Consumption willingness assessment method, device and equipment

Abstract

The invention relates to the technical field of computers, and discloses a consumption willingness evaluation method, device and equipment, and the method comprises the steps: obtaining the sample data of a sample driver, and obtaining the consumption willingness value of the sample driver based on the sample data; extracting a sample sequence of the sample data under the target dimension, constructing a first model based on the sample sequence and the consumption intention value to obtain user data of the target driver, and extracting a feature sequence of the user data under at least one specified dimension; and inputting the feature sequence into the first model to obtain a consumption willingness prediction value of the target driver. The method can improve the prediction accuracy of the consumption willingness of the driver.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and particularly to a method, device and equipment for evaluating consumption willingness. Background Art

[0002] With the rapid development of technology and economy, the consumption behavior of drivers has become one of the focuses of enterprises' attention. More and more enterprises need to enhance the relationship with customers, maximize enterprise sales revenue and improve customer retention. Accurately predicting the consumption willingness of drivers can help enterprises formulate more effective marketing strategies and also contribute to improving the service quality and satisfaction of drivers.

[0003] In related technologies, usually only the behavioral characteristics of drivers are obtained to predict the consumption willingness of drivers. The historical data of drivers is not reused, resulting in low accuracy and loss of potential customers.

[0004] In view of this, a method for evaluating consumption willingness is needed, which can improve the prediction accuracy of drivers' consumption willingness. Summary of the Invention

[0005] In view of this, the present invention provides a method for evaluating consumption willingness, which can improve the prediction accuracy of drivers' consumption willingness.

[0006] In a first aspect, the present invention provides a method for evaluating consumption willingness, the method comprising: obtaining sample data of sample drivers, and based on the sample data, obtaining the consumption willingness values of the sample drivers; extracting a sample sequence of the sample data in a target dimension, and based on the sample sequence and the consumption willingness values, constructing a first model. Obtaining user data of target drivers, and extracting a feature sequence of the user data in at least one specified dimension; inputting the feature sequence into the first model to obtain a predicted value of the consumption willingness of the target drivers.

[0007] In this embodiment, by obtaining the sample data of sample drivers, a sample sequence of the sample data in a target dimension and the consumption willingness values of the sample drivers are obtained to construct a first model. Then, by obtaining the user data of target drivers, extracting a feature sequence of the user data in at least one specified dimension, and sending the feature sequence to the first model, a predicted value of the consumption willingness of the target drivers is obtained. Through the above solution, according to the sample sequence of the sample data of sample drivers in a target dimension and the consumption willingness values of the sample drivers, a first model is constructed, which can make full use of the sample data of sample drivers and improve the prediction accuracy of the consumption willingness of the first model. Extracting a feature sequence of the user data of target drivers in at least one specified dimension and inputting the feature sequence into the first model to obtain a predicted value of the consumption willingness of the target drivers can accurately predict the consumption willingness of the target drivers.

[0008] In an alternative embodiment, the specified dimension includes at least one of basic information features, car rental business features, refueling business features, charging business features, maintenance business features, convenience store business features, comprehensive behavior features, and social features.

[0009] In this embodiment, a feature sequence of user data is extracted under at least one specified dimension, and the specified dimension includes at least one of basic information features, car rental business features, refueling business features, charging business features, maintenance business features, convenience store business features, comprehensive behavior features, and social features. This can simplify data representation and improve the accuracy of the predicted consumer willingness value.

[0010] In an alternative embodiment, after obtaining the predicted consumer willingness value of the target driver, it further includes: determining the classification result of the target driver based on the predicted consumer willingness value of the target driver; and determining the corresponding push strategy for the target driver based on the classification result.

[0011] In this embodiment, the target driver is classified according to the predicted consumer willingness value of the target driver, so as to determine the corresponding push strategy for the target driver. The target driver can be pushed accordingly according to the predicted consumer willingness value, which can optimize the usage experience of the target driver.

[0012] In an alternative embodiment, extracting the sample sequence of the sample data under the target dimension includes: extracting the reference sequences of the sample data under several dimensions; calculating the Pearson correlation coefficient between the reference sequences under different dimensions and the consumer willingness value of the sample driver based on the reference sequences and the consumer willingness value of the sample driver; determining the target dimension based on the Pearson correlation coefficient; and obtaining the sample sequence of the sample data under the target dimension, where the sample sequence is the feature sequence of the sample data under the target dimension.

[0013] In this embodiment, by extracting the reference sequences of the sample data under several dimensions and constructing a set of Pearson correlation coefficients based on the reference sequences and the consumer willingness value of the sample driver, the target dimension is determined. To obtain the sample sequence of the sample data under the target dimension. Dimensions with a high correlation with the consumer willingness value of the sample driver can be obtained, thereby improving the model accuracy.

[0014] In an alternative embodiment, constructing the first model includes: performing data preprocessing on the sample sequence, where the data preprocessing includes at least one of outlier removal and missing value filling; and inputting the sample sequence after data preprocessing into the initial model for training to obtain the first model.

[0015] In this embodiment, by performing data preprocessing on the sample sequence and inputting the sample sequence after data preprocessing into the initial model for training to obtain the first model, the prediction accuracy of the first model can be improved.

