightGBM classifier;

[0112] The first satisfaction attribution probability, the second satisfaction attribution probability and the third satisfaction attribution probability are input into the trained meta-classifier to perform satisfaction prediction to obtain the satisfaction of the target user.

[0113] Among them, each module in the above-mentioned user satisfaction prediction device corresponds to each step in the above-mentioned user satisfaction prediction method embodiment, and its functions and implementation processes are no longer repeated here one by one.

[0114] The user satisfaction prediction device can execute the user satisfaction prediction method provided in the embodiment of the present application. Therefore, it can achieve the beneficial effects that can be achieved by the user satisfaction prediction method provided in the embodiment of the present application. Please refer to the previous embodiment for details and will not be repeated here.

[0115] The method and apparatus of the present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0116] Exemplarily, the above method and apparatus may be implemented in the form of a computer program. The computer program may be implemented in the form of a computer program. Figure 5 Runs on the computer device shown.

[0117] See also Figure 5 , Figure 5 It is a schematic block diagram of the structure of a computer device provided in an embodiment of the present application.

[0118] See also Figure 5 The computer device includes a processor and a memory connected via a system bus, wherein the memory may include a non-volatile storage medium and an internal memory.

[0119] The processor is used to provide computing and control capabilities and support the operation of the entire computer equipment.

[0120] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When the computer program is executed by the processor, the processor can execute any training method of the satisfaction prediction model or the user satisfaction prediction method.

[0121] It should be understood that the processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0122] In one embodiment, the processor is used to run a computer program stored in the memory to implement the following steps:

[0123] Obtain user feedback data on claims services;

[0124] Constructing at least two training sample sets according to the feedback data;

[0125] Using each of the training sample sets, each base classifier of the satisfaction prediction model based on Stacking integration is trained to obtain the predicted satisfaction attribution probability output by each of the base classifiers;

[0126] The predicted satisfaction attribution probability is used to train the meta-classifier of the satisfaction prediction model based on Stacking ensemble to obtain a trained satisfaction prediction model based on Stacking ensemble.

[0127] In one embodiment, when the processor implements the step of training the meta-classifier of the satisfaction prediction model based on Stacking integration by using the predicted satisfaction attribution probability to obtain the trained satisfaction prediction model based on Stacking integration, it is used to implement:

[0128] Combining the predicted satisfaction attribution probabilities to obtain a new training sample set;

[0129] The new training sample set is input into the meta-classifier for training to update the hyperparameters of the meta-classifier until the hyperparameters are less than a preset hyperparameter threshold, thereby obtaining a trained satisfaction prediction model based on Stacking integration.

[0130] In one embodiment, the base classifiers include a random forest classifier, an XGBoost classifier, and a LightGBM classifier.

[0131] In one embodiment, when implementing the step of constructing at least two training sample sets according to the feedback data, the processor is configured to implement:

[0132] Cleaning the feedback data to obtain clean feedback data;

[0133] Performing topic extraction processing on the clean feedback data by using a preset language model to obtain a business topic of the clean feedback data;

[0134] According to the business topic, the cleaning feedback data is divided into at least two training sample sets.

[0135] In one embodiment, the preset language model includes a pre-trained BERT model, and the processor, when implementing the topic extraction processing of the clean feedback data by the preset language model to obtain the business topic of the clean feedback data, is used to implement:

[0136] Using the trained BERT model, performing keyword extraction processing on the clean feedback data to obtain keywords of the clean feedback data;

[0137] Performing subject classification processing on the keywords to obtain the subjects of the keywords;

[0138] The subject of the keyword is used as the business subject.

[0139] In one embodiment, the processor is used to run a computer program stored in the memory to implement the following steps:

[0140] Obtain target feedback data from target users on claims services;

[0141] The target feedback data is predicted for satisfaction by using a trained satisfaction prediction model based on Stacking integration to obtain the target user's satisfaction, wherein the trained satisfaction prediction model based on Stacking integration is obtained by training the satisfaction prediction model using the training method described above.

[0142] In one embodiment, when the processor implements the satisfaction prediction model based on Stacking integration trained to perform satisfaction prediction on the target feedback data to obtain the satisfaction of the target user, it is used to implement:

[0143] Inputting the target feedback data into the trained random forest classifier to perform satisfaction prediction, and obtaining the first satisfaction attribution probability predicted by the trained random forest classifier;

[0144] Inputting the target feedback data into the trained XGBoost classifier to perform satisfaction prediction, and obtaining the second satisfaction attribution probability predicted by the trained XGBoost classifier;

[0145] Input the target feedback data into the trained LightGBM classifier for satisfaction prediction, and obtain the third satisfaction attribution probability predicted by the trained LightGBM classifier;

[0146] The first satisfaction attribution probability, the second satisfaction attribution probability and the third satisfaction attribution probability are input into the trained meta-classifier to perform satisfaction prediction to obtain the satisfaction of the target user.

[0147] The computer device can execute the training method of the satisfaction prediction model or the user satisfaction prediction method provided in the embodiment of the present application. Therefore, it can achieve the beneficial effects that can be achieved by the training method of the satisfaction prediction model or the user satisfaction prediction method provided in the embodiment of the present application. Please see the previous embodiments for details and will not be repeated here.

[0148] An embodiment of the present application also provides a computer-readable storage medium.

[0149] A computer program is stored on the computer-readable storage medium of the present application, and when the computer program is executed by a processor, the steps of the training method of the satisfaction prediction model or the user satisfaction prediction method as described above are implemented.

