CN104866626B - A kind of recommendation method and device of telecommunication service - Google Patents
A kind of recommendation method and device of telecommunication service Download PDFInfo
- Publication number
- CN104866626B CN104866626B CN201510329668.5A CN201510329668A CN104866626B CN 104866626 B CN104866626 B CN 104866626B CN 201510329668 A CN201510329668 A CN 201510329668A CN 104866626 B CN104866626 B CN 104866626B
- Authority
- CN
- China
- Prior art keywords
- telecommunication service
- testing accuracy
- individual
- user
- population
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
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
Abstract
The embodiment provides a kind of recommendation method and device of telecommunication service, wherein this method includes:According to the historical information of target user and be previously obtained for training result that user is examined whether to receive telecommunication service to be recommended, telecommunication service matching prediction is carried out to target user, obtains the result that telecommunication service matching is predicted;According to telecommunication service matching prediction as a result, identifying the user that can receive telecommunication service to be recommended;Telecommunication service recommendation information is sent to the user that can receive telecommunication service to be recommended.The embodiment of the present invention accurately can recommend telecommunication service to user.
Description
Technical field
The present invention relates to field of communication technology, the recommendation method and device of more particularly to a kind of telecommunication service.
Background technology
At this stage, it is greatly reduced along with the universal and charging level of data traffic consumption, three big telecom operators pass
The telecommunication service of system is faced with the destiny of homogeneous competition so that telecom operators must be to intensive management and lean operation
Transition, from previous dependence traditional media, such as the public mould of marketing that TV, broadcast, newspaper, outdoor advertising " blanket type " bomb
Formula to the demand analysed in depth with holding user, is accurately positioned out the type of service of user's adaptation, thus to " being found for client
The accurate marketing Mode change of the business of suitable himself ".
At present, each basic telecom carrier all establishes respective operation analysis system, relies primarily on following several ways
To assist carrying out accurate marketing excavation:
Mode one:Traditional and improved business association rule and method forms multiple business association items to assist carrying out
Marketing is excavated;
Mode two:Basic document based on the telecommunication user, brand classification (Global Link, walk in the Divine Land, M-ZONE), history
The basic informations such as month spending limit carry out simple analysis and data mining.
For mode one, due to only considered the relevance between business in which, not from the angle of client,
Consider the true consumption needs of client, be reflection be in this way service developer thinks recommendable business,
Obviously it is such business recommended the result is that drawn from service developer angle, be not very accurately, can not really meet
The business demand of client.
For mode two, since the real demand of user can not be from the basic document of itself, client's set meal classification, the history moon
It is obtained in spending limit, for example, the client of monthly high consumption is not all interested in expensive business in history, equally in history
Low consumption client also and is not all interested in cheap business.It is clear that such business recommended result is also inaccurate
, it also can not really meet client's real demand.
Invention content
A kind of recommendation method and device for being designed to provide telecommunication service of the embodiment of the present invention, can accurately to
Recommend telecommunication service in family.
In order to achieve the above object, the embodiment provides a kind of recommendation method of telecommunication service, this method packets
It includes:
According to the historical information of target user and be previously obtained for examining whether user can receive telecommunications industry to be recommended
The training result of business carries out target user telecommunication service matching prediction, obtains the result of telecommunication service matching prediction;
According to telecommunication service matching prediction as a result, identifying the user that can receive telecommunication service to be recommended;
Telecommunication service recommendation information is sent to the user that can receive telecommunication service to be recommended.
Wherein, according to telecommunication service matching prediction as a result, identifying the user that can receive telecommunication service to be recommended, specifically
Including:
Judge whether the result of telecommunication service matching prediction is more than preset value;
If the result of telecommunication service matching prediction is more than preset value, it is determined that the target user is can receive telecommunications to be recommended
The user of business.
Wherein, it is to be recommended for user to be examined whether to receive with being previously obtained in the historical information according to target user
The training result of telecommunication service carries out target user telecommunication service matching prediction, obtains the result of telecommunication service matching prediction
Before, method further includes:
It obtains to examine whether user can receive the training result of telecommunication service to be recommended.
