Automobile financial service optimization method, system and equipment based on large language model

Abstract

The invention discloses an automobile financial service optimization method, system and device based on a large language model. The method comprises the steps of obtaining a first text input by a first user; inputting the first text into the large language model to obtain a first financial scheme; obtaining a second text input by the first user; obtaining a first evaluation for the first financial scheme in the second text; obtaining first transaction data of the first user, and clustering the first user according to the first transaction data and the first evaluation to obtain a clustering result; establishing a transaction prediction model according to a clustering result; obtaining dialogue data to be detected, and obtaining future transaction prediction data according to the dialogue data to be detected and the transaction prediction model; and optimizing the financial scheme according to the future transaction prediction data. The problems that an existing automobile financial service financial scheme is slow in feedback response to markets and users, slow in perception of user demand changes and insufficient in accuracy are at least solved.

Description

Technical Field

[0001] The present application relates to the field of financial management technology, and in particular to a method, system and device for optimizing automobile financial services based on a large language model. Background Art

[0002] With the rapid development of the global automobile industry, automobile financial services, as an emerging financial model, have gradually become an important part of the modern automobile consumer market. Automobile financial services mainly refer to the financial support services related to automobiles, such as purchase, leasing, and maintenance, provided by financial institutions to consumers. Its core purpose is to lower the threshold for consumers to buy cars, promote the increase of automobile sales, and meet the needs of different consumers through flexible financing methods. At present, the main forms of automobile financial services include: automobile loans, automobile leasing, financial leasing, insurance services, and car owner value-added services. Automobile loans are the most common financing method. Consumers apply for loans from banks or financial companies to buy cars and repay the principal and interest on schedule. Car leasing means that consumers use cars by paying rent. After the lease period ends, they can choose to buy, renew or return the car. Financial leasing is usually provided by automobile manufacturers or dealers in cooperation with financial institutions. Consumers pay rent during the lease period and can choose to buy the car after the expiration.

[0003] However, traditional auto financial services still have some shortcomings. For example, the traditional application process is cumbersome, the approval time is long, and the customer credit assessment is not accurate enough, which leads to poor experience for consumers when obtaining financial services. In addition, the credit status, car purchase needs and repayment ability of different consumers vary greatly. How to provide personalized and flexible financial solutions is also a major challenge facing current auto financial services.

[0004] However, the existing financial solutions for auto financial services are slow to respond to market and user feedback, and have problems such as slow perception of changes in user demand and lack of accuracy. There is an urgent need for a more intelligent and efficient financial service optimization solution so that financial service institutions can meet the diverse and personalized financial needs of modern consumers. Summary of the invention

[0005] The present invention provides an automobile financial service optimization method, system and device based on a large language model, which at least solves the problems that the existing automobile financial service financial solutions are slow to respond to market and user feedback, slow to perceive changes in user needs and insufficient accuracy, and provides a more intelligent and efficient financial service optimization solution to enable financial service institutions to meet the diversified and personalized financial needs of modern consumers.

[0006] An automobile financial service optimization method based on a large language model comprises the following steps:

[0007] Acquire a first text input by a first user, where the first text is configured as a text in which the user inquires about automobile financial solutions;

[0008] According to the first text, inputting the first text into a large language model to obtain a first financial solution;

[0009] Acquire a second text input by the first user, where the second text is configured as a text sent by the user in response to the first financial plan;

[0010] parsing the second text to obtain a first evaluation of the first financial plan in the second text;

[0011] Acquire first transaction data of the first user, and cluster the first users according to the first transaction data and the first evaluation to obtain a clustering result;

[0012] According to the clustering result, a transaction prediction model is established, wherein the transaction prediction model is configured to be a prediction model that takes evaluation data as input and takes transaction status as output, wherein the evaluation data includes the evaluated financial scheme and the evaluation of the financial scheme;

[0013] Acquire the conversation data to be tested, and obtain future transaction prediction data according to the conversation data to be tested and the transaction prediction model, wherein the conversation data to be tested is configured to include at least new evaluation data received by the large language model;

[0014] The financial plan is optimized according to the future transaction forecast data.

[0015] Optionally, inputting the first text into a large language model according to the first text to obtain a first financial solution includes:

[0016] According to the first text, recall and search the first text in a preset financial knowledge base to obtain a first recall result;

[0017] The first text and the first recall result are input into a large language model to obtain a first financial solution.

[0018] Optionally, the acquiring the first transaction data of the first user, clustering the first users according to the first transaction data and the first evaluation, and obtaining a clustering result includes:

[0019] Acquire first transaction data of the first user, parse the first transaction data, and obtain a first transaction feature corresponding to the first user;

[0020] According to the first evaluation, based on the characteristics of the first solution corresponding to the first evaluation, obtaining a first evaluation feature corresponding to the first user;

[0021] The first users are clustered according to the first transaction feature and the first evaluation feature to obtain a clustering result.

[0022] Optionally, clustering the first users according to the first transaction feature and the first evaluation feature to obtain a clustering result includes:

[0023] According to the first transaction feature and the first evaluation feature, combining the first transaction feature and the first evaluation feature into a first comprehensive vector;

[0024] According to the first comprehensive vector, standardizing the features of the first comprehensive vector;

[0025] Clustering the first user using a DBSCAN algorithm according to the first comprehensive vector to obtain a clustering result;

[0026] The value of MinPts of the DBSCAN algorithm is configured to be obtained according to the number of solutions in the first financial solution.

