CN113343087A - Method and system for acquiring marketing user - Google Patents
Method and system for acquiring marketing user Download PDFInfo
- Publication number
- CN113343087A CN113343087A CN202110641191.XA CN202110641191A CN113343087A CN 113343087 A CN113343087 A CN 113343087A CN 202110641191 A CN202110641191 A CN 202110641191A CN 113343087 A CN113343087 A CN 113343087A
- Authority
- CN
- China
- Prior art keywords
- user data
- marketing
- users
- target domain
- user
- 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.)
- Withdrawn
Links
- 238000000034 method Methods 0.000 title claims abstract description 65
- 238000009826 distribution Methods 0.000 claims abstract description 33
- 238000012216 screening Methods 0.000 claims abstract description 29
- 238000013508 migration Methods 0.000 claims abstract description 21
- 230000005012 migration Effects 0.000 claims abstract description 21
- 238000010801 machine learning Methods 0.000 claims abstract description 14
- 230000003044 adaptive effect Effects 0.000 claims description 13
- 230000008569 process Effects 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 6
- 238000013500 data storage Methods 0.000 claims description 2
- 238000013526 transfer learning Methods 0.000 description 19
- 230000000694 effects Effects 0.000 description 7
- 230000006978 adaptation Effects 0.000 description 6
- 238000012549 training Methods 0.000 description 6
- 230000009286 beneficial effect Effects 0.000 description 4
- 238000012552 review Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 239000002699 waste material Substances 0.000 description 2
- 238000004891 communication Methods 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000013486 operation strategy Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Medical Informatics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method and a system for acquiring marketing users, which are used for acquiring user data similar to a new product during product marketing and storing the user data into a source domain, screening out user data of potential users of the new product from a user data pool and storing the user data into a first target domain, inputting the user data of the source domain and the first target domain into a plurality of data distribution self-adaptive migration learning models, adding a prediction label to the user data in the first target domain and marketing high-quality users. The original method of simply relying on artificial experience is converted into the method of combining machine learning, marketing customer groups are selected intelligently, a machine learning model is perfected in the early stage of new product marketing, the problems that no data exists in the middle stage before a new product is on line, users cannot be analyzed, and the conventional machine learning model is unavailable are solved, and the labor cost and the marketing channel cost are saved.
Description
Technical Field
The invention relates to the technical field of transfer learning, in particular to a method and a system for acquiring marketing users.
Background
The new process is the process of pulling a new user, and the new user is the basis for development and stability of each enterprise, so the process of pulling a new link also plays a very important role in the whole marketing scene. When a new product is just on line, operators often use personal experience to select marketing people when no data is available and data analysis cannot be used as a reference, and the experience of people is limited and not comprehensive enough, so that the defect is more and more obvious, and the method has the concrete expression that the new product is on line too fast and the operators cannot make an effective scheme in time; the operation cost is high, the operation cost is huge and the profit is insufficient due to the fact that the selected passenger groups are not fine enough; the operation scheme is single, and the operation can all give priority to the active customer group when each product is on line, neglects other customer groups with huge potential and larger user base number on the one hand, and on the other hand, a plurality of operation scenes operate the customer groups to repeatedly cause resource waste or resource contention.
A common marketing pull adopts a scoring card scoring system and a tree model supervised learning algorithm to assist in selecting a marketing scene, and the application premises are that after enough samples are accumulated, a model system is established based on enough samples, and the cold start problem of a product just on line and the situation of few samples in the early stage of on line are ignored.
In the actual updating link, based on massive user scale, the marketing cost and the consideration of customer experience are combined, and only very limited users can be subjected to related marketing touch through a short message channel. Therefore, through innovative technical means, the demand user group corresponding to the product is accurately depicted and grasped, and the problems of improving marketing efficiency and reducing operation cost become urgent to be solved.
Disclosure of Invention
The invention aims to provide a method and a system for acquiring marketing users, which are based on a technical framework of transfer learning, utilize a small amount of manual participation, combine the characteristics of accurate transfer learning scheme, high efficiency and iteration in machine learning, solve the problems of the prior new product in the early stage of online operation and the early and middle operation work of accumulating a small amount of samples, and select high-quality users for marketing.
