CN111681051B - Purchase intention prediction method and device, storage medium and terminal - Google Patents
Purchase intention prediction method and device, storage medium and terminal Download PDFInfo
- Publication number
- CN111681051B CN111681051B CN202010512538.6A CN202010512538A CN111681051B CN 111681051 B CN111681051 B CN 111681051B CN 202010512538 A CN202010512538 A CN 202010512538A CN 111681051 B CN111681051 B CN 111681051B
- Authority
- CN
- China
- Prior art keywords
- machine learning
- training
- model
- learning model
- samples
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 64
- 230000006399 behavior Effects 0.000 claims abstract description 34
- 238000012549 training Methods 0.000 claims description 95
- 238000010801 machine learning Methods 0.000 claims description 92
- 238000012360 testing method Methods 0.000 claims description 40
- 230000006870 function Effects 0.000 claims description 28
- 238000006243 chemical reaction Methods 0.000 claims description 15
- 230000008569 process Effects 0.000 claims description 11
- 238000012216 screening Methods 0.000 claims description 4
- 238000007477 logistic regression Methods 0.000 description 15
- 238000012545 processing Methods 0.000 description 7
- 230000009471 action Effects 0.000 description 6
- 230000000694 effects Effects 0.000 description 6
- 230000000875 corresponding effect Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 230000008451 emotion Effects 0.000 description 1
- 230000002996 emotional effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012821 model calculation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0202—Market predictions or forecasting for commercial activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Entrepreneurship & Innovation (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Economics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Game Theory and Decision Science (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application provides a purchase intention prediction method, a device, a storage medium and a terminal, wherein the method is characterized in that target access data capable of representing online matching behaviors of users to be predicted are obtained; extracting a characteristic value of the target access data under a preset characteristic dimension, and inputting the characteristic value of the target access data into a prediction model so that the prediction model classifies the characteristic value of the target access data; and then obtaining the purchase intention degree of the user to be predicted through the prediction model. Therefore, the purchase intention prediction method can simply and effectively mine the potential customers with high purchase intention, so that the subsequent marketing resources can be utilized to the greatest extent, and the sales volume of the automobile under the limited marketing cost is improved.
Description
Technical Field
The application relates to the technical field of automobile internet marketing, in particular to a purchase intention prediction method, a device, a storage medium and a terminal.
Background
The field of automobile internet marketing is an intersecting field of automobile sales and internet sales. On the basis of continuous development of internet technology, on-line customized business of automobiles is generated in order to meet different and different automobile use demands of different users.
The automobile online customization opens a channel for a user to select a configuration list, and the user can go to a 4s store to place an online order or directly place an online order after the internet matching is completed. The process of selecting and matching products is an important channel for a user to know the car, and the user who buys the car is directly determined to be less only by only selecting and matching once.
Therefore, how to find users who really have the intention to purchase from a plurality of users who select and match products becomes a problem to be solved in the present stage.
Disclosure of Invention
In view of the above, the present application provides a purchase intention prediction method, device, storage medium and terminal, which have the following technical scheme:
a method of purchase intent prediction, the method comprising:
acquiring target access data capable of representing online matching behaviors of users to be predicted;
extracting a characteristic value of the target access data under a preset characteristic dimension, and inputting the characteristic value of the target access data into a prediction model so that the prediction model classifies the characteristic value of the target access data, wherein the prediction model is obtained by training a machine learning model by taking access data of online matching behaviors of historical users as samples in advance;
and acquiring the purchase intention degree of the user to be predicted, which is output by the prediction model.
Preferably, the process of training the machine learning model to obtain the prediction model by taking access data of online matching behaviors of the historical user as a sample in advance includes:
taking access data representing online matching behaviors of a historical user as a sample and conversion data representing purchase intention behaviors of the historical user as labels of the sample;
extracting a characteristic value of the sample under the preset characteristic dimension, and inputting the characteristic value and the label of a part of samples used for the training in the sample into a machine learning model so that the machine learning model adjusts model parameters by fitting the sample;
and calculating a loss function value of the machine learning model, and repeatedly inputting the characteristic value and the label of a part of samples used for the current training in the samples into the general machine learning model when the loss function value does not accord with a preset first training ending condition until the loss function value accords with the first training ending condition, and taking the trained machine learning model as a prediction model.
Preferably, the samples include training samples and test samples;
the inputting the characteristic values and the labels of the part of samples used for the training in the sample into the machine learning model comprises the following steps:
inputting the characteristic values and the labels of the part of samples used for the training in the training sample into a machine learning model;
before using the trained machine learning model as the predictive model, the method further includes:
inputting the characteristic values of the test sample into a trained machine learning model so that the trained machine learning model classifies the characteristic values of the test sample;
obtaining the purchase intention outputted by the trained machine learning model and the label of the test sample, calculating the accuracy of the trained machine learning model, and taking the trained machine learning model as a prediction model when the accuracy accords with a preset second training ending condition.