Wherein, the training result for user to be examined whether to receive telecommunication service to be recommended is obtained, is specifically included:
Obtain the required basic data of training;
Basic data is pre-processed, obtains intermediate data;
Intermediate data is trained by genetic algorithm and neural network algorithm, obtains training result.
Wherein, basic data is pre-processed, obtains intermediate data, specifically included:
Basic data is screened according to preset screening requirement, the data after being screened;
According to preset data format, the data after screening are ranked up;
Data after sequence are normalized, obtain the mediant for the data structure that there is genetic algorithm can identify
According to.
Wherein, intermediate data is trained by genetic algorithm and neural network algorithm, obtains training result, it is specific to wrap
It includes:
Operation is iterated to intermediate data by genetic algorithm, obtains initial population;
Initial population is decoded, obtains each individual in initial population;
The testing accuracy of each individual in initial population is optimized by neural network algorithm, obtains intermediate group
Body;
Judge the individual for being more than default testing accuracy in transitional population with the presence or absence of testing accuracy;
If it is tied in transitional population there are the individual that testing accuracy is more than default testing accuracy using the individual as training
Fruit.
Wherein, method further includes:
If there is no the individuals that testing accuracy is more than default testing accuracy, basis in transitional population to preset testing accuracy,
Transitional population is selected by genetic algorithm, is intersected and mutation operation, obtains new group;
New group is decoded, and passes through neural network algorithm and the testing accuracy of decoded each individual is carried out
Optimization, obtains new transitional population;
Judge the individual for being more than default testing accuracy in new transitional population with the presence or absence of testing accuracy;
If there are the individual that testing accuracy is more than default testing accuracy in new transitional population, using the individual as training
As a result;
If hereditary calculation is continued through there is no the individual that testing accuracy is more than default testing accuracy in new transitional population
Method selects new transitional population, intersects and mutation operation, obtains new group, and continue through neural network algorithm pair
The testing accuracy of each individual in new group optimizes, and obtains new transitional population, until being deposited in new transitional population
It is more than the individual of default testing accuracy, and using the individual as training result in testing accuracy.
Wherein, the testing accuracy of each individual in new group is optimized continuing through neural network algorithm,
After obtaining new transitional population, method further includes:
Judge whether the interative computation number of genetic algorithm reaches preset times;
If the interative computation number of genetic algorithm reaches preset times, deconditioning, and last time operation is obtained
New transitional population in testing accuracy it is highest individual be used as training result.
The embodiment of the present invention additionally provides a kind of recommendation apparatus of telecommunication service, which includes:
Prediction module, for the historical information according to target user and be previously obtained for examining whether user can receive
The training result of telecommunication service to be recommended carries out target user telecommunication service matching prediction, obtains telecommunication service matching prediction
Result;
Identification module, for according to telecommunication service matching prediction as a result, telecommunication service to be recommended can be received by identifying
User;
Sending module, for sending telecommunication service recommendation information to the user that can receive telecommunication service to be recommended.
Wherein, identification module includes:
Whether judging unit, the result for judging telecommunication service matching prediction are more than preset value, and when telecommunication service
When result with prediction is more than preset value, determination unit is triggered;
Determination unit for the triggering according to judging unit, determines the target user for that can receive telecommunication service to be recommended
User.
Wherein, device further includes:
Acquisition module, for obtaining the training result for user to be examined whether to receive telecommunication service to be recommended.
Wherein, acquisition module includes:
Acquiring unit, for obtaining the required basic data of training;
Processing unit for being pre-processed to basic data, obtains intermediate data;
Training unit is trained intermediate data for passing through genetic algorithm and neural network algorithm, obtains training knot
Fruit.
Wherein, processing unit includes:
Subelement is screened, for being screened according to preset screening requirement to basic data, the data after being screened;
Sort subelement, for according to preset data format, being ranked up to the data after screening;
Normalizing unit for the data after sequence to be normalized, obtains the number that there is genetic algorithm can identify
According to the intermediate data of structure.
