CN112487295A - 5G package pushing method and device, electronic equipment and computer storage medium - Google Patents
5G package pushing method and device, electronic equipment and computer storage medium Download PDFInfo
- Publication number
- CN112487295A CN112487295A CN202011403554.8A CN202011403554A CN112487295A CN 112487295 A CN112487295 A CN 112487295A CN 202011403554 A CN202011403554 A CN 202011403554A CN 112487295 A CN112487295 A CN 112487295A
- Authority
- CN
- China
- Prior art keywords
- classifier
- package
- model
- determining
- constructing
- 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
- 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
- G06F18/285—Selection of pattern recognition techniques, e.g. of classifiers in a multi-classifier system
Abstract
The application provides a 5G package pushing method and device, electronic equipment and a computer storage medium. The 5G package pushing method comprises the following steps: determining 5G set of meal potential customers by utilizing a two-classification model in a preset multilayer cascade classifier; the multi-layer cascade classifier comprises a two-classification model and a cascade multi-classifier; and 5G set meal pushing is carried out by utilizing the cascade multi-classifier aiming at the 5G set meal hidden passengers. According to the embodiment of the application, the 5G package pushing can be carried out more accurately.
Description
Technical Field
The application belongs to the technical field of operator big data, and particularly relates to a 5G package pushing method and device, electronic equipment and a computer storage medium.
Background
With the accelerated arrival of 5G, operators have also successively introduced many new 5G packages in order to achieve revenue growth and attract customers. It is a challenging task to dig out potential customers who will replace a 5G package from the existing user group and recommend the most appropriate 5G package category for them.
The current mode of package recommendation of operators is mainly manual recommendation in business halls, potential users are not screened in the recommendation mode, complete user recommendation is directly performed, the recommended packages are concentrated on popular packages, incompleteness and subjective colors exist, and only marketing personnel with abundant experience can recommend proper and personalized 5G packages for the users.
The currently common recommendation technologies mainly include collaborative filtering recommendation and construction of a multi-classification model. The first collaborative filtering recommendation mode is mainly used for calculating the similarity among users according to the attributes and behavior habits of the users and recommending 5G packages of the users with high similarity to the users. The other mode is to construct a multi-classification model, take user attributes, consumption behaviors and package service conditions as user characteristics, take 5G package handling categories handled by a user as labels, and model and predict the probability of handling each type of packages by the user.
The existing package recommending mode is to recommend a whole number of users, is not targeted, has low selling efficiency and low selling success rate, and can influence the user perception. In the recommendation technology, the first method of recommending packages based on the collaborative filtering algorithm has high computational complexity and poor computational efficiency due to the large user size. In the second recommendation mode for constructing the multi-classification model, the 5G packages are various at present, and the handling conditions of the packages are extremely unbalanced, so that the prediction accuracy of the mode is low.
Therefore, how to more accurately push the 5G package is a technical problem that needs to be solved by those skilled in the art.
Disclosure of Invention
The embodiment of the application provides a 5G package pushing method and device, electronic equipment and a computer storage medium, and 5G packages can be pushed more accurately.
In a first aspect, an embodiment of the present application provides a 5G package pushing method, including:
determining 5G set of meal potential customers by utilizing a two-classification model in a preset multilayer cascade classifier; the multi-layer cascade classifier comprises a two-classification model and a cascade multi-classifier;
and 5G set meal pushing is carried out by utilizing the cascade multi-classifier aiming at the 5G set meal hidden passengers.
Optionally, before determining the 5G package of meal potential customers by using the binary classification model in the preset multilayer cascade classifier, the method further includes:
deriving the linear strong correlation characteristics based on the service and weighted combined kernel function, and determining the input characteristics of the model;
and constructing a multilayer cascade classifier based on the model input characteristics.
Optionally, deriving the linear strong correlation feature based on the service and weighted combined kernel function, and determining the model input feature includes:
deriving the linear strong correlation characteristics based on the service and weighted combined kernel function, and determining a first input characteristic;
and determining the model input characteristics by using the sample distribution weight mutual information function and the first input characteristics.
Optionally, constructing a multi-layer cascade classifier based on the model input features includes:
constructing a parent classifier in a strategy planning clustering mode based on the model input characteristics;
constructing a sub-classifier for each clustered cluster;
and obtaining the cascade multi-classifier in the multilayer cascade classifier based on the parent classifier and the child classifier.
Optionally, based on the model input features, constructing a parent classifier in a policy planning clustering manner, including:
performing data enhancement on the input features of the model by adopting a data enhancement mode;
and constructing a parent classifier in a strategy planning clustering mode based on the model input characteristics after data enhancement.
In a second aspect, an embodiment of the present application provides a 5G package pushing device, including:
the first determining module is used for determining the 5G package of latent customers by utilizing a binary classification model in a preset multilayer cascade classifier; the multi-layer cascade classifier comprises a two-classification model and a cascade multi-classifier;
and the pushing module is used for pushing the 5G package by utilizing the cascade multi-classifier aiming at the 5G package hidden passengers.