[0016] In an alternative embodiment, when data preprocessing includes missing value filling, data preprocessing of the sample sequence includes: identifying missing values and non-missing values in the sample sequence, and obtaining the average value of the non-missing values in the sample sequence; filling the average value of the non-missing values into the missing values of the sample sequence.

[0017] In this embodiment, by calculating the average value of the non-missing values in the sample sequence and using the average value to fill the missing values of the sample sequence, the data integrity of the sample sequence can be improved, thereby improving the model performance of the first model.

[0018] In an alternative embodiment, inputting the sample sequence after data preprocessing into an initial model for training to obtain a first model includes: obtaining an initial model, where the initial model is an XGBoost model; determining the model parameters of the initial model, where the model parameters include at least one of the maximum depth of the tree, the subsample ratio, and the learning rate; inputting the sample sequence into the initial model and obtaining a first output result of the initial model; determining a first training result of the initial model based on the first output result and the consumption willingness value of the sample driver, where the consumption willingness value of the sample driver corresponds to the sample sequence; adjusting the model parameters of the initial model based on the first training result to obtain the first model.

[0019] In this embodiment, an XGBoost model is constructed and the model is trained using the sample sequence to adjust the model parameters to obtain the first model. The prediction accuracy of the first model can be improved.

[0020] In an alternative embodiment, the method further includes: obtaining test data of test drivers and obtaining the consumption willingness values of the test drivers based on the test data; obtaining a test sequence of the test data in a target dimension and inputting the test sequence into the first model to obtain a first consumption willingness test value; evaluating the first model based on the first consumption willingness test value and the consumption willingness values of the test drivers; adjusting the parameters of the first model based on the evaluation result.

[0021] In this embodiment, the first model is tested by obtaining the test data of the test drivers and the consumption willingness values of the test drivers, and the model parameters of the first model are adjusted according to the test results. The prediction accuracy of the first model can be improved.

[0022] Second aspect, the present invention provides a device for evaluating consumption willingness, the device comprising: an acquisition module, configured to acquire sample data of sample drivers, and based on the sample data, obtain the consumption willingness values of the sample drivers; a construction module, configured to extract sample sequences of the sample data in a target dimension, and based on the sample sequences and the consumption willingness values, construct a first model; an extraction module, configured to acquire user data of target drivers, and extract feature sequences of the user data in at least one specified dimension; an output module, configured to input the feature sequences into the first model to obtain the predicted consumption willingness values of the target drivers.

[0023] Third aspect, the present invention provides a computer device, comprising: a memory and a processor, which are communicatively connected to each other, wherein the memory stores computer instructions, and the processor executes the computer instructions to execute the consumption willingness evaluation method according to the first aspect or any corresponding embodiment thereof. Description of the Drawings

[0024] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0025] Figure 1 It is a flowchart of the consumption willingness method according to an embodiment of the present invention;

[0026] Figure 2 It is a flowchart of another consumption willingness method according to an embodiment of the present invention;

[0027] Figure 3 It is a flowchart of another consumption willingness method according to an embodiment of the present invention;

[0028] Figure 4 It is a structural block diagram of the consumption willingness device according to an embodiment of the present invention;

[0029] Figure 5 It is a hardware structure diagram of the computer device according to an embodiment of the present invention. Detailed Embodiments

[0030] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0031] An embodiment of the present invention provides a method for evaluating consumption willingness. By obtaining sample data of sample drivers, a sample sequence of the sample data in a target dimension and the consumption willingness value of the sample drivers are obtained to construct a first model. Then, by obtaining user data of target drivers, a feature sequence of the user data in at least one specified dimension is extracted and sent to the first model to obtain a predicted consumption willingness value of the target drivers. Through the above solution, according to the sample sequence of the sample data of the sample drivers in the target dimension and the consumption willingness value of the sample drivers, the first model is constructed, which can make full use of the sample data of the sample drivers and improve the prediction accuracy of the consumption willingness of the first model. Extracting the feature sequence of the user data of the target drivers in at least one specified dimension and inputting the feature sequence into the first model to obtain the predicted consumption willingness value of the target drivers can accurately predict the consumption willingness of the target drivers.

[0032] According to an embodiment of the present invention, an embodiment of a method for evaluating consumption willingness is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0033] In this embodiment, a method for evaluating consumption willingness is provided. Figure 1 It is a flowchart of the method for evaluating consumption willingness according to an embodiment of the present invention, as Figure 1 shown. The process includes the following steps:

[0034] Step S101, obtain sample data of sample drivers, and based on the sample data, obtain the consumption willingness value of the sample drivers.

[0035] Among them, the sample drivers can be contracted drivers of a online car-hailing operation company, and the sample data can be the order data of the contracted drivers in the past period of time. Among them, the order data can include the historical orders of the contracted drivers, etc. In some alternative embodiments, the sample data of the sample drivers can also be obtained through methods such as questionnaires. Different types, regions, ages, etc. of the drivers can be considered for stratification, and then sampling is performed to determine the sample drivers. Then, the sample data of the sample drivers is obtained by designing a questionnaire.

[0036] After obtaining the sample data, the sample data can be first sorted out and cleaned to enhance the data quality of the sample data and improve the prediction accuracy of the subsequent first model.

[0037] In some alternative embodiments, when obtaining the sample data, relevant indicators of consumption willingness can be added so as to directly extract the consumption willingness value of the sample drivers.