Wherein, training unit includes:
Operation subelement is iterated operation to intermediate data for passing through genetic algorithm, obtains initial population;
Decoding subunit for being decoded to initial population, obtains each individual in initial population;
Optimize subelement, it is excellent to the testing accuracy progress of each individual in initial population for passing through neural network algorithm
Change, obtain transitional population;
Judgment sub-unit, for judging to be more than the individual of default testing accuracy in transitional population with the presence or absence of testing accuracy,
And when being more than the individual for presetting testing accuracy there are testing accuracy in transitional population, trigger the first determination subelement;
First determination subelement, for the triggering according to judgment sub-unit, using the individual as training result.
Wherein, device further includes:
First operation module, for when testing accuracy being not present in transitional population being more than the individual of default testing accuracy,
According to default testing accuracy, transitional population is selected by genetic algorithm, is intersected and mutation operation, obtains new group;
Decoder module for being decoded to new group, and passes through neural network algorithm to decoded each individual
Testing accuracy optimize, obtain new transitional population;
First judgment module, for judging to be more than default testing accuracy with the presence or absence of testing accuracy in new transitional population
Individual, and when being more than the individual for presetting testing accuracy there are testing accuracy in new transitional population, determining module is triggered, when new
Transitional population in there is no during the individual that testing accuracy is more than default testing accuracy, trigger the second operation module;
Determining module, for the triggering according to the first judgment module, using the individual as training result;
Second operation module for the triggering according to the first judgment module, continues through genetic algorithm to new intermediate group
Body selected, is intersected and mutation operation, obtains new group, and continue through neural network algorithm to every in new group
The testing accuracy of individual optimizes, and obtains new transitional population, to be more than until there are testing accuracies in new transitional population
The individual of default testing accuracy, and using the individual as training result.
Wherein, device further includes:
Whether the second judgment module, the interative computation number for judging genetic algorithm reach preset times, and when heredity
When the interative computation number of algorithm reaches preset times, stopping modular is triggered;
Stopping modular, for the triggering according to the second judgment module, deconditioning, and last time operation obtained new
Transitional population in testing accuracy it is highest individual be used as training result.
The said program of the present invention includes at least following advantageous effect:
In an embodiment of the present invention, it is used for whether examining user with what is be previously obtained by the historical information of target user
It can receive the training result of telecommunication service to be recommended, the user that can receive telecommunication service to be recommended is identified from target user,
And telecommunication service recommendation information is sent to the user that can receive telecommunication service to be recommended, solve accurately to recommend to user
The problem of telecommunication service, has achieved the effect that accurately to recommend telecommunication service to user.
Description of the drawings
Fig. 1 is the flow chart of the recommendation method of telecommunication service in first embodiment of the invention;
Fig. 2 is the flow chart of the recommendation method of telecommunication service in second embodiment of the invention;
Fig. 3 is the flow chart of the recommendation method of telecommunication service in third embodiment of the invention;
Fig. 4 is to obtain the flow chart of training result in third embodiment of the invention;
Fig. 5 is the structure diagram of the recommendation apparatus of telecommunication service in fourth embodiment of the invention.
Specific embodiment
The exemplary embodiment of the disclosure is more fully described below with reference to accompanying drawings.Although the disclosure is shown in attached drawing
Exemplary embodiment, it being understood, however, that may be realized in various forms the disclosure without should be by embodiments set forth here
It is limited.On the contrary, these embodiments are provided to facilitate a more thoroughly understanding of the present invention, and can be by the scope of the present disclosure
Completely it is communicated to those skilled in the art.
First embodiment
As shown in Figure 1, the first embodiment of the present invention provides a kind of recommendation method of telecommunication service, this method includes:
Step S11, according to the historical information of target user and be previously obtained for examining whether user can receive to wait to push away
The training result of telecommunication service is recommended, telecommunication service matching prediction is carried out to target user, obtains the knot of telecommunication service matching prediction
Fruit;
In the first embodiment of the present invention, the historical information of above-mentioned target user mainly includes customer flow information data
With user account basic information, wherein customer flow information data mainly by gateway be grouped wireless service technology (GPRS,
General Packet Radio Service) supporting node collects, and user account basic information is mainly transported by business
Battalion's support system (BOSS) collects.