[0027] Optionally, establishing a transaction prediction model according to the clustering result includes:

[0028] Establishing an initial transaction prediction model, wherein the initial transaction prediction model is configured as a prediction model that takes evaluation data as input and takes transaction status as output;

[0029] According to the clustering results, the initial transaction prediction model is trained.

[0030] Optionally, the acquiring the first transaction data of the first user, parsing the first transaction data, and obtaining a first transaction feature corresponding to the first user includes:

[0031] acquiring first transaction data of the first user, parsing the first transaction data, and obtaining a first transaction time and a first transaction plan from the first transaction data;

[0032] Obtaining a first transaction feature corresponding to the first user according to the first transaction time and the feature of the first transaction plan;

[0033] The obtaining, according to the first evaluation and based on the feature of the first solution corresponding to the first evaluation, a first evaluation feature corresponding to the first user includes:

[0034] According to the first evaluation, obtaining a first evaluation time of the first evaluation;

[0035] obtaining a first evaluation feature corresponding to the first user based on the first evaluation, the feature of the first solution, and the first evaluation time;

[0036] The characteristics of the first solution are of the same category as the characteristic parameters in the characteristics of the first transaction solution.

[0037] Optionally, combining the first transaction feature and the first evaluation feature into a first comprehensive vector according to the first transaction feature and the first evaluation feature includes:

[0038] Obtaining a first interval time according to the first transaction time and the first evaluation time of the first user;

[0039] Obtaining a first solution difference according to the first transaction solution of the first user and the first solution;

[0040] A first comprehensive vector is obtained according to the first interval time and the first solution difference corresponding to each user among the first users.

[0041] In another aspect, an automobile financial service optimization system based on a large language model includes a large language model platform and a financial service platform;

[0042] The large language model platform is configured as follows:

[0043] Acquire a first text input by a first user, where the first text is configured as a text in which the user inquires about automobile financial solutions;

[0044] According to the first text, inputting the first text into a large language model to obtain a first financial solution;

[0045] Acquire a second text input by the first user, where the second text is configured as a text sent by the user in response to the first financial plan;

[0046] parsing the second text to obtain a first evaluation of the first financial plan in the second text;

[0047] The financial service platform is configured to:

[0048] Acquire first transaction data of the first user, and cluster the first users according to the first transaction data and the first evaluation to obtain a clustering result;

[0049] According to the clustering result, a transaction prediction model is established, wherein the transaction prediction model is configured to be a prediction model that takes evaluation data as input and takes transaction status as output, wherein the evaluation data includes the evaluated financial scheme and the evaluation of the financial scheme;

[0050] Acquire the conversation data to be tested, and obtain future transaction prediction data according to the conversation data to be tested and the transaction prediction model, wherein the conversation data to be tested is configured to include at least new evaluation data received by the large language model;

[0051] The financial plan is optimized according to the future transaction forecast data.

[0052] Optionally, the acquiring the first transaction data of the first user, clustering the first users according to the first transaction data and the first evaluation, and obtaining a clustering result includes:

[0053] Acquire first transaction data of the first user, parse the first transaction data, and obtain a first transaction feature corresponding to the first user;

[0054] According to the first evaluation, based on the characteristics of the first solution corresponding to the first evaluation, obtaining a first evaluation feature corresponding to the first user;

[0055] The first users are clustered according to the first transaction feature and the first evaluation feature to obtain a clustering result.

[0056] On the other hand, a computer device is provided. The computer device includes a memory and a processor. The memory stores a computer program. The processor executes the computer program to implement the above method.

[0057] On the other hand, a computer storage medium stores a computer program, and a processor executes the computer program to implement the above method.

[0058] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0059] The present invention provides an automobile financial service optimization method, system and device based on a large language model, comprising: obtaining a first text input by a first user, wherein the first text is configured as a text of the user querying an automobile financial plan; inputting the first text into a large language model according to the first text to obtain a first financial plan; obtaining a second text input by the first user, wherein the second text is configured as a text sent by the user for the first financial plan; parsing the second text to obtain a first evaluation of the first financial plan in the second text; obtaining first transaction data of the first user, clustering the first user according to the first transaction data and the first evaluation to obtain a clustering result; establishing a transaction prediction model according to the clustering result, wherein the transaction prediction model is configured as a prediction model with evaluation data as input and transaction status as output, wherein the evaluation data includes the evaluated financial plan and the evaluation of the financial plan; obtaining conversation data to be tested, obtaining future transaction prediction data according to the conversation data to be tested and the transaction prediction model, wherein the conversation data to be tested is configured to include at least new evaluation data received by the large language model; and optimizing the financial plan according to the future transaction prediction data. At least it solves the problems of existing automobile financial services' slow response to market and user feedback, slow perception of changes in user demand, and lack of accuracy, and provides a more intelligent and efficient financial service optimization solution to enable financial service institutions to meet the diverse and personalized financial needs of modern consumers. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the specific implementation of the present application or the technical solution in the prior art, the following is a brief introduction to the drawings required for the specific implementation or the prior art description. In all drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, each element or part is not necessarily drawn according to the actual scale.