[0038] In some alternative embodiments, according to the sample data, the influence weights of different indicators in the sample data on the consumption willingness can be determined, and based on the sample data and the influence weights, the consumption willingness value of the sample drivers can be obtained by the method of weighted average.

[0039] In some alternative embodiments, the consumption willingness value of the sample drivers can also be labeled by professionals to ensure the accuracy of the data. In an actual application, the consumption willingness value of the drivers is labeled as "high", "relatively high", "medium", "relatively low" and "low".

[0040] After obtaining the sample data of the sample drivers, the sample data can be first cleaned and integrated, etc. Among them, data cleaning can include deleting or imputing missing values, removing outliers, etc. In the case of data missing, the average value of the remaining non-missing data can be calculated to fill the missing data.

[0041] Step S102: Extract the sample sequence of the sample data in the target dimension, and construct a first model based on the sample sequence and the consumption willingness value.

[0042] Among them, the target dimension can be determined according to the actual situation, and can include dimensions such as the driver's monthly average income situation, the expenditure on vehicle operation, and the driver's consumption attitude. Then, the features closely related to the target dimension are screened out from the sample data to generate the sample sequence in the target dimension. Among them, the features can be screened out through feature selection algorithms such as the recursive feature elimination method.

[0043] In some alternative embodiments, several dimensions can be first determined, and then the Spearman correlation coefficient between different dimensions and the consumption willingness value is calculated. According to the calculation results, the features highly correlated with the consumption willingness value are selected to generate the sample sequence in the target dimension.

[0044] Among them, the model type of the first model can be determined according to the actual situation. It can be a classification model such as logistic regression or decision tree. The sample data can be divided into a training set, a validation set, and a test set according to a certain ratio. The selected first model is trained using the training set, and by continuously adjusting the parameters of the model, the accuracy index of the first model on the training set is maximized. Then, the validation set is used to verify and optimize the first model, and based on the verification results, a model configuration with better performance is found. Finally, the optimized first model is evaluated using the test set, and it can be determined whether the first model can meet the application requirements by calculating the corresponding evaluation metrics. In some alternative embodiments, when the first model does not meet the application requirements, the first model can be retrained to improve the prediction accuracy of the first model.

[0045] In a practical application, the sample sequence is divided into a training set and a test set according to a ratio of 8:2. That is, when there are 1000 sample sequences, the training set includes 800 and the test set includes 200.

[0046] Step S103, obtain the user data of the target driver, and extract the feature sequence of the user data under at least one specified dimension.

[0047] The user data of the target driver can be directly obtained through the operation platform. By obtaining the user data of the target driver, the consumption willingness of the target driver can be predicted by the first model, so as to optimize the push strategy and improve the driver's usage experience. Among them, the target driver is the driver for whom the consumption willingness needs to be obtained, and can be selected according to the actual situation.

[0048] Features closely related to the specified dimension can be screened out from the user data to generate a feature sequence under the specified dimension. Among them, feature selection algorithms such as the recursive feature elimination method can be used to screen out features and perform feature extraction. Among them, the specified dimension can be selected from the target dimensions to improve the prediction accuracy of the first model.

[0049] Step S104, input the feature sequence into the first model to obtain the predicted value of the consumption willingness of the target driver.

[0050] After loading the first model, it is necessary to first determine the input requirements of the first model, and preprocess the feature sequence according to the data input requirements of the first model to ensure that the feature sequence meets the input requirements of the first model. Then, the feature sequence is input into the first model, and the first model makes a prediction and returns the corresponding prediction result. Among them, the output result of the first model is the predicted value of the consumption willingness of the target driver.

[0051] The consumption willingness evaluation method provided in this embodiment obtains the sample data of sample drivers, obtains the sample sequence of the sample data in the target dimension and the consumption willingness value of the sample drivers, so as to construct a first model. Then, by obtaining the user data of the target drivers, extracting the feature sequence of the user data in at least one specified dimension, and sending the feature sequence into the first model, the consumption willingness prediction value of the target drivers can be obtained. Through the above solution, according to the sample sequence of the sample data of the sample drivers in the target dimension and the consumption willingness value of the sample drivers, a first model is constructed, which can make full use of the sample data of the sample drivers and improve the prediction accuracy of the consumption willingness of the first model. Extracting the feature sequence of the user data of the target drivers in at least one specified dimension and inputting the feature sequence into the first model to obtain the consumption willingness prediction value of the target drivers can accurately predict the consumption willingness of the target drivers.

[0052] In this embodiment, a consumption willingness evaluation method is provided. Figure 2 It is a flowchart of the consumption willingness evaluation method according to an embodiment of the present invention, as Figure 2 shown, and the process includes the following steps:

[0053] Step S201, obtain the sample data of the sample drivers, and based on the sample data, obtain the consumption willingness value of the sample drivers.

[0054] For details, please refer to Figure 1 Step S101 of the embodiment shown, which will not be elaborated here.

[0055] Step S202, extract the sample sequence of the sample data in the target dimension, and based on the sample sequence and the consumption willingness value, construct a first model.

[0056] Specifically, the above step S202 includes:

[0057] Step S2021, extract the reference sequences of the sample data in several dimensions.