Step S12, according to telecommunication service matching prediction as a result, identifying the user that can receive telecommunication service to be recommended;
Step S13 sends telecommunication service recommendation information to the user that can receive telecommunication service to be recommended.
It in the first embodiment of the present invention, can be by way of short message to the user that can receive telecommunication service to be recommended
Send telecommunication service recommendation information.It is understood that it is not limited in the first embodiment of the present invention to can receive to wait to push away
The user for recommending telecommunication service sends the concrete mode of telecommunication service recommendation information.
It in the first embodiment of the present invention, can will be for examining whether user can receive the instruction of telecommunication service to be recommended
Practice result and be interpreted as a kind of function, therefore when the historical information of target user is imported wherein, a telecommunications can be obtained
The result of business matching prediction.And then according to obtain telecommunication service matching prediction result come judge the target user whether be
It can receive the user of telecommunication service to be recommended, when it is that can receive the user of telecommunication service to be recommended, be sent to telecommunications industry
Business recommendation information.The purpose for accurately recommending telecommunication service to user is just realized in this way.
Second embodiment
As shown in Fig. 2, the second embodiment of the present invention provides a kind of recommendation method of telecommunication service, this method includes:
Step S21, according to the historical information of target user and be previously obtained for examining whether user can receive to wait to push away
The training result of telecommunication service is recommended, telecommunication service matching prediction is carried out to target user, obtains the knot of telecommunication service matching prediction
Fruit;
In the second embodiment of the present invention, the historical information of above-mentioned target user mainly includes customer flow information data
With user account basic information, wherein customer flow information data mainly by gateway be grouped wireless service technology (GPRS,
General Packet Radio Service) supporting node collects, and user account basic information is mainly transported by business
Battalion's support system (BOSS) collects.
Step S22, judges whether the result of telecommunication service matching prediction is more than preset value;
Step S23, if the result of telecommunication service matching prediction is more than preset value, it is determined that the target user is can receive to treat
Recommend the user of telecommunication service;
Step S24 sends telecommunication service recommendation information to the user that can receive telecommunication service to be recommended.
It, can be by way of short message to the user that can receive telecommunication service to be recommended in the second embodiment of the present invention
Send telecommunication service recommendation information.It is understood that it is not limited in the second embodiment of the present invention to can receive to wait to push away
The user for recommending telecommunication service sends the concrete mode of telecommunication service recommendation information.
It, can will be for examining whether user can receive the instruction of telecommunication service to be recommended in the second embodiment of the present invention
Practice result and be interpreted as a kind of function, therefore when the historical information of target user is imported wherein, a telecommunications can be obtained
The result of business matching prediction.And then judge whether the result of telecommunication service matching prediction is more than preset value (such as 0.8), when
When it is more than preset value, it is believed that the target user is that can receive the user of telecommunication service to be recommended, and be sent to telecommunication service
Recommendation information.If the result of certain telecommunication service matching prediction is less than preset value, then it is assumed that the target user is will not receive
The user of telecommunication service to be recommended does not need to be sent to telecommunication service recommendation information, so as to fulfill accurate business marketing mould
Formula achievees the purpose that accurately to recommend telecommunication service to user.
3rd embodiment
As shown in figure 3, the third embodiment of the present invention provides a kind of recommendation method of telecommunication service, this method includes:
Step S31 obtains the training result for user to be examined whether to receive telecommunication service to be recommended;
Step S32, according to the historical information of target user and be previously obtained for examining whether user can receive to wait to push away
The training result of telecommunication service is recommended, telecommunication service matching prediction is carried out to target user, obtains the knot of telecommunication service matching prediction
Fruit;
In the third embodiment of the present invention, the historical information of above-mentioned target user mainly includes customer flow information data
With user account basic information, wherein customer flow information data mainly by gateway be grouped wireless service technology (GPRS,
General Packet Radio Service) supporting node collects, and user account basic information is mainly transported by business
Battalion's support system (BOSS) collects.
Step S33, according to telecommunication service matching prediction as a result, identifying the user that can receive telecommunication service to be recommended;
Step S34 sends telecommunication service recommendation information to the user that can receive telecommunication service to be recommended.