Preferably, before extracting the feature value of the sample in the preset feature dimension, the method further includes:
and screening qualified samples meeting preset forward conversion conditions from the samples.
Preferably, the method further comprises:
and adding the users to be predicted, the purchase intention of which meets the preset high purchase intention condition, into the recommendation form, and outputting the recommendation form.
A purchase intent prediction apparatus, the apparatus comprising: the system comprises a data acquisition module, a prediction module and an intention acquisition module, wherein the prediction module comprises a model training unit;
the model training unit is used for training the machine learning model to obtain a prediction model by taking access data of online matching behaviors of the historical user as a sample in advance;
the data acquisition module is used for acquiring target access data capable of representing online matching behaviors of users to be predicted;
the prediction module is used for extracting the characteristic value of the target access data under the preset characteristic dimension, and inputting the characteristic value of the target access data into the prediction model so that the prediction model classifies the characteristic value of the target access data;
and the intention acquisition module is used for acquiring the purchase intention of the user to be predicted, which is output by the prediction model.
Preferably, the model training unit is specifically configured to:
taking access data representing online matching behaviors of a historical user as a sample and conversion data representing purchase intention behaviors of the historical user as labels of the sample;
extracting a characteristic value of the sample under the preset characteristic dimension, and inputting the characteristic value and the label of a part of samples used for the training in the sample into a machine learning model so that the machine learning model adjusts model parameters by fitting the sample;
and calculating a loss function value of the machine learning model, and repeatedly inputting the characteristic value and the label of a part of samples used for the current training in the samples into the general machine learning model when the loss function value does not accord with a preset first training ending condition until the loss function value accords with the first training ending condition, and taking the trained machine learning model as a prediction model.
Preferably, the samples include training samples and test samples;
the model training unit is used for inputting the characteristic values and the labels of the partial samples used for the training in the samples into a general machine learning model, and is specifically used for:
inputting the characteristic values and the labels of the part of the training samples for the training to a general machine learning model;
model training unit, still be used for:
inputting the characteristic values of the test sample into a trained machine learning model so that the trained machine learning model classifies the characteristic values of the test sample;
obtaining the purchase intention outputted by the trained machine learning model and the label of the test sample, calculating the accuracy of the trained machine learning model, and taking the trained machine learning model as a prediction model when the accuracy accords with a preset second training ending condition.
A computer readable storage medium having stored thereon computer instructions which, when executed, perform the steps of any of the purchase intent prediction methods.
A terminal comprising a memory and a processor, the memory having stored thereon computer instructions executable on the processor, the processor executing the steps of any of the purchase intention prediction methods when the computer instructions are executed.
The purchase intention prediction method, the device, the storage medium and the terminal provided by the application can acquire the target access data when the user to be predicted is selected online, and the purchase intention of the user to be predicted is predicted by processing the characteristic value of the target access data under the preset characteristic dimension by using the prediction model. Therefore, the purchase intention prediction method can simply and effectively mine the potential customers with high purchase intention, so that the subsequent marketing resources can be utilized to the greatest extent, and the sales volume of the automobile under the limited marketing cost is improved.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings that are required to be used in the embodiments or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are only embodiments of the present application, and that other drawings can be obtained according to the provided drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flowchart of a method for predicting intent to purchase according to an embodiment of the present application;
FIG. 2 is a flowchart of a part of a method for predicting purchase intention according to an embodiment of the application;
FIG. 3 is a schematic illustration of an interpretation of a predictive model using a SHAP interpreter, in accordance with an embodiment of the present application;
FIG. 4 is a schematic diagram of the influence result of a prediction model according to an embodiment of the present application;
fig. 5 is a schematic structural diagram of a purchase intention prediction apparatus according to an embodiment of the present application.
Detailed Description
The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
The existing scheme of internet automobile marketing is to use advertisement delivery to attract users to leave information such as contact information (simply called fund) through information such as activities, preferential and the like. Then, customer service personnel call one by one to follow up, confirm the interesting vehicle type and the vehicle purchasing requirement, and recommend the user to a nearby 4s store or dealer. Because the online customized business of the automobile is still in the early development stage, the marketing scheme specially aiming at the business is similar to the traditional automobile internet marketing scheme, or the online activities attract vermicelli registration accounts and login platforms, and then call back is carried out on the resident users, or the call back is selectively carried out according to the online activity degree of the users.
Of course, predictions of on-line custom made purchase intent of a vehicle based on supervised learning are temporarily not studied. In internet marketing in other fields, there is a method of determining emotional tendency by using history transaction data, history transaction information of a user, or a language in a user communication record, though there is a method of using machine learning. The traditional internet marketing is mainly recommended, namely, the commodity suitable for the user is recommended from innumerable commodities, and the application can find the user really having purchase intention from a plurality of users who select the product.