[0058] Among them, the dimensions can be selected according to the actual situation, and can include the personal background, consumption habits, income situation, etc. of the drivers. After determining the dimensions, the features of the sample data in the corresponding dimensions can be directly extracted, and then the reference sequences in the corresponding dimensions can be constructed.

[0059] Step S2022, based on the reference sequences and the consumption willingness value of the sample drivers, calculate the Pearson correlation coefficients between the reference sequences in different dimensions and the consumption willingness value of the sample drivers.

[0060] Among them, when the reference sequence is a numerical value, the Pearson coefficient between the reference sequence and the consumption willingness value can be directly calculated. When the reference sequence is not a numerical value, according to the actual situation, a quantization rule for the reference sequence can be formulated to quantize the reference sequence to obtain a corresponding numerical value, and then the Pearson coefficient between the reference sequence and the consumption willingness value is calculated.

[0061] Step S2023, determine the target dimension based on the Pearson coefficient.

[0062] The correlation between the dimension and the consumption willingness value can be determined according to the Pearson coefficients of different dimensions, and a suitable dimension can be selected as the target dimension according to the strength of the correlation.

[0063] In some alternative embodiments, the Pearson coefficients of different dimensions can be sorted, and the dimension with a higher Pearson coefficient can be selected as the target dimension. The Pearson coefficient can characterize the relationship between the reference sequence and the consumption willingness value under the corresponding dimension. The higher the Pearson coefficient, the stronger the positive correlation between the corresponding dimension and the consumption willingness.

[0064] Step S2024, obtain the sample sequence of the sample data under the target dimension, where the sample sequence is the feature sequence of the sample data under the target dimension.

[0065] The reference sequence under the target dimension can be directly used as the sample sequence.

[0066] Step S2025, construct the first model based on the sample sequence and the consumption willingness value.

[0067] For details, please refer to step S102, which will not be elaborated here.

[0068] In some alternative embodiments, the above step S2025 further includes:

[0069] Step 2025-1, perform data preprocessing on the sample sequence, where the data preprocessing includes at least one of outlier removal and missing value filling.

[0070] Among them, when the data preprocessing includes outlier removal, the definition of outliers can be determined according to the actual situation, and then the outliers in the sample sequence are identified and deleted to improve the data quality of the sample sequence.

[0071] When the data preprocessing includes missing value filling, the filling method can be determined according to the data type of the missing value, and the filling value can be determined by using the average value or the model prediction method for missing value filling.

[0072] In some alternative embodiments, the data preprocessing includes missing value filling, and the above step S2025-1 includes:

[0073] Step a1: Identify the missing values and non-missing values in the sample sequence, and obtain the average value of the non-missing values in the sample sequence.

[0074] First, a comprehensive check can be performed on the sample sequence to determine which missing values are included. Then, for numerical variables, calculate the average value of the remaining non-missing values.

[0075] Step a2: Fill the missing values in the sample sequence with the average value of the non-missing values.

[0076] By calculating the average value of the non-missing values in the sample sequence and using the average value to fill the missing values in the sample sequence, the data integrity of the sample sequence can be improved, thereby improving the model performance of the first model.

[0077] Step S2025-2: Input the preprocessed sample sequence into the initial model for training to obtain the first model.

[0078] In some alternative embodiments, the above step S2025-2 includes:

[0079] Step b1: Obtain the initial model, where the initial model is an XGBoost model.

[0080] Among them, the XGBoost (eXtreme Gradient Boosting) model is a gradient boosting algorithm based on decision tree ensemble and is a type in the ensemble learning framework. It is very suitable for dealing with classification problems and can effectively capture the complex relationship between the sequence and the consumption willingness value, providing accurate prediction results.

[0081] Step b2: Determine the model parameters of the initial model, where the model parameters include at least one of the maximum tree depth, subsample ratio, and learning rate.

[0082] Among them, the model parameters can be determined according to the actual situation. In the case where the number of features in the sample sequence is large, the maximum tree depth can be appropriately reduced.

[0083] In some alternative embodiments, a grid search or random search method is used to determine the model parameters of the initial model.

[0084] Step b3: Input the sample sequence into the initial model and obtain the first output result of the initial model.

[0085] Step b4: Based on the first output result and the consumption willingness value of the sample driver, determine the first training result of the initial model, where the consumption willingness value of the sample driver corresponds to the sample sequence.

[0086] In the case where the difference between the first output result and the consumption willingness value of the sample driver is large, it indicates that the prediction accuracy of the initial model is poor, that is, the first training result is poor. In the case where the difference between the first output result and the consumption willingness value of the sample driver is small, it indicates that the prediction accuracy of the initial model is high, that is, the first training result is good.

[0087] Step b5, based on the first training result, adjust the model parameters of the initial model to obtain the first model.

[0088] In the case where the first training result is poor, corresponding parameter adjustment or retraining is required. In the case where the first training result is good, fine-tuning can be performed on the model parameters to further improve the prediction accuracy of the initial model.

[0089] The parameter adjustment strategy can be determined according to the actual situation.

[0090] Construct an XGBoost model, use the sample sequence to train the model to adjust the model parameters, and obtain the first model. The prediction accuracy of the first model can be improved.

[0091] By performing data preprocessing on the sample sequence and inputting the preprocessed sample sequence into the initial model for training to obtain the first model, the prediction accuracy of the first model can be improved.