Optionally, the apparatus further comprises:
the second determining module is used for deriving the linear strong correlation characteristics based on the service and weighted combined kernel function and determining the model input characteristics;
and the construction module is used for constructing the multilayer cascade classifier based on the model input characteristics.
Optionally, the second determining module includes:
the derivation unit is used for deriving the linear strong correlation characteristics based on the service and weighted combined kernel function and determining first input characteristics;
and the determining unit is used for determining the model input characteristics by utilizing the sample distribution weight mutual information function and the first input characteristics.
Optionally, the building block includes:
the first construction unit is used for constructing a parent classifier in a strategy planning clustering mode based on the model input characteristics;
the second construction unit is used for constructing a sub-classifier for each clustered cluster;
and the obtaining unit is used for obtaining the cascade multi-classifier in the multilayer cascade classifier based on the parent classifier and the child classifier.
Optionally, the first building unit includes:
the data enhancement unit is used for enhancing the data of the input features of the model in a data enhancement mode;
and the construction subunit is used for constructing a parent classifier in a strategy planning clustering mode based on the model input characteristics after the data enhancement.
In a third aspect, an embodiment of the present application provides an electronic device, where the electronic device includes: a processor and a memory storing computer program instructions;
the processor, when executing the computer program instructions, implements the 5G package push method as shown in the first aspect.
In a fourth aspect, the present application provides a computer storage medium, where computer program instructions are stored on the computer storage medium, and when executed by a processor, the computer program instructions implement the 5G package pushing method shown in the first aspect.
The 5G package pushing method and device, the electronic equipment and the computer storage medium can more accurately push the 5G package. The 5G package pushing method determines 5G package hidden passengers by utilizing a two-classification model in a preset multilayer cascade classifier; the multi-layer cascade classifier comprises a two-classification model and a cascade multi-classifier; aiming at 5G set meal hidden passengers, the 5G set meal pushing can be more accurately carried out by utilizing the cascade multi-classifier.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments of the present application will be briefly described below, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a schematic flowchart of a 5G package pushing method according to an embodiment of the present application;
FIG. 2 is a schematic flow chart illustrating selection of initial feature factors according to an embodiment of the present application;
FIG. 3 is a schematic flow chart diagram of feature derivation provided by an embodiment of the present application;
FIG. 4 is a schematic diagram of a flow of prediction result output according to an embodiment of the present application;
fig. 5 is a schematic flowchart of a 5G package pushing method according to another embodiment of the present application;
fig. 6 is a schematic structural diagram of a 5G package pushing device provided in an embodiment of the present application;
fig. 7 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
Detailed Description
Features and exemplary embodiments of various aspects of the present application will be described in detail below, and in order to make objects, technical solutions and advantages of the present application more apparent, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are intended to be illustrative only and are not intended to be limiting. It will be apparent to one skilled in the art that the present application may be practiced without some of these specific details. The following description of the embodiments is merely intended to provide a better understanding of the present application by illustrating examples thereof.
It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
In order to solve the prior art problem, embodiments of the present application provide a method, an apparatus, a device, and a computer storage medium for pushing a 5G package. First, a method for pushing a 5G package provided in an embodiment of the present application is described below.
Fig. 1 shows a flowchart of a 5G package pushing method provided in an embodiment of the present application. As shown in fig. 1, the 5G package pushing method includes:
s101, determining 5G set of meal potential customers by utilizing a preset two-classification model in a multilayer cascade classifier; the multi-layer cascade classifier comprises a two-classification model and a cascade multi-classifier.
In one embodiment, before determining the 5G package of meal potential customers using the binary classification model in the preset multilayer cascade classifier, the method further comprises:
deriving the linear strong correlation characteristics based on the service and weighted combined kernel function, and determining the input characteristics of the model;
and constructing a multilayer cascade classifier based on the model input characteristics.
In one embodiment, deriving the linear strongly correlated features based on a business and weighted combined kernel function, determining model input features, comprises:
deriving the linear strong correlation characteristics based on the service and weighted combined kernel function, and determining a first input characteristic;
and determining the model input characteristics by using the sample distribution weight mutual information function and the first input characteristics.
In one embodiment, building a multi-tiered cascade classifier based on model input features includes:
constructing a parent classifier in a strategy planning clustering mode based on the model input characteristics;
constructing a sub-classifier for each clustered cluster;
and obtaining the cascade multi-classifier in the multilayer cascade classifier based on the parent classifier and the child classifier.
In one embodiment, constructing a parent classifier by means of policy planning clustering based on model input features includes:
performing data enhancement on the input features of the model by adopting a data enhancement mode;
and constructing a parent classifier in a strategy planning clustering mode based on the model input characteristics after data enhancement.
S102, aiming at the 5G set of latent customers, 5G set of meal pushing is carried out by utilizing the cascade multi-classifier.