[0061] Figure 1 A flowchart of an automobile financial service optimization method based on a large language model in this application;

[0062] Figure 2 This is a schematic diagram of the structure of a computer device in this application.

[0063] Markings in the figure: 101 - processor, 102 - communication bus, 103 - network interface, 104 - user interface, 105 - memory.

[0064] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0065] In order to enable those skilled in the art to better understand the scheme of the present disclosure, the technical scheme in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in the field without creative work should fall within the scope of protection of the present disclosure.

[0066] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices. A first user refers to one or more users that meet the first user characteristics, and a second user refers to one or more users that meet the second user characteristics.

[0067] Example 1

[0068] An automobile financial service optimization method based on a large language model comprises the following steps:

[0069] S1. Obtain a first text input by a first user.

[0070] Specifically, the first text is configured as a text used by the first user to query an automobile financial solution.

[0071] Specifically, the first text may be “What is the interest rate for a 3-year loan for XX model of XX brand” or “Which financial plan is recommended for XX model of XX brand now”, etc.

[0072] Optionally, after obtaining the first text input by the first user, a classification model is used to perform classification screening to screen out the text in which the automobile financial solution is queried.

[0073] S2. According to the first text, input the first text into a large language model to obtain a first financial solution;

[0074] Optionally, the first large language model can be a general large language model such as ChatGLM, GPT-4, etc., or it can be a customized large language model.

[0075] S3. Obtain a second text input by the first user.

[0076] Specifically, the second text is configured as a text sent by the first user for the first financial plan;

[0077] Specifically, the second text may be "Is there a plan with a shorter period of time?", "Is there a plan with a lower interest rate?" or "The total interest of this plan is too high."

[0078] S4. Parse the second text to obtain a first evaluation of the first financial solution in the second text.

[0079] Specifically, large language models are usually based on the Transformer architecture and use the self-attention mechanism to process contextual information. Through context tracking, reference analysis, self-attention mechanism and semantic reasoning, we can use the large language model to obtain the first evaluation of the first financial plan in the second text.

[0080] Specifically, when the second text is "Is there a plan with a shorter time period?", "Is there a plan with a lower interest rate?" or "The total interest of this plan is too high", the first evaluation of the first financial plan in the second text obtained by using the large language model is "The loan period is too long", "The interest rate is high" and "The total interest is too high" respectively.

[0081] S5. Obtain first transaction data of the first user, cluster the first user according to the first transaction data and the first evaluation, and obtain a clustering result;

[0082] Optionally, the clustering method includes K-means clustering, DBSCAN clustering, etc. Preferably, this solution adopts DBSCAN clustering.

[0083] Optionally, the step of obtaining first transaction data of a first user, clustering the first user according to the first transaction data and the first evaluation, and obtaining a clustering result includes:

[0084] Acquire first transaction data of the first user, and perform data preprocessing on the first transaction data and the first evaluation. Specifically, the preprocessing method may include data filling, data standardization, etc.;

[0085] Selecting DBSCAN parameters according to the first transaction data and the first evaluation. Specifically, the DBSCAN parameters may be set according to the first transaction data or the number of financial solutions in the first evaluation;

[0086] Apply DBSCAN clustering to cluster the first user and obtain the clustering result.

[0087] S6. Establish a transaction prediction model based on the clustering results.

[0088] Optionally, the transaction prediction model is configured as a prediction model that takes evaluation data as input and takes transaction status as output, and the evaluation data includes the evaluated financial plan and the evaluation of the financial plan;

[0089] Optionally, a transaction prediction model is established based on the clustering results, including:

[0090] According to the clustering result, a clustering label of each first user is obtained;

[0091] We can use the cluster label of each data point as the classification target to establish a transaction prediction model. The transaction prediction model is configured to test and verify the transaction prediction model. Specifically, the method for testing and verifying the transaction prediction model can use cross-validation, accuracy, precision, recall and other indicators.

[0092] S7. Acquire the conversation data to be measured, and obtain future transaction prediction data based on the conversation data to be measured and the transaction prediction model.

[0093] The dialogue data to be tested is configured to include at least new evaluation data received by the large language model.

[0094] The dialogue data to be tested is configured as new evaluation data received by the large language model and the financial scheme corresponding to the evaluation.

[0095] Optionally, the conversation data to be measured is obtained, and future transaction prediction data is obtained according to the conversation data to be measured and the transaction prediction model, including:

[0096] Acquire the conversation data to be tested, and obtain the classification result of the second user corresponding to the conversation data to be tested, that is, the clustering label of the second user, based on the conversation data to be tested and the transaction prediction model; obtain the transaction prediction data of the second user based on the classification result of the second user; and obtain future transaction prediction data based on the transaction prediction data of the second user.

[0097] S8. Optimize financial solutions based on future transaction forecast data.

[0098] Optionally, the financial plan is optimized based on future transaction forecast data and business objectives or business strategies so that future actual transaction data meets the business objectives or business strategies.

[0099] Optionally, the correspondence between the first user, the first text, the first financial plan, the second text, the first evaluation, and the first transaction data can be a one-to-one correspondence or a one-to-many correspondence, but in general, the first text, the first financial plan, the second text and the first evaluation correspond one-to-one, and the first user and the first transaction data correspond one-to-one.