[0092] In some alternative embodiments, after step S2025-2, it further includes:

[0093] Step c1, obtain the test data of the test driver, and based on the test data, obtain the consumption willingness value of the test driver.

[0094] Among them, the sample data and the test data can be obtained together, and the test data does not participate in the training and construction of the first model to prevent affecting the test result of the first model.

[0095] Step c2, obtain the test sequence of the test data in the target dimension, and input the test sequence into the first model to obtain the first consumption willingness test value.

[0096] Step c3, evaluate the first model based on the first consumption willingness test value and the consumption willingness value of the test driver.

[0097] Among them, the evaluation index can be determined according to the actual situation, and can include accuracy, precision, and recall, etc., and then according to the first consumption willingness test value and the consumption willingness value of the test driver, determine the evaluation result of the first model.

[0098] In some alternative embodiments, the evaluation index includes precision, and the calculation method of precision is shown in the following formula:

[0099]

[0100] Among them, Precision is the precision, TP represents the number of samples that are actually positive and predicted as positive, and FP represents the number of samples that are actually negative but predicted as positive.

[0101] In some alternative embodiments, the evaluation metric includes recall rate, and the calculation method of the recall rate is shown in the following formula:

[0102]

[0103] Among them, Recall is the recall rate, and FN represents the number of samples that are actually positive but predicted as negative.

[0104] In some alternative embodiments, the evaluation metric includes F1 score, and the calculation method of the F1 score is shown in the following formula:

[0105]

[0106] Among them, F1 represents the F1 score.

[0107] Step c4, based on the evaluation result, adjust the parameters of the first model.

[0108] When the evaluation result is overfitting, the complexity of the model can be reduced. When the evaluation result is underfitting, the complexity of the model can be enhanced.

[0109] By obtaining the test data of the test driver and the consumption willingness value of the test driver, testing the first model, and adjusting the model parameters of the first model according to the test result, the prediction accuracy of the first model can be improved.

[0110] Step S203, obtain the user data of the target driver, and extract the feature sequence of the user data under at least one specified dimension.

[0111] For details, please refer to Figure 1 Step S103 of the embodiment shown, which will not be elaborated here.

[0112] In some alternative embodiments, the specified dimension includes at least one of basic information features, car rental business features, gas station business features, charging business features, maintenance business features, convenience store business features, comprehensive behavior features, and social features.

[0113] Among them, driver information can be collected through software. The basic information features may include user ID (Identification), user age, user gender, registration time, and geographical location, etc. Among them, the user ID is the unique identifier for each user, the registration time is the registration time of the user on the global travel platform, and the geographical location is the geographical location commonly used by the user, such as city, area, etc.

[0114] The car rental business features may include car rental frequency, car rental duration, car rental type, car rental time preference, number of historical orders, order cancellation rate, and order completion rate, etc. Among them, the car rental frequency is the frequency of car rental by the user, such as the number of car rentals per month; the car rental duration is the average duration of each car rental; the car rental type is the type of vehicle rented by the user, such as economy type, business type, luxury type, etc.; the car rental time preference is the time period preferred by the user for car rental, such as weekdays, weekends, daytime, nighttime, etc.; the number of historical orders is the total number of car rental historical orders; the order cancellation rate is the proportion of car rental orders cancelled by the user; the order completion rate is the proportion of car rental orders completed by the user.

[0115] The refueling business features may include refueling frequency, refueling amount, gas station preference, refueling type, and refueling time preference, etc. Among them, the refueling frequency can be the frequency of refueling by the user on the global travel platform; the refueling amount is the average amount of each refueling; the gas station preference is the location or brand of the gas station often chosen by the user; the refueling type is the type of fuel for refueling, such as No. 92, No. 95; the refueling time preference is the time period preferred by the user for refueling.

[0116] The charging business features may include charging frequency, charging duration, charging amount, charging station preference, charging time preference, etc. Among them, the charging frequency is the frequency of vehicle charging by the user on the global travel platform; the charging duration is the average time of each charging; the charging amount is the average cost of each charging; the charging station preference is the location of the charging station commonly used by the user; the charging time preference is the time period preferred by the user for charging.

[0117] The vehicle maintenance business features may include maintenance frequency, maintenance content, maintenance amount, maintenance time preference, and historical maintenance records, etc. Among them, the maintenance frequency is the frequency of vehicle maintenance by the user; the maintenance content is the maintenance items often chosen by the user, such as oil change, tire maintenance, brake inspection, etc.; the maintenance amount is the average cost of each maintenance; the maintenance time preference is the time period preferred by the user for maintenance; the historical maintenance records are the past maintenance records and content of the user.

[0118] The business characteristics of convenience stores can include purchase frequency, purchase amount, purchase types, and shopping time preferences, etc. Among them, the purchase frequency is the frequency of users shopping in convenience stores; the purchase amount is the average amount per shopping; the purchase types are the types of goods that users often buy, such as beverages, snacks, and daily necessities; the shopping time preference is the time period that users prefer to shop.