The 5G package pushing method determines 5G package hidden passengers by utilizing a two-classification model in a preset multilayer cascade classifier; the multi-layer cascade classifier comprises a two-classification model and a cascade multi-classifier; aiming at 5G set meal hidden passengers, the 5G set meal pushing can be more accurately carried out by utilizing the cascade multi-classifier.
The above technical solution is explained below with a specific example.
The scheme provides a 5G package recommendation method based on a multilayer cascade classifier, and the main solution measures are as follows: 1) constructing a multilayer classifier, preferably constructing a secondary classification model to mine 5G package tenants, and constructing a cascade multi-classifier for package recommendation aiming at the 5G package tenants; 2) deriving the linear strong correlation characteristics based on the service and weighted combined kernel function, and deeply mining the nonlinear information of the linear strong correlation characteristics; 3) introducing a sample distribution weight mutual information function, and reducing the influence of the data imbalance of the 5G set meal hidden passenger mining model on effective feature selection; 4) an integrated classifier is built, and the accuracy and robustness of a 5G set meal hidden passenger mining model are improved; 5) constructing 5G package recommendation cascade multi-classifiers, constructing a parent classifier in an optimal strategy planning clustering mode, and constructing a sub-classifier for each cluster; 6) and balancing the sample data of the parent classifier in a data enhancement mode.
According to the scheme, a 5G package latent passenger mining model is firstly established according to characteristics such as user attributes, consumption behaviors and package service conditions, potential users with high probability of replacing the 5G package are preferably mined out and serve as target users for package recommendation at the next stage, the quantity of target recommended users is reduced, and the recommendation success rate is improved. And then constructing a cascade classifier to recommend the package, dividing the package into a plurality of clusters by a clustering mode according to the sample data of the existing 5G package, and then constructing a sub-classifier in each cluster to finish package recommendation prediction.
The specific implementation steps are as follows:
firstly, a multi-layer classifier is constructed, and 114 initial characteristic factors are selected. From the current user group handling the 5G package, most users are non-5G terminals, so when screening target potential customers, whether the users are 5G terminals or not is preferentially judged, and the users who are 5G terminals but do not replace the 5G package are all used as potential customers recommended by the 5G package. Inputting non-5G terminal users into a 5G package submarine passenger mining model, wherein the overall process is shown in FIG. 2, data acquisition is performed on original data, data preprocessing is performed on the original data, whether the users are 5G terminals is judged, and if yes, packages are recommended by using a package recommending module; if not, judging whether the user is a potential client of the 5G package, and if so, recommending the package by using a package recommending module.
Based on 5g package service, 114 initial characteristic factors are selected according to 9 major dimensional characteristics of user basic attributes, terminal attributes, drag attributes, consumption attributes, historical package ordering preferences, package fitness evaluation attributes, actual traffic usage behaviors, actual voice usage behaviors and APP traffic usage preferences. The relevant feature attributes are shown in table 1, where the first column is a feature broad class, and the second column is a specific feature, and because there are many features, only some of the features in each broad class are shown.
TABLE 1 characteristic Attribute Table
And secondly, deriving linear strongly-correlated characteristics. The historical package using habits of the user are mined, and the multi-month related package characteristics of the user are extracted as input factors, so that the performance of the same characteristics in multiple months can have strong linear correlation, such as actual monthly consumption, actual monthly flow consumption, actual monthly voice consumption and the like of the user. In addition, some non-multi-month features also have strong correlation, such as the current-month call duration and the current-month call number of the user. In order to ensure that some important information is not lost after such strongly correlated features are selected by features, feature derivation is performed on the above two cases in two ways of "deriving based on service features" and "deriving based on weighted combination function features", and a specific flow is shown in fig. 3:
1. deriving based on business features: and aiming at the characteristics with the multi-month values, performing related characteristic derivation by combining service priority, generating statistical values such as variance, mean value, maximum value, minimum value and the like for the characteristics with the multi-month values, and taking the derived characteristics of each variable as a model input factor.
Taking the actual total consumption mean value as an example, the actual consumption value of the user in the last three months per month can be calculated to derive the actual total consumption mean value in the last march, the actual total consumption maximum value in the last march and the actual total consumption variance in the last march.
The characteristics derived from the service characteristics are specifically shown in table 2:
TABLE 2 features based on traffic derivation
2) Deriving based on weighted combination function features: and for the non-multi-month features, mining the non-linear relation between the linear strong correlation features by means of a kernel function. The scheme is characterized in that a weighted combination kernel function is constructed to derive strong correlation characteristics, a Gaussian kernel function and a polynomial kernel function are introduced, and an optimal combination is selected through cross validationAnd (4) weighting. Gaussian kernel function kg (x)i,xj) And a polynomial kernel function kp (x)i,xj) The formula is shown in equations 1 and 2:
kg(xi,xj)＝(axi Txj+c)dequation 1
Wherein x represents a feature vector, and a, c, d and sigma are all hyper-parameters.