It, can be by way of short message to the user that can receive telecommunication service to be recommended in the third embodiment of the present invention
Send telecommunication service recommendation information.It is understood that it is not limited in the third embodiment of the present invention to can receive to wait to push away
The user for recommending telecommunication service sends the concrete mode of telecommunication service recommendation information.
It, can will be for examining whether user can receive the instruction of telecommunication service to be recommended in the third embodiment of the present invention
Practice result and be interpreted as a kind of function, therefore when the historical information of target user is imported wherein, a telecommunications can be obtained
The result of business matching prediction.And then according to obtain telecommunication service matching prediction result come judge the target user whether be
It can receive the user of telecommunication service to be recommended, when it is that can receive the user of telecommunication service to be recommended, be sent to telecommunications industry
Business recommendation information.Accurate business marketing pattern just can be realized in this way, reached and accurately recommended telecommunication service to user
Purpose.
Wherein, in the third embodiment of the present invention, above-mentioned steps S31 is specifically included:It is required that training is obtained first
Basic data;Then basic data is pre-processed, obtains intermediate data;Finally by genetic algorithm and neural network algorithm
Intermediate data is trained, obtains training result.
In the third embodiment of the present invention, the required basic data of training mainly includes:A part is ordered to be recommended
The user of telecommunication service and a part are not subscribed to the customer flow information data and user account of the user of telecommunication service to be recommended
Basic information, wherein customer flow information data are mainly grouped wireless service technology (GPRS, General Packet by gateway
Radio Service) supporting node collects, and user account basic information mainly passes through business operation support system (BOSS)
It collects.Correspondingly, after the required basic data of training is got, it can be pre-processed, obtains intermediate data,
Intermediate data is trained finally by genetic algorithm and neural network algorithm, obtains training result, as specific training
Process can elaborate later.
In the third embodiment of the present invention, by taking 10 yuan of flow upshift set meals (hereinafter referred to as A set meals) as an example, elaborate
The recommendation method of above-mentioned telecommunication service.Find there are eight attribute columns to order A with user in user bill by association analysis early period
There are the larger degrees of association for set meal.So as to use the user bill data in Heilongjiang Province's 1-6 months in 2014, therefrom extract
Basic data of 5000000 Subscriber Number tickets (including ordering A set meals user and being not subscribed to A set meals user) as training, and
Basic data is pre-processed, then the data by pretreatment are instructed by genetic algorithm and neural network algorithm again
Practice, obtain training result.And when needing to recommend A set meals to the user for being not subscribed to A set meals, in order to make third embodiment of the invention
In telecommunication service recommendation method accuracy it is more distinct, extract be not subscribed to 600,000 users 2014 of A set meals herein
Year 1-6 month call bill datas wherein receive the user of A set meals to identify.It is divided into 3 groups, every group of 200,000 users use respectively
Algorithm in the third embodiment of the present invention, neural network algorithm, artificial experience method identify that the meeting respectively predicted receives A set meals
User, respectively take out 5000 numbers in the user that can receive A set meals from three groups, the user that can receive A set meals for these sends out
10086 port A set meals marketing short messages are sent, user can handle by replying letter, from BOSS system statistics user's order situations
As shown in table 1.In addition, similar, with more than same method, 100,000 are not subscribed to the user of emotional affection net telephone expenses set meal into
Row identification, obtains the user that 29744 meetings receive emotional affection net telephone expenses set meal, these can be received with the user of emotional affection net telephone expenses set meal
It carries out emotional affection net and orders sending short messages in groups, share 10321 family users and the business is ordered by uplink short message mode, discrimination reaches
34.7%, and normal short message mass-sending marketing user's order rate is 1~5%.It follows that third embodiment of the invention provided
The recommendation method of telecommunication service is more much more accurate than the method for conventional recommendation telecommunication service.