In order to achieve the above purpose, the invention provides the following technical scheme:
a method for obtaining marketing users, comprising:
acquiring user data similar to a new product during product marketing and storing the user data in a source domain, wherein the user data in the source domain carries a tag used for identifying the purchase/non-purchase of a user;
screening out user data of potential users of the new product from a user data pool and storing the user data in a first target domain, wherein the user data in the first target domain does not carry the label;
inputting the user data of the source domain and the first target domain into a plurality of data distribution self-adaptive migration learning models, adding prediction labels to all/part of the user data in the first target domain, and outputting and storing the user data in a first sample set;
obtaining high-quality users from the first sample set and marketing, wherein the high-quality users are as follows: and carrying the user corresponding to the user data of the label for identifying the user to purchase.
Specifically, the method for screening out the potential user data of the new product from the user data pool comprises the following steps:
collecting user data from a user data pool and distributing the user data to screening personnel for classification, wherein the classification comprises the following steps: potential users and non-potential users;
and acquiring the user data after the screening personnel classification, and storing the user data of the potential users classified as the screening personnel as the potential user data of the new product in a first target domain.
Preferably, the data distribution adaptive migration learning model includes at least one of an edge distribution adaptive migration learning model and a joint distribution adaptive migration learning model.
Specifically, the user data of the source domain corresponds to a first feature group, the user data of the first target domain corresponds to a second feature group, and the first feature group and the second feature group are respectively composed of a plurality of features;
before inputting the user data of the source domain and the first target domain into a plurality of data distribution adaptive migration learning models, screening out common features of the first feature group and the second feature group, and deleting features except the common features from the first feature group and the second feature group respectively.
Further, the method for obtaining the high-quality users from the first sample set for marketing further comprises:
distributing the user data in the first sample set to an expert for auditing, wherein the auditing comprises: auditing and checking the tags of the user data in the first sample set;
acquiring user data which is audited by an expert, screening out user data carrying a label which identifies a user to purchase, and storing the user data as a high-quality user;
and marketing the high-quality users.
Preferably, the method for acquiring marketing users further comprises:
acquiring user data generated in the process of marketing aiming at the high-quality user and storing the user data into a second target domain;
inputting the user data of the source domain and the second target domain into a plurality of sample-based migration learning models, screening out the user data with the potential of purchasing a new product from the source domain based on the user data of the second target domain, and storing the user data into a second sample set;
and marketing aiming at the users corresponding to the user data in the second sample set.
Further, the sample-based migration learning model is realized through a TrAdaBoost algorithm model with different parameters.
Specifically, the method for acquiring the marketing user further comprises the following steps:
acquiring user data generated in the process of marketing aiming at the users corresponding to the user data in the second sample set and storing the user data in a third target domain;
acquiring a third sample set from the source domain based on the third target domain by using a machine learning model, and grading user data in the third sample set;
and selecting a preset amount of user data from the third sample set for marketing based on the grading result.
Further, scoring the user data in the third sample set by using a scoring card model in a Logitics algorithm;
the preset number is determined based on the scoring result, the marketing cost and the marketing budget.
A system for obtaining marketing users, comprising:
the system comprises a user data acquisition module, a data storage module and a data processing module, wherein the user data acquisition module is used for acquiring user data during marketing of products similar to new products and screening out user data of potential users of the new products from a user data pool;
the data processing module is used for predicting labels for user data in a target domain and acquiring high-quality users;
and the marketing module is used for marketing the high-quality users.
Compared with the prior art, the method and the system for acquiring the marketing user have the following beneficial effects:
the method for acquiring the marketing user provided by the invention is based on transfer learning and a small amount of manual participation, changes the prior mode of simply relying on manual experience into the mode of combining machine learning, more intelligently selects marketing customer groups, perfects a machine learning model in the early stage of new product marketing, solves the problems that no data exists in the middle stage before a new product is on line, users cannot be analyzed, and a conventional machine learning model is unavailable, and saves the labor cost and the marketing channel cost.