Final weighted combined kernel function k (x)i,xj) As shown in equation 3:
k(xi,xj)＝λ1kg(xi,xj)+λ2kp(xi,xj) Equation 3
Wherein λ1+λ21, weight λ1And λ2Obtaining an optimal solution in a cross validation mode, and obtaining an optimal solution lambda by adjusting the parameter scheme1＝0.263，λ2＝0.737。
Taking the generation of the linear strongly correlated feature combination of the total call as an example, the feature derivation formula is as follows:
total linear strongly correlated feature set 1 ═ 0.263 x kg (x)Total number of calls,xTotal call duration)+0.737*kp(xTotal number of calls,xTotal call duration)
Other relevant features that are derived based on weighted kernel functions are shown in table 3:
TABLE 3 feature derived variables based on weighted kernel function
Third step, introducing the sampleAnd a weight mutual information function is distributed, so that the influence of the unbalanced data of the 5G set meal hidden passenger mining model on effective feature selection is reduced. Because the current 5G technology is in the early stage of popularization, the 5G package user group is small, the proportion of the 5G package user to the non-5G package user can even reach 1:112, and the sample is in an extremely unbalanced condition. The 5G hidden passenger mining model is constructed by considering the characteristics of all aspects of the user, such as user basic information, terminal information, use behavior habits, use preferences, package ordering conditions and the like, and the related characteristic dimension is high, and reasonable characteristic selection is beneficial to improving the operation efficiency of the model and subsequent traceability analysis. Common feature selection methods tend to select a subset of features that are favorable for most classes, and reasonable feature extraction cannot be performed on unbalanced samples. Therefore, in order to solve the influence of sample imbalance on feature selection, the scheme adds positive and negative sample distribution as a weight to the feature selection process, and modifies the traditional mutual information formula, thereby extracting an efficient feature sequence. Assume that sample feature f and label set L ═ L1,l2},lie.L if feature f and label LiThe mutual information between them is I (f; l)i) Then, the mutual information between the feature f and the label set L is:
H(f；L)＝I(f；l1)+I(f；l2)
the larger the mutual information value, the stronger the association between the representation feature and the tag, and the more important the feature is. When the sample distribution is unbalanced, different labels describe the sample to different extents. Therefore, the scheme assigns different weights w to the labels according to the distribution condition of the positive and negative samples, wherein w+Weight, w, representing positive label-Representing the weight of the negative label, and after the weight is substituted, the mutual information between the characteristic f and the label set L is as follows:
wherein:
if the ratio of positive and negative samples in the 5G package data is 1: n, then
Taking the characteristic age as an example, the final weighted mutual information value calculation formula is as follows:
the final normalized mutual information values of the features calculated by the weighted mutual information function are shown in table 4:
TABLE 4 mutual information value table after normalization of each feature
And finally, sequencing the mutual information values after the normalization of the features, and selecting the first 60% of the features as the final effective input features of the model. The original features are added with the derived features, feature screening is carried out according to the weighted mutual information value, and finally, 81 features are input into the model, wherein the specific features are shown in table 5:
TABLE 5 characteristic Table finally input into the model
From the ranking of the characteristic mutual information values after final normalization, the characteristics of contract expiration time, network access duration, actual consumption average value in about March and historical package types of the user are important characteristics influencing the replacement of 5G packages by the user and accord with business logic.
And fourthly, constructing an integrated classifier, and improving the accuracy and robustness of the 5G set meal submarine passenger mining model. The unbalanced distribution of samples affects not only the feature selection but also the classifier accuracy. In order to reduce the influence of the sample imbalance on the precision of the classifier, an undersampling or oversampling method is generally adopted to balance the data set. Under-sampling, i.e. randomly deleting most types of sample data, so as to balance the distribution of positive and negative samples in the data set, but the random data deletion mode may cause some important information to be lost, so that the performance of the classifier is reduced. The over-sampling approach balances the majority of samples by increasing the number of minority samples, but this approach may result in model over-fitting.
Because the direct undersampling and oversampling are carried out to a certain extentThe method has the advantages that a plurality of balanced data sets are generated in a multi-time undersampling mode, namely, a complete set of a few 5G package processed samples is extracted, a data subset equal to the few 5G package processed samples is sampled for multiple times in the majority of 5G package unprocessed samples, and the sampled data subset is spliced with the few 5G package processed samples to form a plurality of training sample sets. Respectively fitting the generated balance data subsets by using a classification model, respectively performing parameter adjustment training on each base classifier, finally integrating each base classifier by adopting a bagging idea to form a combined classifier, and obtaining a final classification label in a voting mode, namely assuming that m base classifiers are integrated together, if the number of the base classifiers exceeds the number of the base classifiers, determining whether the number of the base classifiers exceeds the number of the base classifiers
And (4) the base classifier predicts that the user is a 5G package of meal potential customers and then labels whether the 5G potential customers exist or not for the user. The integrated model provided by the scheme can not only prevent some important information from being lost, but also prevent the model from being over-fitted. The specific operation flow can be seen in fig. 4.