[0119] The comprehensive behavior characteristics include activity, preference analysis, consumption ability, promotion sensitivity, number of orders, number of payments, and loyalty, etc. Among them, the activity is the overall activity level of users on the global travel platform; the preference analysis is the usage preference of users for each business line; the consumption ability is the total consumption amount and consumption trend of users on each business line of the global travel platform; the promotion sensitivity is the degree of response of users to promotional activities or discounts, such as the usage of coupons and participation in promotional activities, etc.; the number of orders is the number of orders of users on the global travel platform, and the number of payments is the number of times users click the payment button; the loyalty is the usage duration and loyalty index of users on the global travel platform, and the loyalty index can include the renewal rate, repurchase rate, etc.

[0120] The social characteristics can include recommendation behavior and user feedback, etc. The recommendation behavior includes the recommendation behavior of users on the global travel platform, which can include inviting friends, sharing links, etc. The user feedback is the evaluation and feedback of users, which can include comments, ratings, and complaints, etc.

[0121] In the case where the feature is not a specific value, the corresponding feature can be quantified into a specific value for subsequent model prediction.

[0122] Extract the feature sequence of user data under at least one specified dimension, and the specified dimension includes at least one of basic information features, car rental business features, refueling business features, charging business features, maintenance business features, convenience store business features, comprehensive behavior features, and social features. It can simplify data representation and improve the accuracy of the predicted value of consumption willingness.

[0123] Step S204, input the feature sequence into the first model to obtain the predicted value of the consumption willingness of the target driver.

[0124] For details, please refer to Figure 1 Step S104 of the illustrated embodiment, which will not be elaborated here.

[0125] In some optional implementation manners, after step S204, it further includes:

[0126] Step S205, based on the predicted value of the consumption willingness of the target driver, determine the classification result of the target driver.

[0127] Based on the predicted value of the consumption willingness of the target driver, the consumption willingness value interval can be divided, and the target drivers can be classified. In some alternative embodiments, they can be simply classified into three categories: high willingness, medium willingness, and low willingness.

[0128] Step S206: Based on the classification result, determine the push strategy corresponding to the target driver.

[0129] Among them, when the consumption willingness of the target driver is high, high-end automotive products or services can be pushed to him; when the consumption willingness of the target driver is medium, personalized service products can be pushed to him; when the consumption willingness of the target driver is low, some service products with high cost performance can be pushed to him.

[0130] Classify the target drivers according to the predicted value of the consumption willingness of the target driver, so as to determine the push strategy corresponding to the target driver. Corresponding pushes can be made to the target drivers according to the predicted value of the consumption willingness of the target driver, which can optimize the usage experience of the target drivers.

[0131] In some alternative embodiments, some real-time actual data is extracted to cross-validate the output result of the first model to ensure the stability and accuracy of the first model. In some alternative embodiments, new data can be collected at fixed time intervals to evaluate the performance of the first model. The evaluation metrics can include precision, recall, and F1 score, etc. If problems such as data drift or performance degradation are encountered, the first model can be retrained or adjusted according to the actual situation. Among them, whether the performance of the model has degraded can be determined by the evaluation metrics.

[0132] In a practical application, after deploying the first model to the production environment, the real-time data of the current day is input into the model to predict the consumption willingness of drivers in the future time period and store the records. When the data prediction time arrives, the real data is recorded. For the predicted data and the actual data, in case of data drift, new data is collected to retrain the model and adjust the model parameters.

[0133] The consumption willingness evaluation method provided in this embodiment extracts the reference sequences of the sample data in several dimensions, and constructs a Pearson coefficient set based on the reference sequences and the consumption willingness values of the sample drivers, so as to determine the target dimension. To obtain the sample sequence of the sample data in the target dimension. Dimensions with high correlation with the consumption willingness values of the sample drivers can be obtained, thereby improving the model accuracy.

[0134] In this embodiment, a consumption willingness evaluation method is provided. Figure 3 It is the flowchart of the consumption willingness evaluation method according to the embodiment of the present invention, as Figure 3As shown in the figure, first obtain the user's historical behavior data, then based on the historical behavior data and a pre-trained machine learning model, predict the user's consumption willingness in the next stage, and then perform precise push according to the user's consumption willingness.

[0135] Among them, the user's historical behavior data can be collected within a preset time. In a practical application, the preset time is 30 days, that is, the user's behavior data is collected within the preset time of 30 days. The behavior data can include driving habits, consumption records, location information, etc. to form a data set. Each row of data is a record of a certain user's driving habits, consumption records, and location at a certain time.

[0136] The consumption willingness evaluation method provided in this embodiment can predict the consumption behavior of driver users, achieving high accuracy and high efficiency in recommendation. It can not only provide a more personalized and precise user experience, but also effectively reduce the operating costs of enterprises, improve the satisfaction and loyalty of users to the platform. Moreover, it can also provide strong data support for the push activities of automobile enterprises, accurately locate potential user groups, and adopt targeted push strategies. It can improve the order completion rate.

[0137] In this embodiment, a consumption willingness evaluation device is also provided. This device is used to implement the above-mentioned embodiments and preferred implementation manners, and those that have been described will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that can achieve a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware is also possible and contemplated.

[0138] This embodiment provides a consumption willingness evaluation device, as Figure 4 shown, including:

[0139] An acquisition module 401, configured to acquire sample data of sample drivers, and based on the sample data, obtain the consumption willingness values of the sample drivers.