The embodiment of the application provides a purchase intention prediction method, a flow chart of the method is shown in fig. 1, and the method comprises the following steps:
s10, acquiring target access data capable of representing online matching behaviors of users to be predicted.
In the embodiment of the application, most users of selected products can consider whether to purchase automobiles or not through multiple clicking, webpage viewing and other actions on selected webpages. Thus, the internet web page can transmit back some actions of clicking the web page by the user through the embedded point information, and the actions are stored in a server or a storage medium. In summary, the automobile manufacturer or sales organization can legally obtain the access behavior of the user in the matching process.
S20, extracting characteristic values of the target access data under the preset characteristic dimension, and inputting the characteristic values of the target access data into a prediction model so that the prediction model classifies the characteristic values of the target access data, wherein the prediction model is obtained by training a machine learning model by taking access data of on-line matching behaviors of historical users as samples in advance.
In the embodiment of the application, at least one field can be selected from the fields corresponding to the online matching behaviors of the user as the preset feature dimension, so that the data content of the target access data under the preset feature dimension is extracted as the feature value.
Most of the prior machine learning uses in internet marketing focusing on product recommendation of an e-commerce platform, namely products for judging possible needs of users by using historical transaction data, historical transaction information of users, language judgment emotion tendency in user communication records and the like, or one or more products for matching the needs of users in a vertical marketing link. But the method has no practical value for the vertical marketing of the automobile internet of the on-line customized business of the automobile, because the related product brands and the automobile system are fixed, the user can select configuration in the automobile system according to own needs, and after the configuration is finished, the user can place orders on the internet or go to off-line dealers for further negotiations, so that the user needs to be classified in layers.
Therefore, the embodiment of the application utilizes the access data of the historical user to generate the prediction model to predict the purchase intention of the new user, so that the follow-up marketing can be accurately put in.
It should be noted that the preset feature dimension used for extracting the target access data is the same as the preset feature dimension used when the prediction model is trained in advance. Moreover, the construction of the preset feature dimension and the time window for acquiring the target access data can be finished in a self-defined mode according to different scenes and service operation periods.
It should be noted that the general machine learning model may be any machine learning model capable of completing a classification task, and may be any one of a logistic regression model, a random forest model, an SVM model, and a LightGBM model.
In the specific implementation process, in step S20, the process of training the general machine learning model to obtain the prediction model by taking the access data of the online matching behavior of the historical user as a sample in advance includes the following steps, and the method flowchart is shown in fig. 2:
s201, taking access data representing online matching behaviors of the historical user as a sample and conversion data representing purchase intention behaviors of the historical user as a label of the sample.
In the embodiment of the application, some important feature dimensions which can summarize the related user behaviors and operation preferences are selected based on the user's matching behaviors, and then the access data of the historical user is calculated by taking the user as a unit, so that the conversion data of the user, such as information of reserved materials, card construction, sales and the like, is tracked in an Internet marketing system. Therefore, in the embodiment of the application, conversion data such as "funding," "card creation," "sales," and the like can be used as a label of a sample in units of users, and the label can represent the purchase intention degree of the users, namely the degree of the purchase intention. For example, the label of the conversion data of "fund", "build card", "sales", etc. may be set to a positive number whose magnitude indicates the degree of purchase intention with, and the label of the sample without the conversion data may be set to 0 indicating no purchase intention.
In practical applications, the samples and labels of the samples may be stored and used in the form of structured data.
S202, extracting characteristic values of the samples under preset characteristic dimensions, and inputting the characteristic values and labels of part of the samples used for the training in the samples into a machine learning model so that the machine learning model adjusts model parameters by fitting the samples.
In the embodiment of the application, taking a machine learning model as a logistic regression model as an example, selecting part of samples for each training from the samples in turn, and inputting the characteristic values and the labels of the part of samples into the logistic regression model during each training. The logistic regression model after training is finished can be used as a prediction model to judge the purchase intention degree of the user, and then marketing staff can accurately throw the user according to the release vector of the purchase intention degree.
The logistic regression model is described below:
establishing a logistic regression model shown in the following formula (1)
P(y＝1|x；θ)＝h θ (x)＝g(θ T x) (1)
Wherein,,
wherein x is i For feature input in the model, y is a label term, y=1 means that the user is payed off line or is blocked or sold, θ i Post-model fitting x i Is a coefficient of (a).
In order to prevent the overfitting of the model, the embodiment of the application introduces a regular term containing lambda into the loss function J (theta) of the logistic regression model, and the loss function J (theta) is shown in the following formula (2):
where m is the sample data size, J is the number of features in the logistic regression function, and we obtain the required parameter θ when J (θ) is minimum i And a super parameter lambda.