And fifthly, constructing 5G package recommendation cascade multi-classifiers, constructing a parent classifier in an optimal strategy planning clustering mode, and constructing a sub-classifier for each cluster. After the 5G package target user screening is completed, specific package recommendation needs to be carried out on the target user. As the 5G set meal has various types and the samples of all the set meals are distributed unevenly, similar set meals can be combined, a parent classifier is constructed first, and then a child classifier is constructed. In order to more reasonably combine similar packages, a plurality of cluster clusters are formed by the samples through a DBSCAN clustering algorithm, and the cluster clusters are re-planned and adjusted according to the distribution condition of the 5G package samples in each cluster. Assuming that the DBSCAN algorithm regroups the samples into 3 types, and the sample data of package 1 accounts for 70%, 20%, and 10% of the sample data in cluster 1, cluster 2, and cluster 3, respectively, adjusting the sample of package 1 in the rest clusters to cluster 1, and obtaining the center point of each cluster again by averaging the sample point vectors in each package cluster, and so on. And after the replanning of the samples in each cluster is completed, resetting the labels of the samples in each cluster, and training a parent classifier by taking the cluster labels as classification labels. In combination with the business scene, there are 60 types of 5G packages at present, and the optimal cluster number selected by the contour coefficient is 7, that is, the existing 60 types of packages are grouped into seven categories. When package recommendation is subsequently performed on potential customers, the users are roughly classified into a certain package category through the parent classifier, and then the specific 5G package category is recommended to the users through the sub-classifiers. The cascade classifier flow chart can be seen in fig. 5.
And sixthly, balancing the sample data of the parent classifier by adopting a data enhancement mode. In the last step, after a plurality of clusters are formed through a DBSCAN clustering algorithm, the condition that the number of samples in each cluster is unbalanced is considered, and the effect of a father classifier is influenced. In order to balance the number of samples in each cluster more reasonably, a data enhancement mode is adopted by combining a central point in each cluster to expand a few types of samples in the clusters, and the specific mode is as follows:
(a) obtaining a sample central point i of a cluster where a small number of 5G package sample data are located, and marking a feature vector of the sample central point i as xi；
(b) Firstly, finding k neighbors of the central point in the cluster from all T samples of the small number of 5G package samples, and recording the k neighbors as xi(near),near∈{1,…,k}；
(c) Randomly selecting a sample x from the k neighborsi(1)Generating a random number between 0 and 1
Thereby synthesizing a new sample
(d) The previous step is repeated N times thereafter so that N new samples can be synthesized to reach an equilibrium with the majority cluster sample size.
Assuming that there are seven clusters, the original total sample size is M, and the maximum intra-cluster sample size is N, N < < M, the final total sample size is increased to 7 × N through the data enhancement.
Through a data enhancement mode of the sixth step, various sample data in the parent classifier of the 5G package recommendation model are balanced, and therefore the accuracy and the robustness of the 5G package recommendation parent classifier model are improved.
The effect of package recommendation to the full amount of users by adopting collaborative filtering and a multi-classification model is poor at present, and the advantages of the scheme are as follows:
1. constructing a multi-layer classification model, preferably mining 5G package dinning potential customers, defining a target user, and saving manpower and material resources consumed in the subsequent recommendation step;
2. and linear strong correlation characteristic derivation is carried out based on the service and the weighting kernel function, so that more effective information is fully extracted from limited characteristics.
3. The sample weight is introduced into a mutual information formula, so that the influence of unbalanced samples on feature selection is reduced, real and effective feature variables are extracted, and a powerful basis is provided for subsequent traceability analysis;
4. various model prediction results are integrated in a voting mode, so that the model robustness is ensured, and the model accuracy is improved;
5. the package recommendation is carried out by adopting the cascade classifier, so that the problems of large classification error and the like caused by multiple package categories and uneven sample distribution are effectively solved.
6. Through the optimization scheme, the mining accuracy of the 5G package potential customers and the package recommendation success rate are improved. The 5G package of diner divers mined ROC curves of a model training set and a test set, wherein the value of the training set auc is 0.928, and the value of the test set auc is 0.885. The success rate of 5G package recommendation is 0.657, and the model effect is good.
As shown in fig. 6, an embodiment of the present application further provides a 5G package pushing device, including:
the first determining module 601 is configured to determine a 5G package of latent customers by using a binary classification model in a preset multilayer cascade classifier; the multi-layer cascade classifier comprises a two-classification model and a cascade multi-classifier;
the pushing module 602 is configured to perform pushing of the 5G package by using the cascade multi-classifier for the 5G package latent passenger.
In one embodiment, the apparatus further comprises:
the second determining module is used for deriving the linear strong correlation characteristics based on the service and weighted combined kernel function and determining the model input characteristics;
and the construction module is used for constructing the multilayer cascade classifier based on the model input characteristics.
In one embodiment, the second determining module includes:
the derivation unit is used for deriving the linear strong correlation characteristics based on the service and weighted combined kernel function and determining first input characteristics;
and the determining unit is used for determining the model input characteristics by utilizing the sample distribution weight mutual information function and the first input characteristics.