In 3rd embodiment | Neural network is calculated | Artificial experience method |
Algorithm | Method | ||
Subscribed users number | 454 | 339 | 246 |
Recognition accuracy (%) | 9.08% | 6.78% | 4.92% |
Table 1
Wherein, it is above-mentioned that basic data is pre-processed in the third embodiment of the present invention, intermediate data is obtained, is had
Body includes:Basic data is screened according to preset screening requirement first, follow-up genetic algorithm is removed and neural network is calculated
Unwanted data in method, the data after being screened;Then according to preset data format, the data after screening are arranged
Sequence, that is, the data after screening are spliced into data line by line, and uniformly stored according to preset data format;To row
Data after sequence are normalized, will the absolute value of each attribute value become certain relative value relationship, obtain having and lose
The intermediate data for the data structure that propagation algorithm can identify.
Intermediate data is trained by genetic algorithm and neural network algorithm illustrating, before obtaining training result,
First simply illustrate genetic algorithm and the method that neural network algorithm is used in combination.It is as follows:
(1) initial weight distribution is optimized by genetic algorithm, some preferable search is oriented in solution space
Space:
1. parameter coding, by random initializtion connection weight (wi) and Node B threshold (¢ i) by coded representation into heredity
The genotype string structure data in space, as shown in table 2:
W1 | .. | Wk | ¢ 1 | .. | ¢ m |
Table 2
2. enabling t=0, initial population is determined, the node serial number of neural network, connection weight and threshold value are formed into one group of number
According to representing an individual, several such data groups just form initial population.
3. calculating fitness function F, based on the network energy function E of the output node error of neural network, F is selected
=C/E, C is a constant in formula.
4. selecting (duplication) operation, the optimal result to prevent from having searched is lost, fitness in previous generation groups
Maximum 10% is entered directly into next-generation group.In addition the duplication ratio of 90% individual is determined by fitness:
5. swapping operation, the probability P of exchange is suitably chosenc, generally preferably 0.85.
6. carrying out mutation operation, effect is to prevent from losing useful possibility solution, and algorithm is made to have global convergence.
7. enabling t=t+1, continue to calculate fitness fi, i=1,2 ... n.
8. judging whether end condition meets, i.e. t=N (N is the iterations specified in advance) then enables s=0, into (2),
Otherwise it returns 4..
(2) each new individual uses several steps of neural network algorithm iteration in the group obtained to (1) step, then again with something lost
Propagation algorithm is operated:
1. enabling s=s+1, each individual in group is decoded, obtains the connection weight and threshold value of corresponding network;
According to local maxima frequency of training, the connection weight and threshold value of the network are modified by neural network algorithm, then will
The connection weight and threshold value of network carry out coding and generate corresponding new individual, and all these new individuals form a new group.
2. for new group, individual adaptation degree calculating is carried out, if the error in group corresponding to the individual of fitness maximum
Less than lower limit or s=M (M be specify in advance iterations), then it is transferred to 3.;Otherwise, new group is selected (duplication)
1. operation, swap operation, mutation operation turn to.
3. the maximum individual of fitness is decoded, the connection weight and threshold value of required network are obtained.
In entire searching process, the effect of each method is all limited in the range of one " part ", so as to a side
Face is ensured the global convergence of study by genetic algorithm, and Gauss-Newton method is overcome to ask the dependence and local convergence of initial value
Topic;On the other hand, and the combination of " accurate " Gauss-Newton learning algorithm also overcome simple genetic algorithm carried it is random
Property and probability sex chromosome mosaicism, and help to improve its search efficiency.Result of which will be expected to significantly improve the convergence of algorithm
Property, ensure that it restrains direction while its degree of dependence to primary condition is weakened, even if to feelings limited known to problem
Under condition, remain able to obtain satisfied training result.