Further, the embodiment of the application can alsoFor the parameter theta by further adopting a cross check mode i And a super parameter lambda. When the super parameter lambda is smaller, the model is easy to be over-fitted; when the super parameter lambda is larger, the model overfitting degree can be reduced, and meanwhile, the classification error rate is increased to a certain degree.
In addition, in other embodiments, before extracting the feature value of the sample in the preset feature dimension, a qualified sample meeting the preset forward conversion condition may be further selected from the samples. And taking the qualified sample as a basis for obtaining a prediction model through subsequent training.
Specifically, samples which do not meet the requirements can be removed from the practical point of view, and if the corresponding date of the card build information of the user is before the user accesses the online matching platform, the user is the user with high intention before the user accesses the online matching platform, and the user cannot be taken as a qualified sample of forward conversion.
S203, calculating a loss function value of the machine learning model, and when the loss function value does not meet a preset first training ending condition, returning to execute the step S202 of inputting the characteristic value and the label of a part of samples used for the training into the machine learning model until the loss function value meets the first training ending condition, and taking the trained machine learning model as a prediction model.
In the embodiment of the application, if the function value of the loss function is larger than the preset loss function threshold after the training is finished, the machine learning model is continuously trained, and the next training is performed until the loss function value of the machine learning model after the training is finished is not larger than the loss function threshold.
In other embodiments, the samples may be divided into training samples and test samples.
At this time, the step S202 of inputting the feature value and the label of the partial sample for the current training in the sample into the general machine learning model includes:
and inputting the characteristic values and the labels of the part of samples used for the training in the training sample into a general machine learning model.
Before the "taking the trained machine learning model as the prediction model" in step S203, the method further includes the following steps:
inputting the characteristic values of the test sample into a trained machine learning model so that the trained machine learning model classifies the characteristic values of the test sample;
obtaining purchase intention corresponding to the test sample output by the trained machine learning model and the label of the test sample, calculating the accuracy of the trained machine learning model, and taking the trained machine learning model as a prediction model when the accuracy accords with a preset second training ending condition.
In the embodiment of the application, the samples are divided into training samples and test samples, the training samples are used for training a machine learning model to obtain a prediction model, and the test samples are used for verifying the prediction correctness of the prediction model. Specifically, if the purchase intention of the prediction model for one test sample can represent the label of the test sample, the prediction model is accurate for the test sample, otherwise, the prediction model is inaccurate. The prediction model can be represented by the ratio of the number of samples predicted by the prediction model to the total number of samples in the test sample.
For easy understanding of the present application, the following uses an application scenario as an example to describe the offline training logistic regression model and the online prediction process in the present application:
1. and training a logistic regression model offline.
Sample and tag input: the online matching user behavior log records of the Internet are pulled through online buried points to obtain 1800 tens of thousands of online matching user behavior log records, 20 fields are extracted according to service scenes and application targets, off-line reserved/built card/sales information is obtained, and the user is used as id for matching, so that 20646 pieces of data are obtained.
Sample processing: 12 fields are selected from 20 fields as characteristic inputs of the model according to the model requirement, and 10713 pieces of data which are worth predicting are selected from 20646 pieces of data.
Decomposing a training sample and a test sample: 10713 pieces of data are divided into two parts of a training sample and a test sample, wherein the test sample accounts for 20% of the total number, and meanwhile, the proportion of the positive samples of the stay/build card/sell in the training sample and the test sample is consistent by taking off-line stay/build card/sell information as a layering reference.
Feature standardization: and normalizing the 12 input features, and eliminating the influence on the model caused by the size difference between the features of different dimensions so as to carry out the super-parameter adjustment of the model.
Model fitting and parameter adjustment: generating a logistic regression model, and introducing a superparameter C, whereinFitting the training set by ten-fold cross test, searching the optimal super parameter from 0.1 to 100 for C by grid search, and finally obtaining theta 0 Is-0.54, theta 1 To theta 12 1.03e-03, -6.02e-04,1.01e-01,7.26e-05, -1.23e-01, -8.58e-04, -8.26e-04,3.99e-03, -2.45e-02-3.78e-05,3.63e-01,5.22e-03, respectively
Model test results: predictions are made on the test set (2143) using the generated model, correctly predict result 1501, correct rate 70%, where predictions are result 284 of the build. The model is saved for later recall and prediction of new data.
Model interpretation of SHAP: the model was interpreted in the test sample using the SHAP interpreter based on the generated logistic regression model, the results of which are shown in fig. 3. It can be seen that ft3 and ft1 features have more pronounced positive predictive effects on the model, while ft5 and ft6 have more pronounced negative predictive effects on the model. This also coincides with the coefficients fitted by the model. Where ft3 has the strongest influence on the model.
Ft3 was taken as a study object, and its effect on the model was examined as shown in fig. 4. It can be seen that the positive predictive contribution capability of ft3 to the model is linearly and positively correlated with its own value, which is related to the model form of the logistic regression itself.