The system for acquiring the marketing user solves the problems that the number of the marketing user groups is small and no proper conventional supervised learning model is available after the product is on line, not only greatly saves the operation cost, but also obviously improves the operation efficiency and marketing effect and enables the marketing user groups to be distributed more reasonably.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention and not to limit the invention. In the drawings:
FIG. 1 is a flow chart illustrating a method for obtaining marketing users without a sample phase according to an embodiment of the present invention;
FIG. 2 is a flowchart illustrating a transfer learning method for a no-sample stage according to an embodiment of the present invention;
FIG. 3 is a schematic diagram illustrating a transfer learning effect for a no-sample stage according to an embodiment of the present invention;
FIG. 4 is a comparison of acquiring marketing users according to an embodiment of the present invention with the prior art;
FIG. 5 is a flow chart illustrating a method for obtaining marketing users in a small sample stage according to an embodiment of the present invention;
FIG. 6 is a flowchart illustrating a transfer learning method for a small number of sample stages according to an embodiment of the present invention;
fig. 7 is a flowchart illustrating a method for completely acquiring a marketing user according to an embodiment of the present invention.
Detailed Description
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in detail below. It is to be understood that the described embodiments are merely exemplary of the invention, and not restrictive of the full scope of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example one
Referring to fig. 1, a method for acquiring a marketing user for a pull-new marketing scenario without a sample phase includes:
acquiring user data similar to a new product during product marketing and storing the user data in a source domain, wherein the user data in the source domain carries a tag which is used for identifying whether a user purchases or does not purchase;
screening out user data of potential users of a new product from a user data pool, and storing the user data of the potential users in a first target domain, wherein the user data of the first target domain does not carry a label;
inputting user data of a source domain and a first target domain into a plurality of data distribution self-adaptive migration learning models, adding prediction labels to all/part of user data in the first target domain, outputting and storing the user data in a first sample set;
acquiring high-quality users from the first sample set and marketing, wherein the high-quality users are as follows: and carrying the user corresponding to the user data of the label for identifying the user to purchase.
The method for acquiring the marketing user provided by the invention converts the prior method of simply relying on manual experience into the method of combining machine learning, more intelligently selecting the marketing customer group, perfects the machine learning model in the early stage of new product marketing, solves the problems that no data exists in the middle stage before a new product is on line, the user cannot be analyzed, and the available conventional machine learning model does not exist, and saves the labor cost and the marketing channel cost.
The method for screening out potential user data of a new product from the user data pool comprises the following steps:
collecting user data from a user data pool and distributing the user data to screening personnel for classification, wherein the classification comprises the following steps: potential users and non-potential users;
and acquiring the user data after the screening personnel classification, and storing the user data of the potential users classified as the screening personnel into a first target domain as the potential user data of the new product.
Since the new product is not being marketed at this point, the data in the first target domain is untagged.
The data distribution self-adaption is a common transfer learning algorithm, and the basic idea of the method is that the distance of different data distribution is shortened through transformation due to the fact that the probability distribution of data of a source domain is different from that of data of a target domain. Such methods can be further classified into edge distribution adaptation, conditional distribution adaptation, and joint distribution adaptation according to the nature of the data distribution.
In this embodiment, the three data distribution adaptive migration learning algorithms are tried respectively, and two data distribution adaptive methods with relatively better effects, namely, a marginal distribution adaptive (TCA) method and a Joint Distribution Adaptive (JDA) method, are finally selected.
The goal of edge distribution self-adaptation is to reduce the distance between the edge probability distribution of the source domain and the first target domain, thereby completing the transfer learning. Formally, edge distribution adaptation is the approximation of the difference between the two domains by the distance between the edge probabilities of both the source domain and the first target domain.
The goal of joint distribution adaptation is to reduce the distance of the joint probability distribution of the source domain and the first target domain, thereby completing the transfer learning. Formally, the joint distribution adaptation method approximates the difference between the two domains with the distance between the edge probabilities of both the source and target domains and the distance between the conditional probabilities of both the source and target domains.