[0140] A construction module 402, configured to extract sample sequences of the sample data in a target dimension, and based on the sample sequences and the consumption willingness values, construct a first model.

[0141] An extraction module 403, configured to acquire user data of target drivers, and extract feature sequences of the user data in at least one specified dimension.

[0142] An output module 404, configured to input the feature sequences into the first model to obtain the consumption willingness prediction values of the target drivers.

[0143] In some alternative embodiments, the specified dimension includes at least one of basic information features, car rental business features, refueling business features, charging business features, maintenance business features, convenience store business features, comprehensive behavior features, and social features.

[0144] In some alternative embodiments, the consumption willingness evaluation device further includes:

[0145] A classification module, configured to determine a classification result of the target driver based on the consumption willingness prediction value of the target driver.

[0146] A push module, configured to determine a push strategy corresponding to the target driver based on the classification result.

[0147] In some alternative embodiments, the construction module 402 includes:

[0148] An extraction unit, configured to extract a reference sequence of the sample data under a plurality of dimensions.

[0149] A calculation unit, configured to calculate the Pearson correlation coefficient between the reference sequence and the consumption willingness value of the sample driver under different dimensions based on the reference sequence and the consumption willingness value of the sample driver.

[0150] A target dimension determination unit, configured to determine a target dimension based on the Pearson correlation coefficient.

[0151] A sample sequence acquisition unit, configured to acquire a sample sequence of the sample data under the target dimension, where the sample sequence is a feature sequence of the sample data under the target dimension.

[0152] In some alternative embodiments, the construction module 402 includes:

[0153] A data preprocessing unit, configured to perform data preprocessing on the sample sequence, where the data preprocessing includes at least one of outlier removal and missing value filling.

[0154] A training unit, configured to input the sample sequence after data preprocessing into an initial model for training to obtain a first model.

[0155] In some alternative embodiments, the data preprocessing unit includes:

[0156] An identification subunit, configured to identify missing values and non-missing values in the sample sequence and obtain the average value of the non-missing values in the sample sequence.

[0157] A filling subunit, configured to fill the average value of the non-missing values into the missing values of the sample sequence.

[0158] In some alternative embodiments, the training unit includes:

[0159] An initial model acquisition subunit, configured to acquire an initial model, where the initial model is an XGBoost model.

[0160] A model parameter determination subunit, configured to determine model parameters of the initial model, where the model parameters include at least one of the maximum depth of the tree, the subsample ratio, and the learning rate.

[0161] A first output result acquisition subunit, configured to input a sample sequence into the initial model and acquire a first output result of the initial model.

[0162] A first training result determination subunit, configured to determine a first training result of the initial model based on the first output result and the consumption willingness value of the sample driver, where the consumption willingness value of the sample driver corresponds to the sample sequence.

[0163] An initial model parameter adjustment subunit, configured to adjust the model parameters of the initial model based on the first training result to obtain a first model.

[0164] In some alternative embodiments, the construction module 402 includes a testing unit, and the testing unit includes:

[0165] A test data acquisition subunit, configured to acquire test data of a test driver and obtain a consumption willingness value of the test driver based on the test data.

[0166] A test sequence acquisition subunit, configured to acquire a test sequence of the test data in a target dimension, input the test sequence into the first model, and acquire a first consumption willingness test value.

[0167] An evaluation subunit, configured to evaluate the first model based on the first consumption willingness test value and the consumption willingness value of the test driver.

[0168] A first model parameter adjustment subunit, configured to adjust the parameters of the first model based on the evaluation result.

[0169] The further function descriptions of the above-mentioned various modules and units are the same as those in the corresponding foregoing embodiments, and will not be elaborated herein.

[0170] The consumption willingness evaluation device in this embodiment is presented in the form of functional units. Here, the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and a memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0171] An embodiment of the present invention further provides a computer device having the above Figure 4 shown consumption willingness evaluation device.

[0172] Please refer to Figure 5 , Figure 5 which is a schematic structural diagram of a computer device provided by an alternative embodiment of the present invention. As shown in Figure 5 , the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Each component communicates with each other using different buses and can be installed on a common motherboard or installed in other ways as needed. The processor can process instructions executed within the computer device, including instructions stored in the memory or on the memory to display graphical information of the GUI on an external input / output device (such as a display device coupled to the interface). In some alternative embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a set of blade servers, or a multi-processor system). Figure 5 In

[0173] , a single processor 10 is taken as an example.

[0174] The processor 10 can be a central processing unit, a network processor, or a combination thereof. Among them, the processor 10 can further include a hardware chip. The above hardware chip can be an application-specific integrated circuit, a programmable logic device, or a combination thereof. The above programmable logic device can be a complex programmable logic device, a field programmable gate array, a general array logic, or any combination thereof.

[0175] The memory 20 stores instructions executable by at least one processor 10, so that at least one processor 10 executes the method shown in the above embodiments.

[0176] The memory 20 can include a storage program area and a storage data area. Among them, the storage program area can store an operating system and application programs required for at least one function; the storage data area can store data created according to the use of the computer device. In addition, the memory 20 can include high-speed random access memory, and can also include non-transitory memory, such as at least one disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some alternative embodiments, the memory 20 can optionally include a memory remotely set relative to the processor 10, and these remote memories can be connected to the computer device through a network. Examples of the above network include but are not limited to the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0176] The memory 20 can include volatile memory, such as random access memory; the memory can also include non-volatile memory, such as flash memory, a hard disk, or a solid-state drive; the memory 20 can also include a combination of the above types of memory.