2. And (5) online prediction.
Target access data input and processing: similar to the process of logistic regression model training, 20551 pieces of data were obtained.
Purchase intent prediction: and calling out the offline trained prediction model, and predicting on the target access data to obtain corresponding purchase intention degree, wherein the high intention degree prediction result is 1272.
The application solves the problem of how to accurately put limited marketing resources on users with high purchase intention degree in an internet marketing system with an on-line automobile customization service. The purchase intention degree prediction method has the advantage that purchase intention degree layering can be performed for a user who does not have deep knowledge before. In addition, the application has another advantage that more user data can be continuously acquired according to accumulation of the data, and then the iterative model is retrained according to the data, so that the accuracy of the prediction model can be continuously improved along with the increase of samples. The purchase intention degree of the user is layered to automatically generate the result through the logistic regression model calculation, so that manual screening is not needed, and a large amount of human resources are saved.
S30, obtaining the purchase intention of the user to be predicted, which is output by the prediction model.
In the embodiment of the application, the prediction probability output by the prediction model is the purchase intention degree of the user to be predicted, and the purchase intention degree can represent the degree of the purchase intention of the user to be predicted. And if the preset high purchase intention condition is met, for example, the purchase intention is higher than a preset intention threshold, the user to be predicted can be considered as a high intention user. Of course, the intent range can be set for different intent types, namely, fund remaining, card building and sales, so that accurate delivery is realized.
In other embodiments, the user to be predicted whose purchase intention meets the preset high purchase intention condition may be further added to the recommendation form and output.
Specifically, whether the user to be predicted is in the recommended form obtained by the previous prediction can be judged first; if so, indicating that the user to be predicted has recommended, and directly ending; if not, in order to reduce the workload of the marketing part, the embodiment of the application adds the user with high purchase intention to the recommendation form and pushes the user with high purchase intention to the marketing part, wherein the recommendation form at least comprises the identity information and the purchase intention of the user to be predicted, and in addition, the users can be further sequenced according to the order of the purchase intention from high to low, and the marketing department can accurately push the users according to the requirements and the service scene.
The purchase intention prediction method provided by the embodiment of the application can acquire the target access data when the user to be predicted is selected online, and the prediction model is used for processing the characteristic value of the target access data under the preset characteristic dimension to predict the purchase intention of the user to be predicted. Based on the method and the device, the high-intention potential passengers can be simply and effectively mined, so that the subsequent marketing resources can be utilized to the greatest extent, and the sales of the automobiles under the limited marketing cost can be improved.
Based on the purchase intention prediction method provided in the above embodiment, the embodiment of the present application further provides an apparatus for executing the purchase intention prediction method, where a schematic structural diagram of the apparatus is shown in fig. 5, and the apparatus includes: the data acquisition module 10, the prediction module 20 and the intention acquisition module 30, wherein the prediction module comprises a model training unit 201;
the model training unit 201 is configured to train the machine learning model by taking access data of online matching behaviors of the historical user as a sample in advance to obtain a prediction model;
the data acquisition module 10 is used for acquiring target access data capable of representing online matching behaviors of a user to be predicted;
the prediction module 20 is configured to extract a feature value of the target access data in a preset feature dimension, and input the feature value of the target access data into the prediction model, so that the prediction model classifies the feature value of the target access data;
and the intention acquisition module 30 is used for acquiring the purchase intention of the user to be predicted, which is output by the prediction model.
Optionally, the model training unit 201 is specifically configured to:
taking access data representing the online matching behavior of the historical user as a sample and conversion data representing the purchase intention behavior of the historical user as a label of the sample;
extracting a characteristic value of a sample under a preset characteristic dimension, and inputting the characteristic value and a label of a part of samples used for the training in the sample into a machine learning model so that the machine learning model adjusts model parameters by fitting the sample;
and calculating a loss function value of the machine learning model, and repeatedly inputting the characteristic value and the label of a part of samples used for the training in the samples into the machine learning model when the loss function value does not meet a preset first training ending condition until the loss function value meets the first training ending condition, and taking the trained machine learning model as a prediction model.
Optionally, the samples include training samples and test samples;
the model training unit 201, configured to input the feature values and the labels of the partial samples for the current training in the samples into the general machine learning model, is specifically configured to:
inputting the characteristic values and the labels of part of samples used for the training in the training sample into a general machine learning model;
model training unit 201 is further configured to:
inputting the characteristic values of the test sample into a trained machine learning model so that the trained machine learning model classifies the characteristic values of the test sample; and calculating the accuracy of the trained machine learning model based on the purchase intention corresponding to the test sample output by the trained machine learning model and the label of the test sample, and executing the trained machine learning model as a prediction model when the accuracy meets a preset second training ending condition.
Optionally, the model training unit 201 is further configured to:
and screening qualified samples meeting preset forward conversion conditions from the samples.