Referring to fig. 2, in performing the transfer learning, first we keep the features between the source domain and the first target domain consistent, and if there is a feature with only one domain, delete it directly. The specific method comprises the following steps:
the user data of the source domain corresponds to a first characteristic group, the user data of the first target domain corresponds to a second characteristic group, and the first characteristic group and the second characteristic group are respectively composed of a plurality of characteristics;
the common features of the first feature group and the second feature group are screened out, and features other than the common features are deleted from the first feature group and the second feature group, respectively.
And then, respectively substituting the user data of the source domain and the first target domain into JDA and TCA data distribution adaptive migration learning models to carry out model training. The JDA model obtains a target transformation matrix a, and then performs multiple iterations, where each iteration uses a label obtained in the previous iteration as a label, and after the multiple iterations, a more accurate label is obtained, as shown in fig. 3. And the label obtained by iteration is a predicted label.
And finally, comprehensively considering the size of the passenger group to be marketed according to factors such as marketing cost and the like, selecting users from the target passenger group obtained by the JDA model and the TCA model respectively, and storing the users as a first sample set.
The method for obtaining the high-quality users from the first sample set for marketing further comprises the following steps:
distributing the user data in the first sample set to an expert for auditing, wherein the auditing comprises the following steps: auditing and checking the tags of the user data in the first sample set; acquiring user data which is audited by an expert, screening out user data carrying a label which identifies a user to purchase, and storing the user data as a high-quality user; and marketing the high-quality users.
Wherein the standard of the expert review is more severe than the review standard of the screening personnel.
Referring to fig. 5, a method for obtaining a marketing user for a pull-new marketing scenario with a small number of sample phases includes: acquiring user data generated in the process of marketing aiming at high-quality users and storing the user data into a second target domain; inputting user data of a source domain and a second target domain into a plurality of sample-based migration learning models, screening out user data with the potential of purchasing a new product from the source domain based on the user data of the second target domain, and storing the user data into a second sample set; and marketing aiming at the users corresponding to the user data in the second sample set.
After a period of time for marketing of high-quality users, some samples are brought by marketing customer groups, a part of samples which are naturally transformed also exist, and after a small amount of samples are accumulated, the marketing process enters a second part, namely a small amount of sample stage. In this stage, a sample-based Transfer Learning method (instant based Transfer Learning) may be adopted, and the data samples are reused according to a certain weight generation rule to perform Transfer Learning. The sample-based transfer learning method can be realized by a TrAdaBoost algorithm model with different parameters.
The TrAdaBoost algorithm applies the idea of AdaBoost to transfer learning, improves the example weight beneficial to the target classification task, reduces the example weight not beneficial to the target classification task, and deduces the generalization error upper bound of the model based on the PAC theory. Specifically, the TrAdaBoost algorithm adjusts the weight of the training samples, and if the training samples are the samples with the target domain being wrongly divided, the TrAdaBoost algorithm adjusts the training samples according to the classification error rate of the samples with the target domain, and increases the weight of the samples with the target domain being wrongly divided, so that the samples with the target domain being wrongly divided can be paid more attention in the next training; if the samples are samples with the source domain being wrongly divided, the TrAdaBoost algorithm considers the samples to be samples with different data distribution from the target domain, and therefore the weight of the samples is reduced.
Referring to fig. 6, first, two thirds of features in the first feature group and the second feature group are randomly selected for training by using the tragaboost algorithm with two different parameters, and a certain amount of samples are selected from the source domain by using the tragaboost weight updating method, and the two samples are combined to obtain the second sample set. The TrAdaBoost algorithm with two different parameters is used for improving the stability of the model and effectively reducing the problem of high variance caused by screening samples by a single model.
After the sample size in the second sample set is increased, which is sufficient to characterize the user image or distribution of the guest group, a conventional supervised learning solution may be used. The method for acquiring the marketing user in the sufficient sample stage comprises the following steps:
acquiring user data generated in the process of marketing aiming at the users corresponding to the user data in the second sample set and storing the user data in a third target domain; acquiring a third sample set from the source domain based on a third target domain by using a machine learning model, and grading user data in the third sample set; and selecting a preset amount of user data from the third sample set for marketing based on the grading result.