In one embodiment, a building block comprises:
the first construction unit is used for constructing a parent classifier in a strategy planning clustering mode based on the model input characteristics;
the second construction unit is used for constructing a sub-classifier for each clustered cluster;
and the obtaining unit is used for obtaining the cascade multi-classifier in the multilayer cascade classifier based on the parent classifier and the child classifier.
In one embodiment, a first building element comprises:
the data enhancement unit is used for enhancing the data of the input features of the model in a data enhancement mode;
and the construction subunit is used for constructing a parent classifier in a strategy planning clustering mode based on the model input characteristics after the data enhancement.
Each module/unit in the apparatus shown in fig. 6 has a function of implementing each step in fig. 1, and can achieve the corresponding technical effect, and for brevity, the description is not repeated here.
Fig. 7 shows a schematic structural diagram of an electronic device provided in an embodiment of the present application.
The electronic device may include a processor 701 and a memory 702 storing computer program instructions.
Specifically, the processor 701 may include a Central Processing Unit (CPU), or an Application Specific Integrated Circuit (ASIC), or may be configured to implement one or more Integrated circuits of the embodiments of the present Application.
In one example, the Memory 702 may be a Read Only Memory (ROM). In one example, the ROM may be mask programmed ROM, programmable ROM (prom), erasable prom (eprom), electrically erasable prom (eeprom), electrically rewritable ROM (earom), or flash memory, or a combination of two or more of these.
The processor 701 reads and executes the computer program instructions stored in the memory 702 to implement any one of the 5G package pushing methods in the above embodiments.
In one example, the electronic device may also include a communication interface 703 and a bus 710. As shown in fig. 7, the processor 701, the memory 702, and the communication interface 703 are connected by a bus 710 to complete mutual communication.
The communication interface 703 is mainly used for implementing communication between modules, apparatuses, units and/or devices in this embodiment of the application.
In addition, the embodiment of the application can be realized by providing a computer storage medium. The computer storage medium having computer program instructions stored thereon; the computer program instructions, when executed by a processor, implement any one of the 5G package push methods in the above embodiments.
It is to be understood that the present application is not limited to the particular arrangements and instrumentality described above and shown in the attached drawings. A detailed description of known methods is omitted herein for the sake of brevity. In the above embodiments, several specific steps are described and shown as examples. However, the method processes of the present application are not limited to the specific steps described and illustrated, and those skilled in the art can make various changes, modifications, and additions or change the order between the steps after comprehending the spirit of the present application.
The functional blocks shown in the above-described structural block diagrams may be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, it may be, for example, an electronic circuit, an Application Specific Integrated Circuit (ASIC), suitable firmware, plug-in, function card, or the like. When implemented in software, the elements of the present application are the programs or code segments used to perform the required tasks. The program or code segments may be stored in a machine-readable medium or transmitted by a data signal carried in a carrier wave over a transmission medium or a communication link. A "machine-readable medium" may include any medium that can store or transfer information. Examples of a machine-readable medium include electronic circuits, semiconductor memory devices, ROM, flash memory, Erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, Radio Frequency (RF) links, and so forth. The code segments may be downloaded via computer networks such as the internet, intranet, etc.
It should also be noted that the exemplary embodiments mentioned in this application describe some methods or systems based on a series of steps or devices. However, the present application is not limited to the order of the above-described steps, that is, the steps may be performed in the order mentioned in the embodiments, may be performed in an order different from the order in the embodiments, or may be performed simultaneously.
Aspects of the present application are described above with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each block of the flowchart illustrations and/or block diagrams, and combinations of blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions/acts specified in the flowchart and/or block diagram block or blocks. Such a processor may be, but is not limited to, a general purpose processor, a special purpose processor, an application specific processor, or a field programmable logic circuit. It will also be understood that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware for performing the specified functions or acts, or combinations of special purpose hardware and computer instructions.
As described above, only the specific embodiments of the present application are provided, and it can be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working processes of the system, the module and the unit described above may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again. It should be understood that the scope of the present application is not limited thereto, and any person skilled in the art can easily conceive various equivalent modifications or substitutions within the technical scope of the present application, and these modifications or substitutions should be covered within the scope of the present application.
Claims (12)
1. A5G package pushing method is characterized by comprising the following steps:
determining 5G set of meal potential customers by utilizing a two-classification model in a preset multilayer cascade classifier; wherein the multi-layered cascaded classifier comprises the two-classification model and a cascaded multi-classifier;
and aiming at the 5G set of latent customers, pushing the 5G set of packages by utilizing the cascade multi-classifier.
2. The 5G package pushing method according to claim 1, wherein before the determining of the 5G package hidden passengers by using the binary classification model in the preset multilayer cascade classifier, the method further comprises:
deriving the linear strong correlation characteristics based on the service and weighted combined kernel function, and determining the input characteristics of the model;
and constructing the multilayer cascade classifier based on the model input features.