Optionally, the intent acquisition module 30 is further configured to:
and adding the users to be predicted, the purchase intention of which meets the preset high purchase intention condition, into the recommendation form, and outputting the recommendation form.
The purchase intention prediction device provided by the embodiment of the application can acquire the target access data when the user to be predicted is selected online, and the prediction model is used for processing the characteristic value of the target access data under the preset characteristic dimension to predict the purchase intention of the user to be predicted. Based on the method and the device, the high-intention potential passengers can be simply and effectively mined, so that the subsequent marketing resources can be utilized to the greatest extent, and the sales of the automobiles under the limited marketing cost can be improved.
The embodiment of the application also provides a computer readable storage medium, wherein the computer instructions are stored on the computer readable storage medium, and the computer instructions execute the steps of the purchase intention prediction method when running. The computer readable storage medium may include, for example, a non-volatile memory (non-volatile) or a non-transitory memory (non-transitory) and may also include an optical disc, a mechanical hard disc, a solid state hard disc, and the like.
The embodiment of the application also provides a terminal, which comprises a memory and a processor, wherein the memory stores computer instructions capable of running on the processor, and the processor executes the steps of the purchase intention prediction method when running the computer instructions. The terminal comprises, but is not limited to, a mobile phone, a computer, a tablet personal computer and other terminal equipment.
The processor may be a central processing unit (Central Processing Unit, CPU), application-specific integrated circuit (ASIC), digital Signal Processor (DSP), application-specific integrated circuit (ASIC), off-the-shelf programmable gate array (FPGA), or other programmable logic device, or the like.
For example, the computer instructions are stored in the memory, and when the computer instructions in the memory are executed by the processor, all or part of the functions of the purchase intention prediction method in the above embodiment can be implemented. In addition, when all or part of the functions in the above embodiments are implemented by means of computer instructions, the computer instructions may be stored in a storage medium such as a server, another computer, a magnetic disk, an optical disk, a flash disk, or a removable hard disk, and the computer instructions in the memory may be executed by a processor by downloading or copying the computer instructions to a memory of the local terminal, or by updating a version of the system of the local terminal.
The method, the device, the storage medium and the terminal for predicting the purchase intention provided by the application are described in detail, and specific examples are applied to the description of the principle and the implementation mode of the application, and the description of the examples is only used for helping to understand the method and the core idea of the application; meanwhile, as those skilled in the art will have variations in the specific embodiments and application scope in accordance with the ideas of the present application, the present description should not be construed as limiting the present application in view of the above.
It should be noted that, in the present specification, each embodiment is described in a progressive manner, and each embodiment is mainly described as different from other embodiments, and identical and similar parts between the embodiments are all enough to be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant points refer to the description of the method section.
It is further noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include, or is intended to include, elements inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (8)
1. A method of predicting intent to purchase, the method comprising:
acquiring target access data capable of representing online matching behaviors of users to be predicted;
extracting a characteristic value of the target access data under a preset characteristic dimension, and inputting the characteristic value of the target access data into a prediction model so that the prediction model classifies the characteristic value of the target access data, wherein the prediction model is obtained by training a machine learning model by taking access data of online matching behaviors of historical users as samples in advance;
acquiring the purchase intention of the user to be predicted, which is output by the prediction model;
the process of training the machine learning model to obtain the prediction model by taking access data of the online matching behaviors of the historical user as a sample in advance comprises the following steps:
taking access data representing online matching behaviors of a historical user as a sample and conversion data representing purchase intention behaviors of the historical user as labels of the sample;
extracting a characteristic value of the sample under the preset characteristic dimension, and inputting the characteristic value and the label of a part of samples used for the training in the sample into a machine learning model so that the machine learning model adjusts model parameters by fitting the sample;
and calculating a loss function value of the machine learning model, and repeatedly inputting the characteristic value and the label of a part of samples used for the current training in the samples into the general machine learning model when the loss function value does not accord with a preset first training ending condition until the loss function value accords with the first training ending condition, and taking the trained machine learning model as a prediction model.
2. The method of claim 1, wherein the samples comprise training samples and test samples;
inputting the characteristic values and the labels of the partial samples used for the training in the samples into a general machine learning model, wherein the method comprises the following steps:
inputting the characteristic values and the labels of the part of samples used for the training in the training sample into a machine learning model;
before using the trained machine learning model as the predictive model, the method further includes:
inputting the characteristic values of the test sample into a trained machine learning model so that the trained machine learning model classifies the characteristic values of the test sample;
obtaining the purchase intention outputted by the trained machine learning model and the label of the test sample, calculating the accuracy of the trained machine learning model, and taking the trained machine learning model as a prediction model when the accuracy accords with a preset second training ending condition.
3. The method of claim 1, wherein prior to extracting the feature values of the sample in the predetermined feature dimension, the method further comprises:
and screening qualified samples meeting preset forward conversion conditions from the samples.