The user data in the third sample set can be scored by using a scoring card model in a Logitics algorithm; the preset number may be determined based on the scoring result, the marketing cost, and the marketing budget; wherein, the marketing channel can be used by short message, push or electric marketing.
For example, if the score result is 500-; if the scoring result is 400-500 points, the marketing effect on the user is better; if the scoring result is 300-400 points, it indicates that the marketing effect for the user may be lost; if the scoring result is 300 points, the marketing effect on the user is indicated to be a loss.
To this end, the method for completely acquiring the marketing user in this embodiment can be divided into three stages: a no sample stage, a small sample stage and a sufficient sample stage, as shown in fig. 7.
Referring to fig. 4, in a conventional operation strategy for online time of a new product in a pull-in link, expert experience is often used to select a guest group in a large range, which results in inaccurate guest group selection, time and labor waste, incomplete consideration, and easy neglect of a high-quality guest group. The method for acquiring the marketing user provided by the invention is an idea of combining transfer learning and expert experience, a part of samples are selected by using the transfer learning, and a better-quality sample is selected by using proper conditions of the expert experience, so that the method is more accurate on one hand, and on the other hand, a buried potential customer group is more likely to be selected by using an intelligent algorithm.
The migration learning algorithm framework provided by the invention is suitable for scenes from just online no sample available of a product to scenes in which a small amount of samples are accumulated and the traditional supervised learning algorithm is not enough used, and fills the blank of the two scene technical frameworks.
In a label-free scene, the invention combines a plurality of transfer learning algorithms corresponding to different data distributions, is more comprehensive and convincing, and increases the self-adaptive capacity and the coupling of an algorithm framework. Meanwhile, the advantages of traditional expert experience are not completely abandoned or ignored, an algorithm framework is combined with the traditional expert experience in parallel, the traditional and the modernized intelligence are combined together, the two families are grown into a whole, and the whole system or the marketing scheme is full and convincing.
In a scene with a small number of labels, the invention utilizes a small number of samples of the target domain, combines a plurality of migration learning classifiers, converts more samples in the source domain into samples of the target domain, and greatly increases the samples of the target domain by combining the samples, so that the sample size can meet the requirements of a traditional supervision machine learning model, and the samples are seamlessly connected with a traditional operation mode, thereby forming the communication of the whole pull-up scene full link, and enabling the operation work to be more intelligent and convenient.
Example two
A system for acquiring marketing users comprises a user data acquisition module, a data processing module and a marketing module, wherein,
the user data acquisition module is used for acquiring user data during marketing of products similar to the new products and screening out user data of potential users of the new products from a user data pool;
the data processing module is used for predicting labels for user data in a target domain and acquiring high-quality users;
and the marketing module is used for marketing the high-quality users.
By adopting the method for acquiring the marketing user in the first embodiment, the system provided by the invention solves the problems that the number of the marketing user groups is small and no proper conventional supervised learning model is available after the product is on line, greatly saves the operation cost, obviously improves the operation efficiency and marketing effect and enables the marketing user groups to be distributed more reasonably. Compared with the prior art, the system for acquiring the marketing user provided by the embodiment of the invention has the same beneficial effect as the method for acquiring the marketing user provided by the first embodiment, and other technical features in the system for acquiring the marketing user are the same as those disclosed in the method of the previous embodiment, which are not repeated herein.
Claims (10)
1. A method for obtaining marketing users, comprising:
acquiring user data similar to a new product during product marketing and storing the user data in a source domain, wherein the user data in the source domain carries a tag used for identifying the purchase/non-purchase of a user;
screening out user data of potential users of the new product from a user data pool and storing the user data in a first target domain, wherein the user data in the first target domain does not carry the label;
inputting the user data of the source domain and the first target domain into a plurality of data distribution self-adaptive migration learning models, adding prediction labels to all/part of the user data in the first target domain, and outputting and storing the user data in a first sample set;
obtaining high-quality users from the first sample set and marketing, wherein the high-quality users are as follows: and carrying the user corresponding to the user data of the label for identifying the user to purchase.