3. The 5G package pushing method according to claim 2, wherein the deriving the linear strongly correlated features based on the service and weighted combined kernel function to determine the model input features comprises:
deriving the linear strong correlation characteristics based on the service and weighted combined kernel function, and determining a first input characteristic;
and determining the model input characteristics by utilizing a sample distribution weight mutual information function and the first input characteristics.
4. The 5G package pushing method according to claim 2, wherein the building the multilayer cascade classifier based on the model input features comprises:
constructing a parent classifier in a strategy planning clustering mode based on the model input characteristics;
constructing a sub-classifier for each clustered cluster;
and obtaining the cascade multi-classifier in the multilayer cascade classifier based on the parent classifier and the sub-classifier.
5. The 5G package pushing method according to claim 4, wherein the constructing a parent classifier by means of policy planning clustering based on the model input features comprises:
performing data enhancement on the model input features by adopting a data enhancement mode;
and constructing the parent classifier in a strategy planning clustering mode based on the model input characteristics after data enhancement.
6. The utility model provides a 5G package pusher which characterized in that includes:
the first determining module is used for determining the 5G package of latent customers by utilizing a binary classification model in a preset multilayer cascade classifier; wherein the multi-layered cascaded classifier comprises the two-classification model and a cascaded multi-classifier;
and the pushing module is used for pushing the 5G package by utilizing the cascade multi-classifier aiming at the 5G package latent passenger.
7. The 5G package pushing device of claim 6, further comprising:
the second determining module is used for deriving the linear strong correlation characteristics based on the service and weighted combined kernel function and determining the model input characteristics;
and the construction module is used for constructing the multilayer cascade classifier based on the model input characteristics.
8. The 5G package pushing device according to claim 7, wherein the second determining module comprises:
the derivation unit is used for deriving the linear strong correlation characteristics based on the service and weighted combined kernel function and determining first input characteristics;
a determining unit, configured to determine the model input feature by using a sample distribution weight mutual information function and the first input feature.
9. The 5G package pushing device of claim 7, wherein the building module comprises:
the first construction unit is used for constructing a parent classifier in a strategy planning clustering mode based on the model input characteristics;
the second construction unit is used for constructing a sub-classifier for each clustered cluster;
an obtaining unit, configured to obtain the cascaded multi-classifier in the multi-layered cascaded classifier based on the parent classifier and the sub-classifier.
10. The 5G package pushing device according to claim 9, wherein the first building unit comprises:
the data enhancement unit is used for performing data enhancement on the model input characteristics in a data enhancement mode;
and the construction subunit is used for constructing the parent classifier in a strategy planning clustering mode based on the model input characteristics after data enhancement.
11. An electronic device, characterized in that the electronic device comprises: a processor and a memory storing computer program instructions;
the processor, when executing the computer program instructions, implements the 5G package push method of any of claims 1-5.
12. A computer storage medium having computer program instructions stored thereon, wherein the computer program instructions, when executed by a processor, implement the 5G package push method according to any one of claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011403554.8A CN112487295A (en) | 2020-12-04 | 2020-12-04 | 5G package pushing method and device, electronic equipment and computer storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011403554.8A CN112487295A (en) | 2020-12-04 | 2020-12-04 | 5G package pushing method and device, electronic equipment and computer storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112487295A true CN112487295A (en) | 2021-03-12 |
Family
ID=74939486
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011403554.8A Pending CN112487295A (en) | 2020-12-04 | 2020-12-04 | 5G package pushing method and device, electronic equipment and computer storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112487295A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113010785A (en) * | 2021-03-18 | 2021-06-22 | 中国联合网络通信集团有限公司 | User recommendation method and device |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101964059A (en) * | 2009-07-24 | 2011-02-02 | 富士通株式会社 | Method for constructing cascade classifier, method and device for recognizing object |
| CN104766221A (en) * | 2015-03-25 | 2015-07-08 | 华迪计算机集团有限公司 | Method for selecting flow refining marketing parameters |
| CN107403198A (en) * | 2017-07-31 | 2017-11-28 | 广州探迹科技有限公司 | A kind of official website recognition methods based on cascade classifier |
| US20190114677A1 (en) * | 2017-10-13 | 2019-04-18 | Yahoo Holdings, Inc. | Systems and Methods for User Propensity Classification and Online Auction Design |