4. The method according to claim 1, wherein the method further comprises:
and adding the users to be predicted, the purchase intention of which meets the preset high purchase intention condition, into the recommendation form, and outputting the recommendation form.
5. A purchase intent prediction apparatus, the apparatus comprising: the system comprises a data acquisition module, a prediction module and an intention acquisition module, wherein the prediction module comprises a model training unit;
the model training unit is used for training the machine learning model to obtain a prediction model by taking access data of online matching behaviors of the historical user as a sample in advance;
the data acquisition module is used for acquiring target access data capable of representing online matching behaviors of users to be predicted;
the prediction module is used for extracting the characteristic value of the target access data under the preset characteristic dimension, and inputting the characteristic value of the target access data into the prediction model so that the prediction model classifies the characteristic value of the target access data;
the intention acquisition module is used for acquiring the purchase intention of the user to be predicted, which is output by the prediction model;
the model training unit is specifically configured to:
taking access data representing online matching behaviors of a historical user as a sample and conversion data representing purchase intention behaviors of the historical user as labels of the sample;
extracting a characteristic value of the sample under the preset characteristic dimension, and inputting the characteristic value and the label of a part of samples used for the training in the sample into a machine learning model so that the machine learning model adjusts model parameters by fitting the sample;
and calculating a loss function value of the machine learning model, and repeatedly inputting the characteristic value and the label of a part of samples used for the current training in the samples into the general machine learning model when the loss function value does not accord with a preset first training ending condition until the loss function value accords with the first training ending condition, and taking the trained machine learning model as a prediction model.
6. The apparatus of claim 5, wherein the samples comprise training samples and test samples;
the model training unit is used for inputting the characteristic values and the labels of the partial samples used for the training in the samples into a general machine learning model, and is specifically used for:
inputting the characteristic values and the labels of the part of the training samples for the training to a general machine learning model;
model training unit, still be used for:
inputting the characteristic values of the test sample into a trained machine learning model so that the trained machine learning model classifies the characteristic values of the test sample;
obtaining the purchase intention outputted by the trained machine learning model and the label of the test sample, calculating the accuracy of the trained machine learning model, and taking the trained machine learning model as a prediction model when the accuracy accords with a preset second training ending condition.
7. A computer readable storage medium having stored thereon computer instructions which, when run, perform the steps of the purchase intent prediction method as claimed in any of claims 1 to 4.
8. A terminal comprising a memory and a processor, the memory having stored thereon computer instructions executable on the processor, wherein the processor, when executing the computer instructions, performs the steps of the purchase intention prediction method of any one of claims 1 to 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010512538.6A CN111681051B (en) | 2020-06-08 | 2020-06-08 | Purchase intention prediction method and device, storage medium and terminal |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010512538.6A CN111681051B (en) | 2020-06-08 | 2020-06-08 | Purchase intention prediction method and device, storage medium and terminal |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111681051A CN111681051A (en) | 2020-09-18 |
| CN111681051B true CN111681051B (en) | 2023-09-26 |
Family
ID=72454942
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010512538.6A Active CN111681051B (en) | 2020-06-08 | 2020-06-08 | Purchase intention prediction method and device, storage medium and terminal |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111681051B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112258242B (en) * | 2020-11-02 | 2024-06-18 | 上海汽车集团股份有限公司 | Form configuration item data pushing method and device |
| CN112884449B (en) * | 2021-03-12 | 2024-05-14 | 北京乐学帮网络技术有限公司 | User guiding method, device, computer equipment and storage medium |
| CN113360845A (en) * | 2021-05-25 | 2021-09-07 | 浙江大搜车软件技术有限公司 | Vehicle source transaction probability prediction method and device, electronic device and storage medium |
| CN113190599B (en) * | 2021-06-30 | 2021-09-28 | 平安科技（深圳）有限公司 | Processing method, device and equipment for application user behavior data and storage medium |
| CN113763032B (en) * | 2021-08-03 | 2023-08-04 | 北京光速斑马数据科技有限公司 | Commodity purchase intention recognition method and device |
| CN113807921B (en) * | 2021-09-17 | 2023-11-24 | 深圳市数聚湾区大数据研究院 | Data commodity recommendation method and device, electronic equipment and computer readable storage medium |