2. The method for obtaining marketing users according to claim 1, wherein the method of screening potential user data of the new product from a user data pool comprises:
collecting user data from a user data pool and distributing the user data to screening personnel for classification, wherein the classification comprises the following steps: potential users and non-potential users;
and acquiring the user data after the screening personnel classification, and storing the user data of the potential users classified as the screening personnel as the potential user data of the new product in a first target domain.
3. The method for obtaining marketing users according to claim 1, wherein the data distribution adaptive migration learning model comprises at least one of an edge distribution adaptive migration learning model and a joint distribution adaptive migration learning model.
4. The method for obtaining marketing users according to claim 1, wherein the user data of the source domain corresponds to a first feature group, the user data of the first target domain corresponds to a second feature group, and the first feature group and the second feature group are respectively composed of a plurality of features;
before inputting the user data of the source domain and the first target domain into a plurality of data distribution adaptive migration learning models, screening out common features of the first feature group and the second feature group, and deleting features except the common features from the first feature group and the second feature group respectively.
5. The method for obtaining marketing users according to claim 1, wherein the method for obtaining premium users from the first sample set for marketing further comprises:
distributing the user data in the first sample set to an expert for auditing, wherein the auditing comprises: auditing and checking the tags of the user data in the first sample set;
acquiring user data which is audited by an expert, screening out user data carrying a label which identifies a user to purchase, and storing the user data as a high-quality user;
and marketing the high-quality users.
6. The method for obtaining marketing users according to claim 1, further comprising:
acquiring user data generated in the process of marketing aiming at the high-quality user and storing the user data into a second target domain;
inputting the user data of the source domain and the second target domain into a plurality of sample-based migration learning models, screening out the user data with the potential of purchasing a new product from the source domain based on the user data of the second target domain, and storing the user data into a second sample set;
and marketing aiming at the users corresponding to the user data in the second sample set.
7. The method for obtaining marketing users according to claim 6, wherein the sample-based migration learning model is implemented by a TrAdaBoost algorithm model with different parameters.
8. The method for obtaining marketing users according to claim 1, further comprising:
acquiring user data generated in the process of marketing aiming at the users corresponding to the user data in the second sample set and storing the user data in a third target domain;
acquiring a third sample set from the source domain based on the third target domain by using a machine learning model, and grading user data in the third sample set;
and selecting a preset amount of user data from the third sample set for marketing based on the grading result.
9. The method for obtaining marketing users according to claim 8, wherein the user data in the third sample set is scored using a scoring card model in a logistic algorithm;
the preset number is determined based on the scoring result, the marketing cost and the marketing budget.
10. A system for obtaining marketing users, comprising:
the system comprises a user data acquisition module, a data storage module and a data processing module, wherein the user data acquisition module is used for acquiring user data during marketing of products similar to new products and screening out user data of potential users of the new products from a user data pool;
the data processing module is used for predicting labels for user data in a target domain and acquiring high-quality users;
and the marketing module is used for marketing the high-quality users.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110641191.XA CN113343087A (en) | 2021-06-09 | 2021-06-09 | Method and system for acquiring marketing user |
| CA3161985A CA3161985A1 (en) | 2021-06-09 | 2022-06-08 | Marketing-user acquiring method and system thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110641191.XA CN113343087A (en) | 2021-06-09 | 2021-06-09 | Method and system for acquiring marketing user |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113343087A true CN113343087A (en) | 2021-09-03 |
Family
ID=77475675