[0100] Specifically, in the structured data table storing the above data, each piece of data must at least include the user ID of the first user.

[0101] This solution is used to establish a clustering and classification model by analyzing the situation of users consulting financial solutions, feedback on financial solutions and actual transactions, analyze the relationship between users' evaluation of financial solutions and the financial solutions they finally choose, and then predict the user's future transactions based on the newly received situation of users consulting financial solutions and feedback on financial solutions. It solves the problems of the existing financial solutions for automobile financial services that are slow to respond to market and user feedback, slow to perceive changes in user needs, and lack of accuracy, and provides a more intelligent and efficient financial service optimization solution so that financial service institutions can meet the diversified and personalized financial needs of modern consumers.

[0102] Example 2

[0103] This embodiment is based on the first embodiment, and is a method for optimizing automobile financial services based on a large language model, comprising the following steps:

[0104] S1. Obtain a first text input by a first user.

[0105] Specifically, the first text is configured as text input by the first user for querying automobile financial solutions.

[0106] Specifically, the first text may be “What is the interest rate for a 3-year loan for XX model of XX brand” or “Which financial plan is recommended for XX model of XX brand now”, etc.

[0107] Optionally, after obtaining the first text input by the first user, a classification model is used to perform classification screening to screen out the text in which the automobile financial solution is queried.

[0108] S2. According to the first text, input the first text into a large language model to obtain a first financial solution;

[0109] Optionally, the first financial solution is configured as a financial solution output by a large language model and obtained by the first user through the first text. The first financial solution is a financial solution or a template of the financial solution and key parameters corresponding to the template stored in a knowledge base connected to the large language model.

[0110] Optionally, the first large language model can be a general large language model such as ChatGLM, GPT-4, etc., or it can be a customized large language model.

[0111] Optionally, according to the first text, inputting the first text into a large language model to obtain a first financial solution includes:

[0112] According to the first text, recall and search the first text in a preset financial knowledge base to obtain a first recall result;

[0113] The first text and the first recall result are input into the large language model to obtain a first financial solution.

[0114] Specifically, the pseudo code for implementing the steps and main steps of the above method may include:

[0115] S201, obtaining a first text;

[0116] S202: performing a recall search on the first text in a preset financial knowledge base according to the first text to obtain a first recall result;

[0117] S203, inputting a large language model to generate a financial solution;

[0118] The pseudo code of the main function of step S203 includes:

[0119] def main():

[0120] #Enter the first text

[0121] first_text = input_text()

[0122] #Enter the first recall result

[0123] recall_results=recall_from_financial_knowledge_base(first_text)

[0124] #Generate and output financial plan

[0125] financial_plan=generate_financial_plan(first_text,recall_results)

[0126] print("The generated financial plan is:")

[0127] print(financial_plan).

[0128] S3. Obtain a second text input by the first user.

[0129] Specifically, the second text is configured as a text sent by the first user for the first financial plan;

[0130] Specifically, the second text may be "Is there a plan with a shorter period of time?", "Is there a plan with a lower interest rate?" or "The total interest of this plan is too high."

[0131] Large language models are usually based on the Transformer architecture and use a self-attention mechanism to process contextual information. This allows the model to capture long-distance dependencies in the input - even in very long texts, the model can find connections between different parts. Its core components include: Self-attention mechanism: The model understands the importance of each part of the sentence by calculating the relationship between each word or token and other words. This allows the model to capture "contextual" information in the conversation, especially the relationship between pronouns and the previous and next sentences; Position encoding: The Transformer model does not have a cyclic structure, such as RNN and LSTM, so position encoding is needed to maintain the relationship between the order of words, which is crucial to understanding the order and context of the conversation; Multi-layer stacked Transformer modules: Each layer of Transformer will perform a higher level of abstract processing on the input to better understand complex language structures and contexts.

[0132] S4. Parse the second text to obtain a first evaluation of the first financial solution in the second text.

[0133] Specifically, the large language model is usually based on the Transformer architecture and uses the self-attention mechanism to process contextual information. Through context tracking, reference analysis, self-attention mechanism and semantic reasoning, we can use the large language model to obtain the first evaluation of the first financial plan in the second text.

[0134] Specifically, when the second text is "Is there a shorter plan?", "Is there a plan with a lower interest rate?" or "The total interest of this plan is too high", the first evaluation of the first financial plan in the second text obtained by using the large language model is "The loan period is too long", "The interest rate is high" and "The total interest is too high" respectively.

[0135] Optionally, when parsing the second text and obtaining the first evaluation of the first financial plan in the second text, the first evaluation can be converted into standardized data or vector to facilitate subsequent data processing; specifically, when the sample size is large, the first evaluation is converted into a standardized vector, and when the sample size is insufficient, the first evaluation is converted into standardized data.

[0136] Specifically, the method of converting the first evaluation into standardized data or vectors may adopt an indicator system method based on keyword and / or context analysis, an expert scoring method, or an AI model trained based on expert scoring data.

[0137] S5. Obtain first transaction data of the first user, cluster the first user according to the first transaction data and the first evaluation, and obtain a clustering result;

[0138] Optionally, the clustering method includes K-means clustering, DBSCAN clustering, etc. Preferably, this solution adopts DBSCAN clustering.