[0150] The computer-readable storage medium may be an internal storage unit of the training device of the satisfaction prediction model, the user satisfaction prediction device, or the computer device described in the aforementioned embodiment, such as a hard disk or memory of the training device of the satisfaction prediction model, the user satisfaction prediction device, or the computer device. The computer-readable storage medium may also be an external storage device of the training device of the satisfaction prediction model, the user satisfaction prediction device, or the computer device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital card (Secure Digital Card, SDCard), a flash card (Flash Card), etc., equipped on the training device of the satisfaction prediction model, the user satisfaction prediction device, or the computer device.

[0151] Since the computer program stored in the computer-readable storage medium can execute any one of the satisfaction prediction model training methods or user satisfaction prediction methods provided in the embodiments of the present application, the beneficial effects that can be achieved by any one of the satisfaction prediction model training methods or user satisfaction prediction methods provided in the embodiments of the present application can be achieved. Please see the previous embodiments for details and will not be repeated here.

[0152] Furthermore, the computer-readable storage medium may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function, etc.; the data storage area may store data created according to the use of the blockchain node, etc.

[0153] The blockchain referred to in this application is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0154] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or system including the element.

[0155] The above description is only a specific implementation mode of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or substitutions within the technical scope disclosed in the present application, and these modifications or substitutions should be included in the protection scope of the present application.

Claims

1. A training method for a satisfaction prediction model, characterized in that: The training method of the satisfaction prediction model includes: Obtain user feedback data on claims services; Constructing at least two training sample sets according to the feedback data; Using each of the training sample sets, each base classifier of the satisfaction prediction model based on Stacking integration is trained to obtain the predicted satisfaction attribution probability output by each of the base classifiers; The predicted satisfaction attribution probability is used to train the meta-classifier of the satisfaction prediction model based on Stacking ensemble to obtain a trained satisfaction prediction model based on Stacking ensemble.

2. The training method of the satisfaction prediction model according to claim 1, characterized in that: The method of using the predicted satisfaction attribution probability to train the meta-classifier of the satisfaction prediction model based on Stacking integration to obtain a trained satisfaction prediction model based on Stacking integration includes: Combining the predicted satisfaction attribution probabilities to obtain a new training sample set; The new training sample set is input into the meta-classifier for training to update the hyperparameters of the meta-classifier until the hyperparameters are less than a preset hyperparameter threshold, thereby obtaining a trained satisfaction prediction model based on Stacking integration.

3. The training method of the satisfaction prediction model according to claim 1, characterized in that: The base classifiers include random forest classifier, XGBoost classifier and LightGBM classifier.

4. The training method of the satisfaction prediction model according to claim 1, characterized in that: The step of constructing at least two training sample sets according to the feedback data includes: Cleaning the feedback data to obtain clean feedback data; Performing topic extraction processing on the clean feedback data by using a preset language model to obtain a business topic of the clean feedback data; According to the business topic, the cleaning feedback data is divided into at least two training sample sets.

5. The training method of the satisfaction prediction model according to claim 4, characterized in that: The preset language model includes a pre-trained BERT model, and the topic extraction processing is performed on the clean feedback data by the preset language model to obtain the business topic of the clean feedback data, including: Using the trained BERT model, performing keyword extraction processing on the clean feedback data to obtain keywords of the clean feedback data; Performing subject classification processing on the keywords to obtain the subjects of the keywords; The subject of the keyword is used as the business subject.

6. A method for predicting user satisfaction, characterized in that: The user satisfaction prediction method comprises: Obtain target feedback data from target users on claims services; The target feedback data is predicted for satisfaction by using a trained satisfaction prediction model based on Stacking integration to obtain the target user's satisfaction, wherein the trained satisfaction prediction model based on Stacking integration is obtained by training using the training method for the satisfaction prediction model described in any one of claims 1 to 5.

7. The user satisfaction prediction method according to claim 6, characterized in that: The trained satisfaction prediction model based on Stacking integration includes a trained random forest classifier, a trained XGBoost classifier, a trained LightGBM classifier and a trained meta classifier. The trained satisfaction prediction model based on Stacking integration is used to perform satisfaction prediction on the target feedback data to obtain the satisfaction of the target user, including: Inputting the target feedback data into the trained random forest classifier to perform satisfaction prediction, and obtaining the first satisfaction attribution probability predicted by the trained random forest classifier; Inputting the target feedback data into the trained XGBoost classifier to perform satisfaction prediction, and obtaining the second satisfaction attribution probability predicted by the trained XGBoost classifier; Input the target feedback data into the trained LightGBM classifier for satisfaction prediction, and obtain the third satisfaction attribution probability predicted by the trained LightGBM classifier; The first satisfaction attribution probability, the second satisfaction attribution probability and the third satisfaction attribution probability are input into the trained meta-classifier to perform satisfaction prediction to obtain the satisfaction of the target user.

8. A training device for a satisfaction prediction model, characterized in that: The device comprises: Feedback data acquisition module, used to obtain user feedback data on claims services; A training sample set construction module, used to construct at least two training sample sets according to the feedback data; A base classifier training module, used to train each base classifier of the satisfaction prediction model based on Stacking integration by using each of the training sample sets, and obtain the predicted satisfaction attribution probability output by each of the base classifiers; The meta-classifier training module is used to train the meta-classifier of the satisfaction prediction model based on Stacking integration by using the predicted satisfaction attribution probability to obtain a trained satisfaction prediction model based on Stacking integration.

9. A computer device, characterized in that: The computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the training method of the satisfaction prediction model described in any one of claims 1 to 5 or the user satisfaction prediction method described in any one of claims 6 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, it implements the training method of the satisfaction prediction model described in any one of claims 1 to 5 or the user satisfaction prediction method described in any one of claims 6 to 7.

Images