| CN110019924A (en) * | 2017-08-14 | 2019-07-16 | 中兴通讯股份有限公司 | A kind of method, apparatus of song recommendations, computer equipment and storage medium |
| CN110191085A (en) * | 2019-04-09 | 2019-08-30 | 中国科学院计算机网络信息中心 | Based on polytypic intrusion detection method, device and storage medium |
| CN110378739A (en) * | 2019-07-23 | 2019-10-25 | 中国联合网络通信集团有限公司 | A kind of data traffic matching process and device |
| CN110619349A (en) * | 2019-08-12 | 2019-12-27 | 深圳市识农智能科技有限公司 | Plant image classification method and device |
| CN110955828A (en) * | 2019-11-19 | 2020-04-03 | 浙江鸿程计算机系统有限公司 | Multi-factor embedded personalized package recommendation method based on deep neural network |
| CN111078878A (en) * | 2019-12-06 | 2020-04-28 | 北京百度网讯科技有限公司 | Text processing method, device and equipment and computer readable storage medium |
| CN111209977A (en) * | 2020-01-16 | 2020-05-29 | 北京百度网讯科技有限公司 | Method, apparatus, device and medium for training and using classification model |
| CN111639584A (en) * | 2020-05-26 | 2020-09-08 | 深圳壹账通智能科技有限公司 | Risk identification method and device based on multiple classifiers and computer equipment |
-
2020
- 2020-12-04 CN CN202011403554.8A patent/CN112487295A/en active Pending
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101964059A (en) * | 2009-07-24 | 2011-02-02 | 富士通株式会社 | Method for constructing cascade classifier, method and device for recognizing object |
| CN104766221A (en) * | 2015-03-25 | 2015-07-08 | 华迪计算机集团有限公司 | Method for selecting flow refining marketing parameters |
| CN107403198A (en) * | 2017-07-31 | 2017-11-28 | 广州探迹科技有限公司 | A kind of official website recognition methods based on cascade classifier |
| CN110019924A (en) * | 2017-08-14 | 2019-07-16 | 中兴通讯股份有限公司 | A kind of method, apparatus of song recommendations, computer equipment and storage medium |
| US20190114677A1 (en) * | 2017-10-13 | 2019-04-18 | Yahoo Holdings, Inc. | Systems and Methods for User Propensity Classification and Online Auction Design |
| CN110191085A (en) * | 2019-04-09 | 2019-08-30 | 中国科学院计算机网络信息中心 | Based on polytypic intrusion detection method, device and storage medium |
| CN110378739A (en) * | 2019-07-23 | 2019-10-25 | 中国联合网络通信集团有限公司 | A kind of data traffic matching process and device |
| CN110619349A (en) * | 2019-08-12 | 2019-12-27 | 深圳市识农智能科技有限公司 | Plant image classification method and device |
| CN110955828A (en) * | 2019-11-19 | 2020-04-03 | 浙江鸿程计算机系统有限公司 | Multi-factor embedded personalized package recommendation method based on deep neural network |
| CN111078878A (en) * | 2019-12-06 | 2020-04-28 | 北京百度网讯科技有限公司 | Text processing method, device and equipment and computer readable storage medium |
| CN111209977A (en) * | 2020-01-16 | 2020-05-29 | 北京百度网讯科技有限公司 | Method, apparatus, device and medium for training and using classification model |
| CN111639584A (en) * | 2020-05-26 | 2020-09-08 | 深圳壹账通智能科技有限公司 | Risk identification method and device based on multiple classifiers and computer equipment |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113010785A (en) * | 2021-03-18 | 2021-06-22 | 中国联合网络通信集团有限公司 | User recommendation method and device |
| CN113010785B (en) * | 2021-03-18 | 2023-06-30 | 中国联合网络通信集团有限公司 | User recommendation method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107766929B (en) | Model analysis method and device | |
| CN111291816B (en) | Method and device for carrying out feature processing aiming at user classification model | |
| CN108921604B (en) | Advertisement click rate prediction method based on cost-sensitive classifier integration | |
| CN110310114B (en) | Object classification method, device, server and storage medium | |
| WO2016155493A1 (en) | Data processing method and apparatus | |
| CN108304853A (en) | Acquisition methods, device, storage medium and the electronic device for the degree of correlation of playing | |
| CN111797320B (en) | Data processing method, device, equipment and storage medium | |
| WO2020024444A1 (en) | Group performance grade recognition method and apparatus, and storage medium and computer device | |
| CN113543117B (en) | Prediction method and device for number portability user and computing equipment | |
| CN111210072A (en) | Prediction model training and user resource limit determining method and device | |
| CN109460872B (en) | Mobile communication user loss imbalance data prediction method | |
| CN107194815B (en) | Client segmentation method and system | |
| CN112487295A (en) | 5G package pushing method and device, electronic equipment and computer storage medium | |
| CN110196805A (en) | Data processing method, device, storage medium and electronic device | |
| CN104766221A (en) | Method for selecting flow refining marketing parameters | |
| CN110807159B (en) | Data marking method and device, storage medium and electronic equipment | |
| CN106447397A (en) | Tobacco retail customer pricing method based on decision tree algorithm | |
| CN111831892A (en) | Information recommendation method, information recommendation device, server and storage medium | |
| CN110880117A (en) | False service identification method, device, equipment and storage medium | |
| CN111984842B (en) | Bank customer data processing method and device | |
| CN112214675A (en) | Method, device and equipment for determining user machine purchasing and computer storage medium | |
| CN109308565B (en) | Crowd performance grade identification method and device, storage medium and computer equipment | |
| CN107291722B (en) | Descriptor classification method and device | |
| CN115982634A (en) | Application program classification method and device, electronic equipment and computer program product | |
| CN111382210B (en) | Classification method, device and equipment |
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 |