[0139] Optionally, the step of obtaining first transaction data of a first user, clustering the first user according to the first transaction data and the first evaluation, and obtaining a clustering result includes:

[0140] Acquire first transaction data of the first user, and perform data preprocessing on the first transaction data and the first evaluation. Specifically, the preprocessing method may include data filling, data standardization, etc.;

[0141] Selecting DBSCAN parameters according to the first transaction data and the first evaluation. Specifically, the DBSCAN parameters may be set according to the first transaction data or the number of financial solutions in the first evaluation;

[0142] Apply DBSCAN clustering to cluster the first user and obtain the clustering result.

[0143] Optionally, obtaining first transaction data of a first user, clustering the first user according to the first transaction data and the first evaluation, and obtaining a clustering result includes:

[0144] Acquire first transaction data of a first user, parse the first transaction data, and obtain a first transaction feature corresponding to the first user;

[0145] According to the first evaluation, based on the characteristics of the first solution corresponding to the first evaluation, obtaining a first evaluation feature corresponding to the first user;

[0146] The first user is clustered according to the first transaction feature and the first evaluation feature to obtain a clustering result.

[0147] Optionally, clustering the first user according to the first transaction feature and the first evaluation feature to obtain a clustering result includes:

[0148] According to the first transaction feature and the first evaluation feature, the first transaction feature and the first evaluation feature are combined into a first comprehensive vector;

[0149] According to the first comprehensive vector, a feature of the first comprehensive vector is normalized;

[0150] Clustering the first user using a DBSCAN algorithm according to the first comprehensive vector to obtain a clustering result;

[0151] The value of MinPts of the DBSCAN algorithm is configured to be obtained according to the number of solutions in the first financial solution.

[0152] Optionally, the eps of the DBSCAN algorithm is configured to be obtained via a K-distance graph.

[0153] Specifically, the value of MinPts of the DBSCAN algorithm is configured to be obtained by dividing the number of first users by the number of schemes in the first financial scheme and then dividing or subtracting a preset parameter.

[0154] Specifically, all features of the first comprehensive vector are standardized so that the dimensions of different features are consistent to avoid excessive influence of certain features.

[0155] Optionally, obtaining first transaction data of the first user, parsing the first transaction data, and obtaining a first transaction feature corresponding to the first user includes:

[0156] Acquire first transaction data of a first user, parse the first transaction data, and obtain a first transaction time and a first transaction plan from the first transaction data;

[0157] Obtaining a first transaction feature corresponding to the first user according to the first transaction time and the feature of the first transaction plan;

[0158] According to the first evaluation, based on the feature of the first solution corresponding to the first evaluation, obtaining a first evaluation feature corresponding to the first user includes:

[0159] According to the first evaluation, the first evaluation time of the first evaluation is obtained;

[0160] obtaining a first evaluation feature corresponding to the first user based on the first evaluation, the feature of the first solution, and the first evaluation time;

[0161] The characteristics of the first plan and the categories of the characteristic parameters in the characteristics of the first deal plan are the same.

[0162] Specifically, the characteristics of the first transaction plan include the amount of financial service fees, down payment ratio, interest rate, loan term, total interest, etc. obtained from the characteristics of the first transaction plan.

[0163] Specifically, the first evaluation feature includes one or more of the amount of financial service fee, down payment ratio, interest rate, loan term, total interest, etc. obtained from the characteristics of the first solution; and also includes standardized data or vectors obtained from the first evaluation.

[0164] Optionally, according to the first transaction feature and the first evaluation feature, the first transaction feature and the first evaluation feature are combined into a first comprehensive vector, including:

[0165] Obtaining a first interval time according to a first transaction time and a first evaluation time of a first user;

[0166] Obtaining a first plan difference according to the first transaction plan of the first user and the first plan;

[0167] A first comprehensive vector is obtained according to the first interval time and the first solution difference corresponding to each user among the first users.

[0168] According to the first transaction plan and the first plan of the first user, a method for obtaining the difference between the first plan may be to convert the first transaction plan and the first plan into vectors according to the features and then calculate them, or to calculate them first and then convert them into vectors.

[0169] S6. Establish a transaction prediction model based on the clustering results.

[0170] Optionally, the transaction prediction model is configured as a prediction model that takes evaluation data as input and takes transaction status as output, and the evaluation data includes the evaluated financial plan and the evaluation of the financial plan;

[0171] Optionally, a transaction prediction model is established based on the clustering results, including:

[0172] According to the clustering result, a clustering label of each first user is obtained;

[0173] We can use the cluster label of each data point as the classification target to establish a transaction prediction model. The transaction prediction model is configured to test and verify the transaction prediction model. Specifically, the method for testing and verifying the transaction prediction model can use cross-validation, accuracy, precision, recall and other indicators.

[0174] Optionally, a transaction prediction model is established based on the clustering results, including:

[0175] Establishing an initial transaction prediction model, which is configured as a prediction model that takes evaluation data as input and transaction status as output;

[0176] Based on the clustering results, the initial transaction prediction model is trained.

[0177] Optionally, the transaction prediction model can use classification models such as logistic regression, random forest, and SVM.