[0177] The computer device further includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30, and the output device 40 may be connected through a bus or other means. Figure 5 Here, taking the connection through the bus as an example.

[0178] The input device 30 can receive input digital or character information, and generate key signal inputs related to the user settings and function controls of the computer device, such as a touch screen, a keypad, a mouse, a trackpad, a touchpad, a pointing stick, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 may include a display device, an auxiliary lighting device (e.g., an LED), and a haptic feedback device (e.g., a vibration motor), etc. The above display device includes, but is not limited to, a liquid crystal display, a light-emitting diode, a display, and a plasma display. In some alternative embodiments, the display device may be a touch screen.

[0179] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or be implemented as computer code that can be recorded on a storage medium, or be implemented by downloading through a network and originally stored in a remote storage medium or a non-transitory machine-readable storage medium and will be stored in a local storage medium, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory, a random access memory, a flash memory, a hard disk, or a solid-state drive, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by the computer, the processor, or the hardware, the method shown in the above embodiment is implemented.

[0180] A part of the present invention can be applied as a computer program product, such as computer program instructions. When executed by a computer, through the operation of the computer, the method and / or technical solution according to the present invention can be called or provided. Those skilled in the art should be able to understand that the forms of existence of computer program instructions in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to the computer.

[0181] While embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A method for evaluating consumption willingness, characterized in that: The method comprises: Acquire sample data of sample drivers, and obtain consumption willingness values ​​of the sample drivers based on the sample data; Extracting a sample sequence of the sample data under a target dimension, and constructing a first model based on the sample sequence and the consumption willingness value; Acquire user data of a target driver, and extract a feature sequence of the user data in at least one specified dimension; The feature sequence is input into the first model to obtain the consumption intention prediction value of the target driver.

2. The method according to claim 1, characterized in that The designated dimensions include at least one of basic information characteristics, car rental business characteristics, refueling business characteristics, charging business characteristics, maintenance business characteristics, convenience store business characteristics, comprehensive behavior characteristics and social characteristics.

3. The method according to claim 1, characterized in that After obtaining the consumption intention prediction value of the target driver, the method further includes: Determining a classification result of the target driver based on the predicted value of the consumption intention of the target driver; Based on the classification result, a push strategy corresponding to the target driver is determined.

4. The method according to claim 1, characterized in that: The extracting a sample sequence of the sample data under a target dimension includes: Extracting reference sequences of the sample data in several dimensions; Based on the reference sequence and the consumption willingness values ​​of the sample drivers, calculating the Pearson coefficients of the reference sequence and the consumption willingness values ​​of the sample drivers in different dimensions; Based on the Pearson coefficient, determining a target dimension; A sample sequence of the sample data under the target dimension is obtained, wherein the sample sequence is a feature sequence of the sample data under the target dimension.

5. The method according to claim 1, characterized in that: The constructing of the first model comprises: Performing data preprocessing on the sample sequence, wherein the data preprocessing includes at least one of outlier removal and missing value filling; The sample sequence after data preprocessing is input into the initial model for training to obtain the first model.

6. The method according to claim 5, characterized in that When the data preprocessing includes missing value filling, the data preprocessing on the sample sequence includes: Identify missing values ​​and non-missing values ​​in the sample sequence, and obtain an average value of the non-missing values ​​in the sample sequence; The missing values ​​of the sample sequence are filled with the mean values ​​of the non-missing values.

7. The method according to claim 5, characterized in that The sample sequence after data preprocessing is input into the initial model for training to obtain the first model, including: Obtaining an initial model, wherein the initial model is an XGBoost model; Determining model parameters of the initial model, wherein the model parameters include at least one of a maximum tree depth, a subsample ratio, and a learning rate; Inputting the sample sequence into an initial model and obtaining a first output result of the initial model; Determining a first training result of the initial model based on the first output result and the consumption willingness value of the sample driver, wherein the consumption willingness value of the sample driver corresponds to the sample sequence; Based on the first training result, the model parameters of the initial model are adjusted to obtain a first model.

8. The method according to claim 7, characterized in that The method further comprises: Acquire test data of the test driver, and obtain the consumption willingness value of the test driver based on the test data; Obtaining a test sequence of the test data under the target dimension, and inputting the test sequence into the first model to obtain a first consumption willingness test value; evaluating the first model based on the first consumption willingness test value and the consumption willingness value of the test driver; Based on the evaluation results, the parameters of the first model are adjusted.

9. A consumption willingness evaluation device, characterized in that: The device comprises: An acquisition module, used to acquire sample data of a sample driver, and obtain a consumption willingness value of the sample driver based on the sample data; A construction module, used to extract a sample sequence of the sample data under a target dimension, and construct a first model based on the sample sequence and the consumption willingness value; An extraction module, used to obtain user data of a target driver and extract a feature sequence of the user data in at least one specified dimension; An output module is used to input the feature sequence into the first model to obtain the consumption willingness prediction value of the target driver.

10. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the consumer willingness evaluation method according to any one of claims 1 to 8 by executing the computer instructions.

Images