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110641191.XA Withdrawn CN113343087A (en) | 2021-06-09 | 2021-06-09 | Method and system for acquiring marketing user |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN113343087A (en) |
| CA (1) | CA3161985A1 (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108304876A (en) * | 2018-01-31 | 2018-07-20 | 国信优易数据有限公司 | Disaggregated model training method, device and sorting technique and device |
| CN110020121A (en) * | 2017-10-16 | 2019-07-16 | 上海交通大学 | Software crowdsourcing item recommendation method and system based on transfer learning |
| CN111507768A (en) * | 2020-04-17 | 2020-08-07 | 腾讯科技（深圳）有限公司 | Determination method of potential user, model training method and related device |
| CN112086146A (en) * | 2020-08-24 | 2020-12-15 | 南京邮电大学 | Small molecule drug virtual screening method and device based on deep parameter transfer learning |
| CN112347392A (en) * | 2020-10-21 | 2021-02-09 | 上海淇玥信息技术有限公司 | Anti-fraud assessment method and device based on transfer learning and electronic equipment |
| CN112801718A (en) * | 2021-02-22 | 2021-05-14 | 平安科技（深圳）有限公司 | User behavior prediction method, device, equipment and medium |
-
2021
- 2021-06-09 CN CN202110641191.XA patent/CN113343087A/en not_active Withdrawn
-
2022
- 2022-06-08 CA CA3161985A patent/CA3161985A1/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110020121A (en) * | 2017-10-16 | 2019-07-16 | 上海交通大学 | Software crowdsourcing item recommendation method and system based on transfer learning |
| CN108304876A (en) * | 2018-01-31 | 2018-07-20 | 国信优易数据有限公司 | Disaggregated model training method, device and sorting technique and device |
| CN111507768A (en) * | 2020-04-17 | 2020-08-07 | 腾讯科技（深圳）有限公司 | Determination method of potential user, model training method and related device |
| CN112086146A (en) * | 2020-08-24 | 2020-12-15 | 南京邮电大学 | Small molecule drug virtual screening method and device based on deep parameter transfer learning |
| CN112347392A (en) * | 2020-10-21 | 2021-02-09 | 上海淇玥信息技术有限公司 | Anti-fraud assessment method and device based on transfer learning and electronic equipment |
| CN112801718A (en) * | 2021-02-22 | 2021-05-14 | 平安科技（深圳）有限公司 | User behavior prediction method, device, equipment and medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CA3161985A1 (en) | 2022-12-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| White et al. | Imputation in us manufacturing data and its implications for productivity dispersion | |
| CN108764584B (en) | Enterprise electric energy substitution potential evaluation method | |
| CN107766929B (en) | Model analysis method and device | |
| CN108205766A (en) | Information-pushing method, apparatus and system | |
| CN111967971B (en) | Bank customer data processing method and device | |
| CN110766438B (en) | Method for analyzing user behavior of power grid user through artificial intelligence | |
| CN108304890A (en) | A kind of generation method and device of disaggregated model | |
| CN108804577B (en) | Method for estimating interest degree of information tag | |
| CN108734296A (en) | Optimize method, apparatus, electronic equipment and the medium of the training data of supervised learning | |
| CN111369299A (en) | Method, device and equipment for identification and computer readable storage medium | |
| CN110263979A (en) | Method and device based on intensified learning model prediction sample label | |
| CN111611488A (en) | Information recommendation method and device based on artificial intelligence and electronic equipment | |
| CN116415017B (en) | Advertisement sensitive content auditing method and system based on artificial intelligence | |
| CN111949795A (en) | Work order automatic classification method and device | |
| CN110147389A (en) | Account number treating method and apparatus, storage medium and electronic device | |
| CN112561568A (en) | Target customer prediction method, device and storage medium | |
| Mokarram et al. | Net-load forecasting of renewable energy systems using multi-input LSTM fuzzy and discrete wavelet transform | |
| CN113469730A (en) | Customer repurchase prediction method and device based on RF-LightGBM fusion model under non-contract scene | |
| CN112488551A (en) | XGboost algorithm-based hot line intelligent order dispatching method | |
| CN109033378A (en) | A kind of application method of Zero-shot Learning in intelligent customer service system | |
| CN113343087A (en) | Method and system for acquiring marketing user | |
| CN109241249B (en) | Method and device for determining burst problem | |
| CN114549035A (en) | Construction method of financial user accurate customer acquisition label based on telecommunication big data | |
| CN113313615A (en) | Method and device for quantitatively grading and grading enterprise judicial risks | |
| CN113407827A (en) | Information recommendation method, device, equipment and medium based on user value classification |
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 | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20210903 |