[0178] Specifically, a pseudo code example for implementing the above solution uses K-means clustering to cluster the first user and uses logistic regression as the transaction prediction model:

[0179] #Data preprocessing

[0180] #Assume that X is the feature data and y is the target variable

[0181] #X = pd.DataFrame(....)

[0182] #y=pd.Series(....)

[0183] #Cluster analysis (using K-means clustering)

[0184] kmeans = KMeans(n_clusters = 3, random_state = 42) # Assume we are divided into 3 clusters

[0185] cluster_labels=kmeans.fit_predict(X_scaled)

[0186] #Add cluster labels as new features to the original data

[0187] X_with_labels=np.hstack([X_scaled,cluster_labels.reshape(-1,1)])

[0188] #Build a prediction model

[0189] # Split the data set into training and test sets

[0190] X_train,X_test,y_train,y_test=train_test_split(X_with_labels,y,test_size=0.2,random_state=42)

[0191] #Use logistic regression as a prediction model

[0192] model = LogisticRegression()

[0193] model.fit(X_train,y_train).

[0194] S7. Acquire the conversation data to be measured, and obtain future transaction prediction data based on the conversation data to be measured and the transaction prediction model.

[0195] The dialogue data to be tested is configured to include at least new evaluation data received by the large language model.

[0196] The dialogue data to be tested is configured as new evaluation data received by the large language model and the financial scheme corresponding to the evaluation.

[0197] Optionally, the conversation data to be measured is obtained, and future transaction prediction data is obtained according to the conversation data to be measured and the transaction prediction model, including:

[0198] The conversation data to be tested is obtained, and a classification result of a second user corresponding to the conversation data to be tested, that is, a clustering label of the second user, is obtained according to the conversation data to be tested and a transaction prediction model; transaction prediction data of the second user is obtained according to the classification result of the second user; and future transaction prediction data is obtained according to the transaction prediction data of the second user. The second user is configured as a user corresponding to the conversation data to be tested.

[0199] Specifically, the conversation data to be tested is structured data, including at least one second user, a second financial plan corresponding to the second user, and a second evaluation corresponding to the second financial plan. In the structured data table storing the above data, each data item must at least include the user ID of the second user.

[0200] After the above-mentioned conversation data to be tested is input into the transaction prediction model, the future transaction prediction data is obtained as the transaction data corresponding to the second user, which represents the future transaction status of each user in the second user, such as one user is expected to complete the transaction in 3 months, and another user is expected to complete the transaction after the interest rate is lowered, etc. That is, after the above-mentioned conversation data to be tested is input into the transaction prediction model, the prediction of the transaction status under different conditions in the future can be obtained.

[0201] S8. Optimize financial solutions based on future transaction forecast data.

[0202] Optionally, the financial plan is optimized based on future transaction forecast data and business objectives or business strategies so that future actual transaction data meets the business objectives or business strategies.

[0203] Specifically, the optimized financial plan can be combined with future transaction forecast data to obtain expected transaction data for the optimized financial plan, and based on the expected transaction data for the optimized financial plan, it can be determined whether the business objectives or business strategies are met.

[0204] Example 3

[0205] An automobile financial service optimization system based on a large language model, comprising a large language model platform and a financial service platform;

[0206] The large language model platform is configured as follows:

[0207] Acquire a first text input by a first user, where the first text is configured as a text in which the user queries for an automobile financial solution;

[0208] According to the first text, inputting the first text into a large language model to obtain a first financial solution;

[0209] Acquire a second text input by the first user, where the second text is configured as a text sent by the user for the first financial plan;

[0210] Parsing the second text to obtain a first evaluation of the first financial plan in the second text;

[0211] The financial services platform is configured to:

[0212] Acquire first transaction data of a first user, cluster the first user according to the first transaction data and the first evaluation, and obtain a clustering result;

[0213] According to the clustering results, a transaction prediction model is established. The transaction prediction model is configured to be a prediction model that takes the evaluation data as input and the transaction status as output. The evaluation data includes the evaluated financial plan and the evaluation of the financial plan.

[0214] Obtain the conversation data to be measured, and obtain future transaction prediction data based on the conversation data to be measured and the transaction prediction model;

[0215] Optimize financial solutions based on future transaction forecast data.

[0216] Optionally, according to the first text, inputting the first text into a large language model to obtain a first financial solution includes:

[0217] According to the first text, recall and search the first text in a preset financial knowledge base to obtain a first recall result;

[0218] The first text and the first recall result are input into the large language model to obtain a first financial solution.

[0219] Optionally, obtaining first transaction data of a first user, clustering the first user according to the first transaction data and the first evaluation, and obtaining a clustering result includes:

[0220] Acquire first transaction data of a first user, parse the first transaction data, and obtain a first transaction feature corresponding to the first user;

[0221] According to the first evaluation, based on the characteristics of the first solution corresponding to the first evaluation, obtaining a first evaluation feature corresponding to the first user;

[0222] The first user is clustered according to the first transaction feature and the first evaluation feature to obtain a clustering result.

[0223] Optionally, clustering the first user according to the first transaction feature and the first evaluation feature to obtain a clustering result includes:

[0224] According to the first transaction feature and the first evaluation feature, the first transaction feature and the first evaluation feature are combined into a first comprehensive vector;

[0225] According to the first comprehensive vector, a feature of the first comprehensive vector is normalized;

[0226] Clustering the first user using a DBSCAN algorithm according to the first comprehensive vector to obtain a clustering result;

[0227] The value of MinPts of the DBSCAN algorithm is configured to be obtained according to the number of solutions in the first financial solution.

[0228] Optionally, a transaction prediction model is established based on the clustering results, including:

[0229] Establishing an initial transaction prediction model, which is configured as a prediction model that takes evaluation data as input and transaction status as output;

[0230] Based on the clustering results, the initial transaction prediction model is trained.

[0231] Optionally, obtaining first transaction data of the first user, parsing the first transaction data, and obtaining a first transaction feature corresponding to the first user includes:

[0232] Acquire first transaction data of a first user, parse the first transaction data, and obtain a first transaction time and a first transaction plan from the first transaction data;

[0233] Obtaining a first transaction feature corresponding to the first user according to the first transaction time and the feature of the first transaction plan;

[0234] According to the first evaluation, based on the feature of the first solution corresponding to the first evaluation, obtaining a first evaluation feature corresponding to the first user includes:

[0235] According to the first evaluation, the first evaluation time of the first evaluation is obtained;

[0236] obtaining a first evaluation feature corresponding to the first user based on the first evaluation, the feature of the first solution, and the first evaluation time;

[0237] The characteristics of the first plan and the categories of the characteristic parameters in the characteristics of the first deal plan are the same.

[0238] Optionally, according to the first transaction feature and the first evaluation feature, the first transaction feature and the first evaluation feature are combined into a first comprehensive vector, including:

[0239] Obtaining a first interval time according to a first transaction time and a first evaluation time of a first user;

[0240] Obtaining a first plan difference according to the first transaction plan of the first user and the first plan;

[0241] A first comprehensive vector is obtained according to the first interval time and the first solution difference corresponding to each user among the first users.

[0242] Example 4

[0243] This embodiment provides a computer device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement any of the above methods.

[0244] Specifically, Figure 2 As shown, Figure 2 This is a structural diagram of a computer device in the present application, and the computer device may include: a processor 101, such as a central processing unit (CPU), a communication bus 102, a user interface 104, a network interface 103, and a memory 105. Among them, the communication bus 102 is used to realize the connection and communication between these components. The user interface 104 may include a display screen (Display), an input unit such as a keyboard (Keyboard), and the user interface 104 may also include a standard wired interface and a wireless interface. The network interface 103 may optionally include a standard wired interface and a wireless interface (such as a wireless fidelity (WIreless-FIdelity, WI-FI) interface). The memory 105 may optionally be a storage device independent of the aforementioned processor 101. The memory 105 may be a high-speed random access memory (RAM) memory, or it may be a stable non-volatile memory (NVM), such as at least one disk storage; the processor 101 may be a general-purpose processor, including a central processing unit, a network processor, etc., or it may be a digital signal processor, an application-specific integrated circuit, a field programmable gate array or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component.

[0245] Those skilled in the art will appreciate that Figure 2 The structure shown in the figure does not constitute a limitation on the electronic device, and may include more or less components than shown in the figure, or combine certain components, or arrange the components differently.

[0246] like Figure 2 As shown, the memory 105 as a storage medium may include an operating system, a network communication module, a user interface module, and an application program for implementing a method for optimizing automobile financial services based on a large language model.

[0247] exist Figure 2In the electronic device shown, the network interface 103 is mainly used for data communication with a network server; the user interface 104 is mainly used for data interaction with a user; the processor 101 and the memory 105 in the present application can be set in the electronic device, and the electronic device calls the application stored in the memory 105 for implementing a method for optimizing automobile financial services based on a large language model through the processor 101 to implement the above method.

[0248] Example 5

[0249] This embodiment provides a computer-readable storage medium, on which a computer program is stored, and a processor executes the computer program to implement any of the above methods.

[0250] In some embodiments, the computer readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or various devices including one or any combination of the above memories. The computer may be various computing devices including intelligent terminals and servers.

[0251] In the above embodiments of the present disclosure, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0252] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0253] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed over multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0254] In addition, each functional unit in each embodiment of the present disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0255] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable non-volatile storage medium. Based on this understanding, the technical solution of the present disclosure is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a non-volatile storage medium, including a number of instructions to enable a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the various embodiments of the present disclosure. The aforementioned non-volatile storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk and other media that can store program codes.

[0256] The above are only preferred embodiments of the present disclosure. It should be pointed out that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present disclosure. These improvements and modifications should also be regarded as the protection scope of the present disclosure.

Claims

1. A method for optimizing automobile financial services based on a large language model, characterized in that: The following steps are involved: Acquire a first text input by a first user, where the first text is configured as a text in which the user inquires about automobile financial solutions; According to the first text, inputting the first text into a large language model to obtain a first financial solution; Acquire a second text input by the first user, where the second text is configured as a text sent by the user in response to the first financial plan; parsing the second text to obtain a first evaluation of the first financial scheme in the second text; Acquire first transaction data of the first user, and cluster the first users according to the first transaction data and the first evaluation to obtain a clustering result; According to the clustering result, a transaction prediction model is established, wherein the transaction prediction model is configured to be a prediction model that takes evaluation data as input and takes transaction status as output, wherein the evaluation data includes the evaluated financial scheme and the evaluation of the financial scheme; Acquire the conversation data to be tested, and obtain future transaction prediction data according to the conversation data to be tested and the transaction prediction model, wherein the conversation data to be tested is configured to include at least new evaluation data received by the large language model; The financial plan is optimized according to the future transaction forecast data.

2. The automobile financial service optimization method based on a large language model according to claim 1 is characterized in that: The step of inputting the first text into a large language model to obtain a first financial solution includes: According to the first text, recall and search the first text in a preset financial knowledge base to obtain a first recall result; The first text and the first recall result are input into a large language model to obtain a first financial solution.

3. The automobile financial service optimization method based on a large language model according to claim 1 is characterized in that: The acquiring the first transaction data of the first user, clustering the first user according to the first transaction data and the first evaluation, and obtaining a clustering result includes: Acquire first transaction data of the first user, parse the first transaction data, and obtain a first transaction feature corresponding to the first user; According to the first evaluation, based on the characteristics of the first solution corresponding to the first evaluation, obtaining a first evaluation feature corresponding to the first user; The first users are clustered according to the first transaction feature and the first evaluation feature to obtain a clustering result.

4. The automobile financial service optimization method based on a large language model according to claim 3 is characterized in that: Clustering the first users according to the first transaction feature and the first evaluation feature to obtain a clustering result includes: According to the first transaction feature and the first evaluation feature, combining the first transaction feature and the first evaluation feature into a first comprehensive vector; According to the first comprehensive vector, standardizing the features of the first comprehensive vector; Clustering the first user using a DBSCAN algorithm according to the first comprehensive vector to obtain a clustering result; The value of MinPts of the DBSCAN algorithm is configured to be obtained according to the number of solutions in the first financial solution.

5. The automobile financial service optimization method based on a large language model according to claim 1, characterized in that: The step of establishing a transaction prediction model according to the clustering result includes: Establishing an initial transaction prediction model, wherein the initial transaction prediction model is configured as a prediction model that takes evaluation data as input and takes transaction status as output; According to the clustering results, the initial transaction prediction model is trained.

6. The automobile financial service optimization method based on a large language model according to claim 3 is characterized in that: The acquiring the first transaction data of the first user, parsing the first transaction data, and obtaining the first transaction feature corresponding to the first user includes: acquiring first transaction data of the first user, parsing the first transaction data, and obtaining a first transaction time and a first transaction plan from the first transaction data; Obtaining a first transaction feature corresponding to the first user according to the first transaction time and the feature of the first transaction plan; The obtaining, according to the first evaluation and based on the feature of the first solution corresponding to the first evaluation, a first evaluation feature corresponding to the first user includes: According to the first evaluation, obtaining a first evaluation time of the first evaluation; obtaining a first evaluation feature corresponding to the first user based on the first evaluation, the feature of the first solution, and the first evaluation time; The characteristics of the first solution are of the same category as the characteristic parameters in the characteristics of the first transaction solution.

7. The automobile financial service optimization method based on a large language model according to claim 6 is characterized in that: The step of combining the first transaction feature and the first evaluation feature into a first comprehensive vector according to the first transaction feature and the first evaluation feature includes: Obtaining a first interval time according to the first transaction time and the first evaluation time of the first user; Obtaining a first solution difference according to the first transaction solution of the first user and the first solution; A first comprehensive vector is obtained according to the first interval time and the first solution difference corresponding to each user among the first users.

8. An automobile financial service optimization system based on a large language model, characterized in that: Including large language model platform and financial service platform; The large language model platform is configured as follows: Acquire a first text input by a first user, where the first text is configured as a text in which the user inquires about automobile financial solutions; According to the first text, inputting the first text into a large language model to obtain a first financial solution; Acquire a second text input by the first user, where the second text is configured as a text sent by the user in response to the first financial plan; parsing the second text to obtain a first evaluation of the first financial scheme in the second text; The financial service platform is configured to: Acquire first transaction data of the first user, and cluster the first users according to the first transaction data and the first evaluation to obtain a clustering result; According to the clustering result, a transaction prediction model is established, wherein the transaction prediction model is configured to be a prediction model that takes evaluation data as input and takes transaction status as output, wherein the evaluation data includes the evaluated financial scheme and the evaluation of the financial scheme; Acquire the conversation data to be tested, and obtain future transaction prediction data according to the conversation data to be tested and the transaction prediction model, wherein the conversation data to be tested is configured to include at least new evaluation data received by the large language model; The financial plan is optimized according to the future transaction forecast data.

9. The automobile financial service optimization system based on a large language model according to claim 8, characterized in that: The acquiring the first transaction data of the first user, clustering the first user according to the first transaction data and the first evaluation, and obtaining a clustering result includes: Acquire first transaction data of the first user, parse the first transaction data, and obtain a first transaction feature corresponding to the first user; According to the first evaluation, based on the characteristics of the first solution corresponding to the first evaluation, obtaining a first evaluation feature corresponding to the first user; The first users are clustered according to the first transaction feature and the first evaluation feature to obtain a clustering result.

10. A device, characterized in that: The device comprises a memory and a processor, wherein a computer program is stored in the memory, and the processor executes the computer program to implement the method according to any one of claims 1 to 7.

Images