| CN115309737A (en) * | 2022-10-11 | 2022-11-08 | 深圳市明源云客电子商务有限公司 | Visitor intention analysis method and system, terminal device and readable storage medium |
| CN117217852B (en) * | 2023-08-03 | 2024-02-27 | 广州兴趣岛信息科技有限公司 | Behavior recognition-based purchase willingness prediction method and device |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105389639A (en) * | 2015-12-15 | 2016-03-09 | 上海汽车集团股份有限公司 | Logistics transportation route planning method, device and system based on machine learning |
| CN105488697A (en) * | 2015-12-09 | 2016-04-13 | 焦点科技股份有限公司 | Potential customer mining method based on customer behavior characteristics |
| JP2016076109A (en) * | 2014-10-07 | 2016-05-12 | 国立大学法人九州工業大学 | Device and method for predicting customers's purchase decision |
| CN107993088A (en) * | 2017-11-20 | 2018-05-04 | 北京三快在线科技有限公司 | A kind of Buying Cycle Forecasting Methodology and device, electronic equipment |
| CN109741112A (en) * | 2019-01-10 | 2019-05-10 | 博拉网络股份有限公司 | A kind of user's purchase intention prediction technique based on mobile big data |
| TWM589312U (en) * | 2019-10-08 | 2020-01-11 | 富邦人壽保險股份有限公司 | Product purchase evaluation system based on user web browsing history |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9275342B2 (en) * | 2012-04-09 | 2016-03-01 | 24/7 Customer, Inc. | Method and apparatus for intent modeling and prediction |
| US20150134401A1 (en) * | 2013-11-09 | 2015-05-14 | Carsten Heuer | In-memory end-to-end process of predictive analytics |
| US9910930B2 (en) * | 2014-12-31 | 2018-03-06 | TCL Research America Inc. | Scalable user intent mining using a multimodal restricted boltzmann machine |
-
2020
- 2020-06-08 CN CN202010512538.6A patent/CN111681051B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2016076109A (en) * | 2014-10-07 | 2016-05-12 | 国立大学法人九州工業大学 | Device and method for predicting customers's purchase decision |
| CN105488697A (en) * | 2015-12-09 | 2016-04-13 | 焦点科技股份有限公司 | Potential customer mining method based on customer behavior characteristics |
| CN105389639A (en) * | 2015-12-15 | 2016-03-09 | 上海汽车集团股份有限公司 | Logistics transportation route planning method, device and system based on machine learning |
| CN107993088A (en) * | 2017-11-20 | 2018-05-04 | 北京三快在线科技有限公司 | A kind of Buying Cycle Forecasting Methodology and device, electronic equipment |
| CN109741112A (en) * | 2019-01-10 | 2019-05-10 | 博拉网络股份有限公司 | A kind of user's purchase intention prediction technique based on mobile big data |
| TWM589312U (en) * | 2019-10-08 | 2020-01-11 | 富邦人壽保險股份有限公司 | Product purchase evaluation system based on user web browsing history |
Non-Patent Citations (1)
| Title |
|---|
| 张李义 ; 李一然 ; 文璇 ; .新消费者重复购买意向预测研究.数据分析与知识发现.2018,(第11期),全文. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111681051A (en) | 2020-09-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111681051B (en) | Purchase intention prediction method and device, storage medium and terminal | |
| CN110413877B (en) | Resource recommendation method and device and electronic equipment | |
| US20210027357A1 (en) | Systems and methods for credit card selection based on a consumer's personal spending | |
| CN107016026B (en) | User tag determination method, information push method, user tag determination device, information push device | |
| CN109582876B (en) | Tourist industry user portrait construction method and device and computer equipment | |
| CN110135901A (en) | A kind of enterprise customer draws a portrait construction method, system, medium and electronic equipment | |
| US20150213111A1 (en) | Obtaining social relationship type of network subjects | |
| CN110008973B (en) | Model training method, method and device for determining target user based on model | |
| CN110689164B (en) | Prediction method and system for user offer behavior | |
| CN112365283B (en) | Coupon issuing method and device, terminal equipment and storage medium | |
| CN111061979B (en) | User tag pushing method and device, electronic equipment and medium | |
| CN112380449B (en) | Information recommendation method, model training method and related device | |
| CN109118316B (en) | Method and device for identifying authenticity of online shop | |
| CN112001754A (en) | User portrait generation method, device, equipment and computer readable medium | |
| CN111582538A (en) | Community value prediction method and system based on graph neural network | |
| CN111666275A (en) | Data processing method and device, electronic equipment and storage medium | |
| CN110555749A (en) | credit behavior prediction method and device based on neural network | |
| CN113034168B (en) | Content item delivery method, device, computer equipment and storage medium | |
| CN113780479A (en) | Periodic prediction model training method and device, and periodic prediction method and equipment | |
| CN112200215B (en) | Label feature extraction method and device, storage medium and electronic equipment | |
| KR102086996B1 (en) | Platform system for brand management using responses of consumers, method for product planning using the same and computer program for the same | |
| CN110889716A (en) | Method and device for identifying potential registered user | |
| CN110738529A (en) | User diffusion method and device, readable storage medium and electronic equipment | |
| CN116010707A (en) | Commodity price anomaly identification method, device, equipment and storage medium | |
| CN115496420A (en) | New user quality evaluation method and device, computer equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |