CN111199429A - Prediction model training method and device - Google Patents
Prediction model training method and device Download PDFInfo
- Publication number
- CN111199429A CN111199429A CN202010014918.7A CN202010014918A CN111199429A CN 111199429 A CN111199429 A CN 111199429A CN 202010014918 A CN202010014918 A CN 202010014918A CN 111199429 A CN111199429 A CN 111199429A
- Authority
- CN
- China
- Prior art keywords
- equity
- prediction model
- type
- training
- 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.)
- Pending
Links
Images
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
- G06Q30/0207—Discounts or incentives, e.g. coupons or rebates
-
- 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
- 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
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Finance (AREA)
- Development Economics (AREA)
- Accounting & Taxation (AREA)
- Theoretical Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Game Theory and Decision Science (AREA)
- Economics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The embodiment of the specification provides a method and a device for training a prediction model, wherein the method comprises the steps of allocating a right for each collected user and dividing the right according to right types; dividing the equity characteristics of each equity type and the corresponding user characteristics of the user into training data and test data; training an initial prediction model corresponding to each equity type based on the training data of each equity type to obtain a prediction model corresponding to each equity type; and inputting the test data of each rights and interests type into the prediction model corresponding to each rights and interests type to obtain a test result, and adjusting the prediction model based on all the test results.
Description
Technical Field
The embodiment of the specification relates to the technical field of model training, in particular to a training method of a prediction model. One or more embodiments of the present specification also relate to a training apparatus for a predictive model, a computing device, and a computer-readable storage medium.
Background
At present, in a network marketing scene, in order to stimulate user consumption, a plurality of shopping platforms issue red packets, coupons and the like aiming at users, which can stimulate the rights and interests of the user consumption, and the rights and interests all have certain cost, if the users are not interested in the rights and interests, the rights and interests issued by the shopping platforms do not enable the users to consume based on the rights and interests, and the rights and interests are wasted; therefore, the prediction model for accurately predicting the response condition of the user under the influence of different rights and interests according to the user characteristics is particularly important for reasonable putting of the rights and interests.
Disclosure of Invention
In view of this, the embodiments of the present disclosure provide a method for training a prediction model. One or more embodiments of the present disclosure also relate to a training apparatus for a prediction model, a computing device, and a computer-readable storage medium, which are used to solve the technical problems of the prior art.
According to a first aspect of embodiments herein, there is provided a method for training a prediction model, including:
distributing a right for each collected user, and dividing the right according to the right type;
dividing the equity characteristics of each equity type and the corresponding user characteristics of the user into training data and test data;
training an initial prediction model corresponding to each equity type based on the training data of each equity type to obtain a prediction model corresponding to each equity type;
and inputting the test data of each rights and interests type into the prediction model corresponding to each rights and interests type to obtain a test result, and adjusting the prediction model based on all the test results.
Optionally, the training the initial prediction model corresponding to each equity type based on the training data of each equity type, and obtaining the prediction model corresponding to each equity type includes:
and training the initial prediction model corresponding to each equity type based on the training data of each equity type to obtain a prediction model which corresponds to each equity type and outputs the response rate of the user characteristics to the equity characteristics.
Optionally, the test result includes a test response rate of the user characteristic for the benefit characteristic;
correspondingly, the inputting the test data of each equity type into the prediction model corresponding to each equity type to obtain the test result, and adjusting the prediction model based on all the test results includes:
and inputting the test data of each equity type into the prediction model corresponding to each equity type to obtain the test response rate of the user characteristics of each equity type to the equity characteristics, and adjusting the prediction model based on the test response rate.
Optionally, before adjusting the prediction model based on the test response rate, the method further includes:
and acquiring the preset response rate of the user characteristics output by the prediction model of each equity type to the equity characteristics.
Optionally, the adjusting the prediction model based on the test response rate includes:
and calculating a difference value between a preset response rate of the user characteristics output by the prediction model of each interest type aiming at the interest characteristics and a corresponding test response rate, and adjusting the prediction model based on the difference value when the difference value is greater than or equal to a preset threshold value.
Optionally, the adjusting the prediction model based on the test response rate includes:
sorting the user characteristics output by the prediction models of all the rights and interests types according to the preset response rate of the rights and interests characteristics to obtain a sorting result of the preset response rate;
sequencing the user characteristics output by the prediction models of all the rights and interests aiming at the test response rate of the rights and interests characteristics to obtain a sequencing result of the test response rate;
and under the condition that the sequencing result of the preset response rate is different from the sequencing result of the test response rate, adjusting a prediction model based on the test response rate.
Optionally, the adjusting the prediction model based on the test response rate includes:
and adjusting at least one prediction model corresponding to the ownership type based on the test response rate.
Optionally, the training the initial prediction model corresponding to each equity type based on the training data of each equity type, and obtaining the prediction model corresponding to each equity type includes:
and training the same initial prediction model corresponding to each equity type based on the training data of each equity type to obtain the same prediction model corresponding to each equity type.
Optionally, obtaining the prediction model corresponding to each equity type includes:
and training different initial prediction models corresponding to each equity type based on the training data of each equity type to obtain different prediction models corresponding to each equity type.
Optionally, before dividing the at least two collected rights and interests according to the rights and interests types and allocating different users to the rights and interests of each divided rights and interests type, the method further includes:
multiple users and multiple rights are collected.
Optionally, the user characteristics include attribute characteristics and behavior characteristics of the user.
According to a second aspect of embodiments herein, there is provided a training apparatus for a predictive model, including:
the right and interest division module is configured to distribute one right and interest to each collected user and divide the right and interest according to the right and interest type;
a data partitioning module configured to partition the equity features of the equity of each equity type and the user features of the corresponding user into training data and test data;
the model training module is configured to train the initial prediction model corresponding to each equity type based on the training data of each equity type to obtain a prediction model corresponding to each equity type;
and the model adjusting module is configured to input the test data of each interest type into the prediction model corresponding to each interest type to obtain a test result, and adjust the prediction model based on all the test results.
Optionally, the model training module is further configured to:
and training the initial prediction model corresponding to each equity type based on the training data of each equity type to obtain a prediction model which corresponds to each equity type and outputs the response rate of the user characteristics to the equity characteristics.
Optionally, the test result includes a test response rate of the user characteristic for the benefit characteristic;
accordingly, the model adjustment module is further configured to:
and inputting the test data of each equity type into the prediction model corresponding to each equity type to obtain the test response rate of the user characteristics of each equity type to the equity characteristics, and adjusting the prediction model based on the test response rate.
Optionally, the apparatus further includes:
and the preset response rate acquisition module is configured to acquire the preset response rate of the user characteristics output by the prediction model of each equity type to the equity characteristics.
Optionally, the model adjusting module is further configured to:
and calculating a difference value between a preset response rate of the user characteristics output by the prediction model of each interest type aiming at the interest characteristics and a corresponding test response rate, and adjusting the prediction model based on the difference value when the difference value is greater than or equal to a preset threshold value.
Optionally, the model adjusting module is further configured to:
sorting the user characteristics output by the prediction models of all the rights and interests types according to the preset response rate of the rights and interests characteristics to obtain a sorting result of the preset response rate;
sequencing the user characteristics output by the prediction models of all the rights and interests aiming at the test response rate of the rights and interests characteristics to obtain a sequencing result of the test response rate;
and under the condition that the sequencing result of the preset response rate is different from the sequencing result of the test response rate, adjusting a prediction model based on the test response rate.
Optionally, the model adjusting module is further configured to:
and adjusting at least one prediction model corresponding to the ownership type based on the test response rate.
Optionally, the model training module is further configured to:
and training the same initial prediction model corresponding to each equity type based on the training data of each equity type to obtain the same prediction model corresponding to each equity type.
Optionally, the model training module is further configured to:
and training different initial prediction models corresponding to each equity type based on the training data of each equity type to obtain different prediction models corresponding to each equity type.
Optionally, the apparatus further includes:
an acquisition module configured to acquire a plurality of users and a plurality of rights and interests.
Optionally, the user characteristics include attribute characteristics and behavior characteristics of the user.
According to a third aspect of embodiments herein, there is provided a computing device comprising:
a memory and a processor;
the memory is to store computer-executable instructions, and the processor is to execute the computer-executable instructions to:
distributing a right for each collected user, and dividing the right according to the right type;
dividing the equity characteristics of each equity type and the corresponding user characteristics of the user into training data and test data;
training an initial prediction model corresponding to each equity type based on the training data of each equity type to obtain a prediction model corresponding to each equity type;
and inputting the test data of each rights and interests type into the prediction model corresponding to each rights and interests type to obtain a test result, and adjusting the prediction model based on all the test results.
According to a fourth aspect of embodiments herein, there is provided a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of any one of the methods for training a predictive model.
One embodiment of the specification realizes a training method and a device of a prediction model, wherein the method comprises the steps of allocating a right for each collected user and dividing the right according to right types; dividing the equity characteristics of each equity type and the corresponding user characteristics of the user into training data and test data; training an initial prediction model corresponding to each equity type based on the training data of each equity type to obtain a prediction model corresponding to each equity type; inputting the test data of each equity type into the prediction model corresponding to each equity type to obtain a test result, and adjusting the prediction model based on all the test results;
according to the method, each prediction model only optimizes the rights and interests of one rights and interests type in a way of training the prediction models through the rights and interests, the data is simple, the fitting effect of each trained prediction model is good, the training data of each rights and interests type corresponds to one prediction model, and the training efficiency of the prediction models can be greatly improved by adopting the distributed parallel training prediction model; and the prediction model can be adjusted based on the training result of the prediction model, so that the prediction rationality and accuracy of the prediction model are greatly ensured.
Drawings
FIG. 1 is a flow chart of a method for training a predictive model provided in one embodiment of the present description;
FIG. 2 is a flowchart illustrating model training and model testing in a predictive model training method according to an embodiment of the present disclosure;
FIG. 3 is a flowchart illustrating user response prediction after weighted interest modeling in a method for training a predictive model according to an embodiment of the present disclosure;
FIG. 4 is a schematic structural diagram of a training apparatus for a prediction model according to an embodiment of the present disclosure;
fig. 5 is a block diagram of a computing device according to an embodiment of the present disclosure.
Detailed Description
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present description. This description may be embodied in many different forms and should not be construed as limited to the embodiments set forth herein, as those skilled in the art will be able to make and use the present disclosure without departing from the spirit and scope of the present disclosure.
The terminology used in the description of the one or more embodiments is for the purpose of describing the particular embodiments only and is not intended to be limiting of the description of the one or more embodiments. As used in one or more embodiments of the present specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used in one or more embodiments of the present specification refers to and encompasses any and all possible combinations of one or more of the associated listed items.
It will be understood that, although the terms first, second, etc. may be used herein in one or more embodiments to describe various information, these information should not be limited by these terms. These terms are only used to distinguish one type of information from another. For example, a first can also be referred to as a second and, similarly, a second can also be referred to as a first without departing from the scope of one or more embodiments of the present description. The word "if" as used herein may be interpreted as "at … …" or "when … …" or "in response to a determination", depending on the context.
First, the noun terms to which one or more embodiments of the present specification relate are explained.
The right and interests: in marketing campaigns, items with costs, such as red packs, coupons, or coupons, are issued to users to incentivize the users to participate in the targeted campaign.
Response rate: the probability that a user will engage in a target activity with some kind of benefit.
In the present specification, a training method of a predictive model is provided, and the present specification also relates to a training apparatus of a predictive model, a computing device, and a computer-readable storage medium, which are described in detail in the following embodiments one by one.
Fig. 1 shows a flowchart of a training method of a predictive model according to an embodiment of the present disclosure, which includes steps 102 to 108.
Step 102: and allocating a right for the collected user characteristics of each user, and dividing the right according to the right types.
Wherein the rights and interests include but are not limited to cash red packet, full discount red packet, telephone charge red packet, free coupon, discount coupon, etc.; the equity types can be determined according to the actual application scenarios of the collected equity, for example, if the acquired equity includes a cash red packet, a free ticket, a discount ticket and the like, since the cash red packet can be directly used for cash offset in the actual application scenarios, the free ticket is used for interest offset when the user pays in installments in the actual application scenarios, and the discount ticket is used for discounting the purchased goods when the user purchases the goods in the actual application scenarios, the equity types can be determined as the three equity types according to the several different application scenarios, at this time, the equity types can be understood as the application scenarios, and then the equity types are divided according to the three equity types.
In addition, the rights and interests type can also be determined according to the amount; for example, in the same application scenario, the cash red packages with different quotas also need to be divided according to different quotas, at this time, the equity type can be understood as a quota, and then the equity is divided according to the quotas.
In practical application, the rights and interests are divided according to the rights and interests types, can be divided according to specific application scenes, can also be divided according to the quota or cost of the rights and interests, can also be divided according to the specific application scenes firstly, and then the rights and interests divided according to the specific application scenes are divided in detail according to the quota or cost of the rights and interests, and the specific dividing mode of the rights and interests is determined based on specific rights and interests contents, and can also be divided by adopting other rights and interests types, so that the collected rights and interests can be classified in detail, and no limitation is made herein.
The rights and interests are divided according to the rights and interests types so that the rights and interests can be divided and trained when the prediction model is trained subsequently, the rights and interests of the prediction model corresponding to one rights and interests type are trained, data are simpler, and the fitting effect of the trained prediction model is better.
In another embodiment of this specification, before allocating a right to each collected user and dividing the right according to the right type, the method further includes:
multiple users and multiple benefits are collected.
In specific implementation, in order to ensure the accuracy of prediction performed by the prediction model, a plurality of rights and interests and a plurality of user-formed training data and prediction data are collected in advance, wherein the number of the rights and interests is three or more.
In practical application, a plurality of users and a plurality of rights are collected, namely the user characteristics of each user and the rights characteristics of each right are collected; wherein, the equity characteristics of the equity include but are not limited to characteristics of discount rate, service life, amount, etc.; the right characteristics of the rights of each right type are the same, taking the interest-free ticket as an example, the interest-free ticket of each amount has the characteristics of discount rate, service life and the like, taking the red packet as an example, the red packet of each amount has the characteristics of service life, service range of purchased goods and the like;
the user characteristics of the user include, but are not limited to, attribute characteristics of the user including, but not limited to, the user's age, gender, native place, school calendar, etc., and behavior characteristics including, but not limited to, the user's shopping habits, reading habits, consumption records, etc.
And after the rights and the users are collected, distributing one right for each collected user, and dividing the rights according to the right types.
For example, the collected plurality of users includes user 1, user 2, user 3, user 4, user 5, user 6, and the collected plurality of benefits includes a 1-tuple full minus red packet, a 3-tuple full minus red packet, a 5-tuple full minus red packet, a 6-tuple full minus red packet, a 10-tuple full minus red packet, and an 11-tuple full minus red packet;
the right to assign to each collected user may be: assign a 1-tuple full minus red envelope to user 1, a 3-tuple full minus red envelope to user 2, a 5-tuple full minus red envelope to user 3, a 6-tuple full minus red envelope to user 4, a 10-tuple full minus red envelope to user 5, an 11-tuple full minus red envelope to user 6,
the rights and interests are determined to be three types according to the quota, namely, a 1-5 yuan full red packet is one type of rights and interests, a 6-10 yuan full red packet is one type of rights and interests, and a 11-20 yuan full red packet is one type of rights and interests,
then, the rights and interests are divided according to the three rights and interests types, and divided into rights and interests sets of three rights and interests types, namely the rights and interests of the first rights and interests type comprise a 1-element full red packet, a 3-element full red packet and a 5-element full red packet, the rights and interests of the second rights and interests type comprise a 6-element full red packet and a 10-element full red packet, and the rights and interests of the third rights and interests type comprise a 11-element full red packet.
And the users corresponding to the rights and interests of the first rights and interests type comprise a user 1, a user 2 and a user 3; the users corresponding to the rights and interests of the second rights and interests type comprise a user 4 and a user 5; and the interests of the third interest type correspond to the users 6.
In the embodiment of the specification, a plurality of users and a plurality of rights are collected, a right is allocated to each collected user, and the rights are divided according to the rights types so as to base on the determined training data in the training of the prediction model.
Step 104: the equity features of the equity of each equity type and the user features of the corresponding user are classified into training data and test data.
Specifically, the right and interest features of the right and interest of each right and interest type and the user features of the corresponding users are divided into training data and test data, the extracted right and interest features of the right and interest of the first right and interest type and the user features of the users 1, 2 and 3 corresponding to the right and interest of the first right and interest type are divided into two parts, one part is used as training data, the other part is used as test data, so that the training data includes the right and interest features of the right and interest of the first right and interest type and the user features of the users 1, 2 and 3 corresponding to the right and interest of the first right and interest type, the test data includes the right and interest features of the first right and interest type and the user features of the users 1, 2 and 3 corresponding to the right and interest of the first right and interest type, for example, the training data includes the right and interest features of the right and interest of the first right and interest type User characteristics of user 1, user 2 and user 3 corresponding to the equity, the test data including the equity characteristics of the equity of the remaining twenty percent of the first equity type and the user characteristics of user 1, user 2 and user 3 corresponding to the equity of the first equity type; for example, the specific manner of dividing the equity features and the user features of the equity of the second equity type into the training data and the testing data and the equity features and the user features of the equity of the third equity type into the training data and the testing data can be referred to as the method of dividing the equity features and the user features of the equity of the first equity type into the training data and the testing data, and will not be described herein again.
In the embodiment of the description, the equity features of the equity of each equity type and the user features of the corresponding users are extracted, and the equity features of the equity of each equity type and the user features of the corresponding users are divided into training data and test data, so that the trained prediction model is predicted through the test data after the prediction model is trained through the training data, and the accuracy of the prediction result of the prediction model is ensured.
Step 106: and training the initial prediction model corresponding to each equity type based on the training data of each equity type to obtain the prediction model corresponding to each equity type.
Taking the above as an example, if each equity type corresponds to the same initial prediction model, training the initial prediction model corresponding to each equity type based on the training data of each equity type to obtain the prediction model corresponding to each equity type, and training the initial prediction model a by using the training data of the first equity type to obtain the prediction model a; training the initial prediction model b by using the training data of the second rights and interests type to obtain a prediction model b; training the initial prediction model c by using training data of a third interest type to obtain a prediction model c; a, b and c are initial prediction models of the same type, for example, a, b and c are all Back-ProPagation (BP) neural Network prediction models;
if each equity type corresponds to different initial prediction models, training the initial prediction model corresponding to each equity type based on the training data of each equity type to obtain the prediction model corresponding to each equity type, and training the initial prediction model a by using the training data of the first equity type to obtain the prediction model a; training the initial prediction model b by using the training data of the second rights and interests type to obtain a prediction model b; training the initial prediction model c by using training data of a third interest type to obtain a prediction model c; and a, b, and c are different types of initial prediction models, for example, a is a BP (Back-ProPagation Network) neural Network prediction model, b is a regression prediction model, and c is a trend extrapolation prediction model, or a and b are the same type of initial prediction model, c is a type of initial prediction model, or a and c are the same type of initial prediction model, and b is a type of initial prediction model, that is, the three interest types correspond to different types of initial prediction models.
Specifically, the implementation manner of each equity type corresponding to the same initial prediction model is as follows:
the training of the initial prediction model corresponding to each equity type based on the training data of each equity type to obtain the prediction model corresponding to each equity type comprises:
and training the same initial prediction model corresponding to each equity type based on the training data of each equity type to obtain the same prediction model corresponding to each equity type.
The implementation of the same initial prediction model for each equity type is as follows:
the training of the initial prediction model corresponding to each equity type based on the training data of each equity type to obtain the prediction model corresponding to each equity type comprises:
and training different initial prediction models corresponding to each equity type based on the training data of each equity type to obtain different prediction models corresponding to each equity type.
In the embodiment of the specification, training data of various interest types respectively train corresponding initial prediction models in parallel, the distributed parallel training mode is adopted to enable the training speed of the initial prediction models to be high, the training efficiency of the initial prediction models can be greatly improved, each initial prediction model only needs to optimize the interest of one interest type through dividing the interest modeling according to the interest types, the complexity and the learning difficulty of the initial prediction models are greatly reduced through the mode, and the fitting effect of the prediction models obtained after training is better; in addition, in order to increase the richness of the prediction model, the types of the initial prediction models corresponding to each interest type do not need to be kept consistent, and an initial prediction model with a better interest effect on the current interest type can be selected.
In another embodiment of the present specification, the training the initial prediction model corresponding to each equity type based on the training data of each equity type, and obtaining the prediction model corresponding to each equity type includes:
and training the initial prediction model corresponding to each equity type based on the training data of each equity type to obtain a prediction model which corresponds to each equity type and outputs the response rate of the user characteristics to the equity characteristics.
Specifically, training an initial prediction model corresponding to each equity type based on training data of each equity type results in a prediction model corresponding to each equity type, wherein the prediction model outputs a response rate of a user feature to the equity feature.
In practical application, the prediction model corresponding to each equity type can output the response rate of the user characteristics corresponding to the equity type to the equity characteristics.
In the embodiment of the description, the prediction model corresponding to each equity type can output the response rate of the user characteristics corresponding to the equity type to the equity characteristics, and the understanding of an equity issuer to the user characteristics and the equity characteristics is enhanced; in the split-interest modeling, each prediction model can directly learn the response probability from the user corresponding to the interest of each interest type to the interest of the interest type, and can help the interest issuer to find the relationship between the user characteristics and the interest characteristics according to the output result of each prediction model, so that the interest issuer can more effectively analyze the user characteristics and the interest characteristics according to the marketing scene in the following and then reflect the analysis into the actual application scene, thereby increasing the development of the interest and promoting the development of the service; for example, according to the output result of the prediction model, the response rate of the user between 30 and 40 years old to the full red packet with larger money is higher, the equity issuer can analyze that the consumption level of the user between 30 and 40 years old is higher, and new equity can be added to the user group through the information to stimulate the consumption of the user group, and the like.
Step 108: and inputting the test data of each rights and interests type into the prediction model corresponding to each rights and interests type to obtain a test result, and adjusting the prediction model based on all the test results.
Taking the above as an example, the test data of each equity type is input into the prediction model corresponding to each equity type to obtain the test result, that is, the test data of the first equity type is input into the prediction model a to obtain the test result a1, the test data of the second equity type is input into the prediction model b to obtain the test result b1, and the test data of the third equity type is input into the prediction model c to obtain the test result c 1; and adjusting the prediction model a, the prediction model b and/or the prediction model c based on the test results a1, b1, c 1.
In practical application, the adjustment of the prediction model comprises two parts, wherein the first part is to adjust the internal parameters of a corresponding single prediction model based on the training data of each equity type so as to train to obtain an optimal prediction model; and the other part is to test the corresponding trained prediction model based on the test data of each equity type and then adjust the prediction model based on the test result.
Wherein the test result comprises a test response rate of the user characteristic to the equity characteristic;
correspondingly, the inputting the test data of each equity type into the prediction model corresponding to each equity type to obtain the test result, and adjusting the prediction model based on all the test results includes:
and inputting the test data of each equity type into the prediction model corresponding to each equity type to obtain the test response rate of the user characteristics of each equity type to the equity characteristics, and adjusting the prediction model based on the test response rate.
In the embodiment of the present specification, the preset response rate is an average preset response rate, and the test response rate is an average test response rate.
Still taking the above as an example, inputting the test data of each interest type into the prediction model corresponding to each interest type to obtain the test response rate of the user feature of each interest type for the interest feature, that is, inputting the test data of the first interest type into the prediction model a to obtain the test response rate a1 of the user feature of the user 1, the user feature of the user 2, and the user feature of the user 3 for the interest feature of the rights and interests [ 1 yuan full minus red packet, 3 yuan full minus red packet, 5 yuan full minus red packet ], and for the test response rates obtained by inputting the test data of other interest types into the prediction models corresponding to the interest types thereof, reference may be made to this manner, and details are not repeated herein.
In another embodiment of the present specification, before the adjusting the prediction model based on the test response rate, the method further includes:
and acquiring the preset response rate of the user characteristics output by the prediction model of each equity type to the equity characteristics.
Specifically, the preset response rate of the user feature of each equity type to the equity feature is a real response rate of the user feature of each equity type to the equity feature calculated according to real data, for example, the calculated preset response rate of the user feature of the first equity type to the equity feature is set to 30%, the calculated preset response rate of the user feature of the second equity type to the equity feature is set to 32%, and the calculated preset response rate of the user feature of the third equity type to the equity feature is set to 35%.
In another embodiment of the present specification, the adjusting the prediction model based on the test response rate comprises:
and calculating a difference value between a preset response rate of the user characteristics output by the prediction model of each interest type aiming at the interest characteristics and a corresponding test response rate, and adjusting the prediction model based on the difference value when the difference value is greater than or equal to a preset threshold value.
In practical application, when a prediction model is predicted based on test data, the test data of each interest type is input into the prediction model corresponding to each interest type to obtain a test response rate, that is, the average test response rate of all user characteristics to the interest characteristics is as close as possible to a preset response rate, if the difference between the preset response rate of the user characteristics output by the prediction model of a certain interest type to the interest characteristics and the corresponding test response rate is too large and is greater than or equal to a preset threshold value, it is indicated that the prediction result of the prediction model is unreasonable, and the prediction model needs to be adjusted; the preset threshold may be set according to actual needs, and is not limited herein.
In a specific implementation, the adjusting the prediction model based on the test response rate includes:
sorting the user characteristics output by the prediction models of all the rights and interests types according to the preset response rate of the rights and interests characteristics to obtain a sorting result of the preset response rate;
sequencing the user characteristics output by the prediction models of all the rights and interests aiming at the test response rate of the rights and interests characteristics to obtain a sequencing result of the test response rate;
and under the condition that the sequencing result of the preset response rate is different from the sequencing result of the test response rate, adjusting a prediction model based on the test response rate.
And under the condition that the difference value between the preset response rate of the user feature output by the prediction model of each interest type for the interest feature and the corresponding test response rate is smaller than the preset threshold, the test response rates of the prediction models of each interest type are integrally compared, so that the reality of the prediction results of the integral prediction models corresponding to all the interest types is consistent with the reality of the scene data, for example, the response conditions of the first interest, the second interest and the third interest are sequentially increased on the real data, and the prediction results also need to keep the same size relationship.
Taking the above as an example, if the first equity prediction average response rate is 31.5%, the second equity prediction average response rate is 31%, and the third equity prediction average response rate is 34.5%, the difference between the average preset response rate of the user feature output by the single prediction model, calculated in real time, for the equity feature, and the corresponding average test response rate is close, but an error occurs in the equity sequence in the magnitude sequence, and the prediction model also needs to be adjusted.
In a specific implementation, the adjusting the prediction model based on the test response rate includes:
and adjusting at least one prediction model corresponding to the ownership type based on the test response rate.
Specifically, the obtained preset response rates of the prediction models of each interest type are sorted from small to large or from large to small, the obtained test response rates of the prediction models of each interest type are sorted according to the sorting mode of the preset response rates, finally, the two sorting results are compared, if the sorting results are the same, the trained prediction model of each interest type is reasonable in result, and if the sorting results are different, the trained prediction model of each interest type is unreasonable in result, at this time, all or one of the prediction models of each interest type needs to be adjusted according to the sorting results.
For example, the rights and interests divided according to the rights and interests types are 1-element red packet and 2-10-element red packet, the rights and interests of the 1-element red packet correspond to the user a, the rights and interests of the 2-10-element red packet correspond to the user b, training data and test data are then derived based on the user characteristics of user a and user b and their corresponding equity characteristics, then obtaining a trained prediction model 1 and a prediction model 2 according to training data, obtaining that the preset response rate of the user characteristic of the user a output by the prediction model 1 to the equity characteristic of the equity 1-element red packet is 50 percent, obtaining that the preset response rate of the user characteristic of the user b output by the prediction model 2 to the equity characteristic of the equity 2-10-element red packet is 80 percent, in an actual application scene, people generally have a lower response rate to the red packet with smaller money amount and a higher response rate to the red packet with larger money amount; at this time, the sequence of the preset response rates is [ 50% and 80% ], and the sequence result of the preset response rates is as follows: the preset response rate of the user characteristic of the user a to the right characteristic of the right 1-element red packet is smaller than the preset response rate of the user characteristic of the user b to the right characteristic of the right 2-10-element red packet;
the preset response rate of the user feature of the user a output by the prediction model 1 based on the test data to the equity feature of the equity 1-element red packet is 80%, the preset response rate of the user feature of the user b output by the prediction model 2 to the equity feature of the equity 2-10-element red packet is 30%, the test response rates are ranked in the order of [ 80% and 30% ], and the ranking result of the test response rates is as follows: the test response rate of the user characteristic of the user a to the equity characteristic of the equity 1-element red packet is greater than the test response rate of the user characteristic of the user b to the equity characteristic of the equity 2-10-element red packet;
therefore, the sequencing result of the preset response rate is different from the sequencing result of the test response rate, and obviously does not accord with the normal behavior of the user to the rights and interests, so that the prediction model 1 and the prediction model 2 need to be adjusted; the specific adjustment mode may be to replace the prediction model 1 and the prediction model 2 with other types of prediction models to ensure the rationality of the prediction result of the prediction model.
In practical application, even if the preset response rate of the user characteristics of the user a output by the prediction model 1 based on the test data to the equity characteristics of the equity 1-element red packet is 50%, and the prediction response rate is consistent with the test response rate, the test response rate of the user characteristics of the user a to the equity characteristics of the equity 1-element red packet is greater than the test response rate of the user characteristics of the user b to the equity characteristics of the equity 2-10-element red packet, the prediction model 1 needs to be adjusted, and the type of the prediction model 1 can be changed into the type of the prediction model 2, so that the overall rationality of the prediction model can be corrected.
Taking the above as an example, in practical applications, when the equity issuer issues the equity, the equity is determined according to the response rate of the equity feature by the user feature output by the prediction model, if the response rate of the user for the 1-tuple red packet output by the prediction model is higher, the equity issuer is misled to issue the equity with lower amount to the user, and in practical applications, the equity with lower amount cannot stimulate the consumption of the user, which causes the equity issuer to waste a large amount of time to issue invalid equity, thereby greatly reducing the user experience.
The training method of the prediction model provided by the embodiment of the specification learns the response rate performance of different user characteristics under different equity characteristics by using an equity model of each equity type, and in a scoring stage based on prediction of the prediction model, due to the equity modeling, the response conditions of users under each equity can be directly obtained in parallel for the same user, and the execution efficiency is high.
In addition, the safety and privacy of data can be guaranteed through rights and interests type-based rights and interests modeling, a prediction model can be trained autonomously by each rights and interests data source party through own rights and interests data, and finally, the rationality of the prediction result of the trained prediction model is determined according to the training method of the prediction model for subsequent use; moreover, through the fractional rights modeling, the training data faced by a single prediction model are rights and interests of the same rights and interests type, the prediction models corresponding to a plurality of rights and interests types directly represent the distinctiveness of the rights and interests, the modeling is carried out according to the rights and interests types and the prediction models are independent from each other, so that the problem of data imbalance among the rights and interests of different rights and interests types can be solved, when the rights and interests change, only the prediction model corresponding to the corresponding rights and interests type needs to be deleted or the prediction model corresponding to the newly added rights and interests type needs to be added, when a certain prediction model line shows fall back, only the prediction model corresponding to the corresponding interest type needs to be fitted again, meanwhile, in the subsequent prediction scoring stage of the prediction model, only the user characteristics of a certain user need to be given, the prediction model corresponding to each interest type can synchronously run in parallel, and meanwhile, the response prediction conditions of the user under different interests are given.
Referring to fig. 2, fig. 2 shows a flowchart of model training and model testing in the training method of the predictive model provided in this specification.
In practical application, "data" in fig. 2 includes user characteristics of a user and equity characteristics of equity, "data is divided according to equity types" as described in step 102 and step 104 of the above embodiment, equity types are divided into equity 1 to equity n, equity 1 to equity n are trained in parallel and synchronously according to training data of each equity type, equity 1 to equity n prediction models are obtained after training, equity 1 to equity n prediction models are tested according to test data of each equity type, test results are obtained respectively, test results of equity 1 to equity n prediction models are summarized, rationality of the trained prediction models is detected according to the summarized result, and specific test and detection methods are described in step 108 of the above embodiment, and will not be described in detail herein.
Referring to fig. 3, fig. 3 is a flowchart illustrating user response prediction after weighted interest modeling in the training method of the predictive model provided in the present specification.
In fig. 3, after obtaining the prediction models, namely the equity 1 prediction model to the equity n prediction model, based on the training method of the prediction model in this specification, first obtain the user characteristics of the user to be predicted, then input the user characteristics into the trained equity 1 prediction model to the equity n prediction model, and obtain the response rates of the user to be predicted in the equity 1 to the equity n prediction model, respectively, in parallel, and by adopting the manner of equity-divided modeling, the same user can directly obtain the response conditions of the user under each equity in parallel, and the execution efficiency is high.
Corresponding to the above method embodiment, the present specification further provides an embodiment of a training apparatus for a prediction model, and fig. 4 shows a schematic structural diagram of a training apparatus for a prediction model provided in an embodiment of the present specification.
As shown in fig. 4, the apparatus includes:
the equity division module 402 is configured to allocate an equity to each collected user and divide the equity according to the equity types;
a data partitioning module 404 configured to partition the equity features of the equity of each equity type and the user features of the corresponding user into training data and test data;
a model training module 406 configured to train an initial prediction model corresponding to each equity type based on the training data of each equity type, so as to obtain a prediction model corresponding to each equity type;
and the model adjusting module 408 is configured to input the test data of each equity type into the prediction model corresponding to each equity type to obtain a test result, and adjust the prediction model based on all the test results.
Optionally, the model training module 406 is further configured to:
and training the initial prediction model corresponding to each equity type based on the training data of each equity type to obtain a prediction model which corresponds to each equity type and outputs the response rate of the user characteristics to the equity characteristics.
Optionally, the test result includes a test response rate of the user characteristic for the benefit characteristic;
accordingly, the model adjustment module 408 is further configured to:
and inputting the test data of each equity type into the prediction model corresponding to each equity type to obtain the test response rate of the user characteristics of each equity type to the equity characteristics, and adjusting the prediction model based on the test response rate.
Optionally, the apparatus further includes:
and the preset response rate acquisition module is configured to acquire the preset response rate of the user characteristics output by the prediction model of each equity type to the equity characteristics.
Optionally, the model adjusting module 408 is further configured to:
and calculating a difference value between a preset response rate of the user characteristics output by the prediction model of each interest type aiming at the interest characteristics and a corresponding test response rate, and adjusting the prediction model based on the difference value when the difference value is greater than or equal to a preset threshold value.
Optionally, the model adjusting module 408 is further configured to:
sorting the user characteristics output by the prediction models of all the rights and interests types according to the preset response rate of the rights and interests characteristics to obtain a sorting result of the preset response rate;
sequencing the user characteristics output by the prediction models of all the rights and interests aiming at the test response rate of the rights and interests characteristics to obtain a sequencing result of the test response rate;
and under the condition that the sequencing result of the preset response rate is different from the sequencing result of the test response rate, adjusting a prediction model based on the test response rate.
Optionally, the model adjusting module 408 is further configured to:
and adjusting at least one prediction model corresponding to the ownership type based on the test response rate.
Optionally, the model training module 406 is further configured to:
and training the same initial prediction model corresponding to each equity type based on the training data of each equity type to obtain the same prediction model corresponding to each equity type.
Optionally, the model training module 406 is further configured to:
and training different initial prediction models corresponding to each equity type based on the training data of each equity type to obtain different prediction models corresponding to each equity type.
Optionally, the apparatus further includes:
an acquisition module configured to acquire a plurality of users and a plurality of rights and interests.
Optionally, the user characteristics include attribute characteristics and behavior characteristics of the user.
The training device for the prediction models provided by the embodiment of the specification optimizes the rights and interests of only one rights and interests type for each prediction model in a way of training the prediction models in a way of dividing the rights and interests, has simple data, and is good in fitting effect of each prediction model obtained by training, and the training data of each rights and interests type corresponds to one prediction model; and the prediction model can be adjusted based on the training result of the prediction model, so that the prediction rationality and accuracy of the prediction model are greatly ensured.
The above is an exemplary scheme of a training apparatus of a prediction model according to this embodiment. It should be noted that the technical solution of the training apparatus for the prediction model and the technical solution of the training method for the prediction model described above belong to the same concept, and details of the technical solution of the training apparatus for the prediction model, which are not described in detail, can be referred to the description of the technical solution of the training method for the prediction model described above.
FIG. 5 illustrates a block diagram of a computing device 500 provided in accordance with one embodiment of the present description. The components of the computing device 500 include, but are not limited to, a memory 510 and a processor 520. Processor 520 is coupled to memory 510 via bus 530, and database 550 is used to store data.
Computing device 500 also includes access device 540, access device 540 enabling computing device 500 to communicate via one or more networks 560. Examples of such networks include the Public Switched Telephone Network (PSTN), a Local Area Network (LAN), a Wide Area Network (WAN), a Personal Area Network (PAN), or a combination of communication networks such as the internet. The access device 540 may include one or more of any type of network interface, e.g., a Network Interface Card (NIC), wired or wireless, such as an IEEE802.11 Wireless Local Area Network (WLAN) wireless interface, a worldwide interoperability for microwave access (Wi-MAX) interface, an ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a bluetooth interface, a Near Field Communication (NFC) interface, and so forth.
In one embodiment of the present description, the above-described components of computing device 500, as well as other components not shown in FIG. 5, may also be connected to each other, such as by a bus. It should be understood that the block diagram of the computing device architecture shown in FIG. 5 is for purposes of example only and is not limiting as to the scope of the present description. Those skilled in the art may add or replace other components as desired.
Computing device 500 may be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., tablet, personal digital assistant, laptop, notebook, netbook, etc.), mobile phone (e.g., smartphone), wearable computing device (e.g., smartwatch, smartglasses, etc.), or other type of mobile device, or a stationary computing device such as a desktop computer or PC. Computing device 500 may also be a mobile or stationary server.
Wherein processor 520 is configured to execute the following computer-executable instructions:
distributing a right for each collected user, and dividing the right according to the right type;
dividing the equity characteristics of each equity type and the corresponding user characteristics of the user into training data and test data;
training an initial prediction model corresponding to each equity type based on the training data of each equity type to obtain a prediction model corresponding to each equity type;
and inputting the test data of each rights and interests type into the prediction model corresponding to each rights and interests type to obtain a test result, and adjusting the prediction model based on all the test results.
The above is an illustrative scheme of a computing device of the present embodiment. It should be noted that the technical solution of the computing device and the technical solution of the above-mentioned training method of the predictive model belong to the same concept, and details that are not described in detail in the technical solution of the computing device can be referred to the description of the technical solution of the above-mentioned training method of the predictive model.
An embodiment of the present specification further provides a computer readable storage medium storing computer instructions which, when executed by a processor, implement the steps of any one of the methods for training a predictive model.
The above is an illustrative scheme of a computer-readable storage medium of the present embodiment. It should be noted that the technical solution of the storage medium and the technical solution of the above-mentioned training method of the prediction model belong to the same concept, and for details that are not described in detail in the technical solution of the storage medium, reference may be made to the description of the technical solution of the above-mentioned training method of the prediction model.
The foregoing description has been directed to specific embodiments of this disclosure. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.
The computer instructions comprise computer program code which may be in the form of source code, object code, an executable file or some intermediate form, or the like. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, usb disk, removable hard disk, magnetic disk, optical disk, computer Memory, Read-Only Memory (ROM), Random Access Memory (RAM), electrical carrier wave signals, telecommunications signals, software distribution medium, and the like. It should be noted that the computer readable medium may contain content that is subject to appropriate increase or decrease as required by legislation and patent practice in jurisdictions, for example, in some jurisdictions, computer readable media does not include electrical carrier signals and telecommunications signals as is required by legislation and patent practice.
It should be noted that, for the sake of simplicity, the foregoing method embodiments are described as a series of acts, but those skilled in the art should understand that the present embodiment is not limited by the described acts, because some steps may be performed in other sequences or simultaneously according to the present embodiment. Further, those skilled in the art should also appreciate that the embodiments described in this specification are preferred embodiments and that acts and modules referred to are not necessarily required for an embodiment of the specification.
In the above embodiments, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.
The preferred embodiments of the present specification disclosed above are intended only to aid in the description of the specification. Alternative embodiments are not exhaustive and do not limit the invention to the precise embodiments described. Obviously, many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the embodiments and the practical application, to thereby enable others skilled in the art to best understand and utilize the embodiments. The specification is limited only by the claims and their full scope and equivalents.
Claims (16)
1. A method of training a predictive model, comprising:
distributing a right for each collected user, and dividing the right according to the right type;
dividing the equity characteristics of each equity type and the corresponding user characteristics of the user into training data and test data;
training an initial prediction model corresponding to each equity type based on the training data of each equity type to obtain a prediction model corresponding to each equity type;
and inputting the test data of each rights and interests type into the prediction model corresponding to each rights and interests type to obtain a test result, and adjusting the prediction model based on all the test results.
2. The method for training the prediction model according to claim 1, wherein the training the initial prediction model corresponding to each equity type based on the training data of each equity type, and obtaining the prediction model corresponding to each equity type includes:
and training the initial prediction model corresponding to each equity type based on the training data of each equity type to obtain a prediction model which corresponds to each equity type and outputs the response rate of the user characteristics to the equity characteristics.
3. The training method of the predictive model according to claim 2, wherein the test result comprises a test response rate of a user feature to a profit feature;
correspondingly, the inputting the test data of each equity type into the prediction model corresponding to each equity type to obtain the test result, and adjusting the prediction model based on all the test results includes:
and inputting the test data of each equity type into the prediction model corresponding to each equity type to obtain the test response rate of the user characteristics of each equity type to the equity characteristics, and adjusting the prediction model based on the test response rate.
4. The training method of the prediction model according to claim 3, further comprising, before the adjusting the prediction model based on the test response rate:
and acquiring the preset response rate of the user characteristics output by the prediction model of each equity type to the equity characteristics.
5. The method of training a predictive model of claim 4, the adjusting a predictive model based on the test response rate comprising:
and calculating a difference value between a preset response rate of the user characteristics output by the prediction model of each interest type aiming at the interest characteristics and a corresponding test response rate, and adjusting the prediction model based on the difference value when the difference value is greater than or equal to a preset threshold value.
6. The method of training a predictive model of claim 5, the adjusting a predictive model based on the test response rate comprising:
sorting the user characteristics output by the prediction models of all the rights and interests types according to the preset response rate of the rights and interests characteristics to obtain a sorting result of the preset response rate;
sequencing the user characteristics output by the prediction models of all the rights and interests aiming at the test response rate of the rights and interests characteristics to obtain a sequencing result of the test response rate;
and under the condition that the sequencing result of the preset response rate is different from the sequencing result of the test response rate, adjusting a prediction model based on the test response rate.
7. The method of training a predictive model of claim 6, the adjusting a predictive model based on the test response rate comprising:
and adjusting at least one prediction model corresponding to the ownership type based on the test response rate.
8. The method for training the prediction model according to claim 1, wherein the training the initial prediction model corresponding to each equity type based on the training data of each equity type, and obtaining the prediction model corresponding to each equity type includes:
and training the same initial prediction model corresponding to each equity type based on the training data of each equity type to obtain the same prediction model corresponding to each equity type.
9. The method for training the prediction model according to claim 1, wherein the training the initial prediction model corresponding to each equity type based on the training data of each equity type, and obtaining the prediction model corresponding to each equity type includes:
and training different initial prediction models corresponding to each equity type based on the training data of each equity type to obtain different prediction models corresponding to each equity type.
10. The training method of prediction model according to claim 1, before dividing the collected at least two interests according to the interest types and allocating different users to the interests of each divided interest type, further comprising:
multiple users and multiple rights are collected.
11. The training method of the prediction model according to claim 1, wherein the user characteristics include attribute characteristics and behavior characteristics of the user.
12. An apparatus for training a predictive model, comprising:
the right and interest division module is configured to distribute one right and interest to each collected user and divide the right and interest according to the right and interest type;
a data partitioning module configured to partition the equity features of the equity of each equity type and the user features of the corresponding user into training data and test data;
the model training module is configured to train the initial prediction model corresponding to each equity type based on the training data of each equity type to obtain a prediction model corresponding to each equity type;
and the model adjusting module is configured to input the test data of each interest type into the prediction model corresponding to each interest type to obtain a test result, and adjust the prediction model based on all the test results.
13. The training apparatus of the predictive model of claim 12, the model training module further configured to:
and training the initial prediction model corresponding to each equity type based on the training data of each equity type to obtain a prediction model which corresponds to each equity type and outputs the response rate of the user characteristics to the equity characteristics.
14. The training device of the predictive model according to claim 13, wherein the test result comprises a test response rate of a user characteristic to a benefit characteristic;
accordingly, the model adjustment module is further configured to:
and inputting the test data of each equity type into the prediction model corresponding to each equity type to obtain the test response rate of the user characteristics of each equity type to the equity characteristics, and adjusting the prediction model based on the test response rate.
15. A computing device, comprising:
a memory and a processor;
the memory is to store computer-executable instructions, and the processor is to execute the computer-executable instructions to:
distributing a right for each collected user, and dividing the right according to the right type;
dividing the equity characteristics of each equity type and the corresponding user characteristics of the user into training data and test data;
training an initial prediction model corresponding to each equity type based on the training data of each equity type to obtain a prediction model corresponding to each equity type;
and inputting the test data of each rights and interests type into the prediction model corresponding to each rights and interests type to obtain a test result, and adjusting the prediction model based on all the test results.
16. A computer readable storage medium storing computer instructions which, when executed by a processor, carry out the steps of the method of training a predictive model according to any one of claims 1 to 11.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010014918.7A CN111199429A (en) | 2020-01-07 | 2020-01-07 | Prediction model training method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010014918.7A CN111199429A (en) | 2020-01-07 | 2020-01-07 | Prediction model training method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111199429A true CN111199429A (en) | 2020-05-26 |
Family
ID=70746560
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010014918.7A Pending CN111199429A (en) | 2020-01-07 | 2020-01-07 | Prediction model training method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111199429A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108416624A (en) * | 2018-02-27 | 2018-08-17 | 深圳乐信软件技术有限公司 | A kind of discount coupon method for pushing, device, storage medium and intelligent terminal |
| CN109961191A (en) * | 2019-04-03 | 2019-07-02 | 北京奇艺世纪科技有限公司 | A kind of discount coupon distribution method and device |
| CN110046932A (en) * | 2019-03-29 | 2019-07-23 | 阿里巴巴集团控股有限公司 | A kind of data processing method, device and equipment |
| WO2019201310A1 (en) * | 2018-04-20 | 2019-10-24 | Beijing Didi Infinity Technology And Development Co., Ltd. | Systems and methods for coupon issuing |
-
2020
- 2020-01-07 CN CN202010014918.7A patent/CN111199429A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108416624A (en) * | 2018-02-27 | 2018-08-17 | 深圳乐信软件技术有限公司 | A kind of discount coupon method for pushing, device, storage medium and intelligent terminal |
| WO2019201310A1 (en) * | 2018-04-20 | 2019-10-24 | Beijing Didi Infinity Technology And Development Co., Ltd. | Systems and methods for coupon issuing |
| CN110046932A (en) * | 2019-03-29 | 2019-07-23 | 阿里巴巴集团控股有限公司 | A kind of data processing method, device and equipment |
| CN109961191A (en) * | 2019-04-03 | 2019-07-02 | 北京奇艺世纪科技有限公司 | A kind of discount coupon distribution method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20200327449A1 (en) | User retention platform | |
| US10565525B2 (en) | Collaborative filtering method, apparatus, server and storage medium in combination with time factor | |
| WO2020108605A1 (en) | Recommendation method, apparatus, and storage medium | |
| Chong | A two-staged SEM-neural network approach for understanding and predicting the determinants of m-commerce adoption | |
| US11816727B2 (en) | Credit scoring method and server | |
| CN109902849A (en) | User's behavior prediction method and device, behavior prediction model training method and device | |
| US11113720B2 (en) | System and a method for surveying advertisements in mobile applications | |
| US20170083937A1 (en) | Micro-moment analysis | |
| US10528959B2 (en) | Micro-moment analysis | |
| CN110555749B (en) | Credit behavior prediction method and device based on neural network | |
| CN107463580A (en) | Train clicking rate prediction model method and apparatus, clicking rate predictor method and device | |
| CN106991577A (en) | A kind of method and device for determining targeted customer | |
| CN109949089A (en) | A kind of method, apparatus and terminal of determining displaying rate | |
| Postigo-Boix et al. | A social model based on customers’ profiles for analyzing the churning process in the mobile market of data plans | |
| CN106909560A (en) | Point of interest sort method | |
| Xu et al. | Individual differences and mobile service adoption: An empirical analysis | |
| Leng et al. | Long-range social influence in phone communication networks on offline adoption decisions | |
| CN110796379B (en) | Risk assessment method, device and equipment of business channel and storage medium | |
| CN110717785A (en) | Decision method, system and device based on label distribution learning | |
| CN111199429A (en) | Prediction model training method and device | |
| CN115510318A (en) | Training method of user characterization model, user characterization method and device | |
| CN111144990B (en) | Recommendation method and system | |
| CN113487389A (en) | Information recommendation method and device | |
| CN113822698A (en) | Content pushing method and device, computer equipment and storage medium | |
| CN111882360A (en) | User group expansion method and device |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200526 |