As shown in figure 4, in the third embodiment of the present invention, above by genetic algorithm and neural network algorithm to centre
Data are trained, and obtain training result, are specifically included:
Step S401 is iterated operation to intermediate data by genetic algorithm, obtains initial population;
Step S402, is decoded initial population, obtains each individual in initial population;
Step S403 optimizes the testing accuracy of each individual in initial population by neural network algorithm, obtains
To transitional population;
Step S404 judges the individual for being more than default testing accuracy in transitional population with the presence or absence of testing accuracy;It is if intermediate
There are the individual that testing accuracy is more than default testing accuracy in group, step S405 is performed, otherwise performs step S406;
Step S405, using the individual as training result;
Step S406, according to default testing accuracy, select transitional population, intersected and made a variation behaviour by genetic algorithm
Make, obtain new group;
Step S407 is decoded new group, and passes through inspection of the neural network algorithm to decoded each individual
It tests precision to optimize, obtains new transitional population;
Step S408 judges the individual for being more than default testing accuracy in new transitional population with the presence or absence of testing accuracy, if
There are the individual that testing accuracy is more than default testing accuracy in new transitional population, then step S409 is performed, otherwise performs step
S410;
Step S409, using the individual as training result;
Step S410 continues through genetic algorithm and new transitional population is selected, is intersected and mutation operation, obtains new
Group, and continue through neural network algorithm and the testing accuracy of each individual in new group optimized, obtain new
Transitional population, until there are the individual that testing accuracy is more than default testing accuracy in new transitional population, and by individual work
For training result.
In the third embodiment of the present invention, the process of training result is being obtained by genetic algorithm and neural network algorithm
In, it can judge whether the interative computation number of genetic algorithm reaches preset times, when the interative computation number of genetic algorithm reaches
During preset times, even if there be no testing accuracy be more than the individual of default testing accuracy also can deconditioning, and will last time
The highest individual of testing accuracy in the new transitional population that operation obtains is used as training result.
Wherein, the switching of neural network algorithm to genetic algorithm can be by the default testing accuracy or local maxima that are limited
Step number realizes, the switching of genetic algorithm to neural network algorithm can be operated by primary complete genetic operator (such as to be selected
Operator, report to the leadship after accomplishing a task operator or mutation operator) it realizes, so mutually using the training result of other side as the initial weight of oneself or initial
Group alternately and repeatedly trains, until reach the default testing accuracy limited or genetic algorithm interative computation number it is default
Number.
Fourth embodiment
The embodiment of the recommendation apparatus of a kind of telecommunication service provided below for fourth embodiment of the invention.The telecommunications industry
The embodiment of the recommendation apparatus of business belongs to same design, the embodiment of the recommendation apparatus of telecommunication service with above-mentioned embodiment of the method
In not detailed description detail content, above method embodiment can be referred to.
As shown in figure 5, the fourth embodiment of the present invention provides a kind of recommendation apparatus of telecommunication service, which includes:
Prediction module 51, for the historical information according to target user and be previously obtained for examining whether user can connect
By the training result of telecommunication service to be recommended, telecommunication service matching prediction is carried out to target user, it is pre- to obtain telecommunication service matching
The result of survey;
Identification module 52, for according to telecommunication service matching prediction as a result, telecommunication service to be recommended can be received by identifying
User;
Sending module 53, for sending telecommunication service recommendation information to the user that can receive telecommunication service to be recommended.
Wherein, identification module 52 includes:
Whether judging unit, the result for judging telecommunication service matching prediction are more than preset value, and when telecommunication service
When result with prediction is more than preset value, determination unit is triggered;
Determination unit for the triggering according to judging unit, determines the target user for that can receive telecommunication service to be recommended
User.
Wherein, device further includes:
Acquisition module, for obtaining the training result for user to be examined whether to receive telecommunication service to be recommended.
Wherein, acquisition module includes:
Acquiring unit, for obtaining the required basic data of training;
Processing unit for being pre-processed to basic data, obtains intermediate data;
Training unit is trained intermediate data for passing through genetic algorithm and neural network algorithm, obtains training knot
Fruit.
Wherein, processing unit includes:
Subelement is screened, for being screened according to preset screening requirement to basic data, the data after being screened;
Sort subelement, for according to preset data format, being ranked up to the data after screening;
Normalizing unit for the data after sequence to be normalized, obtains the number that there is genetic algorithm can identify
According to the intermediate data of structure.
Wherein, training unit includes:
Operation subelement is iterated operation to intermediate data for passing through genetic algorithm, obtains initial population;
Decoding subunit for being decoded to initial population, obtains each individual in initial population;
Optimize subelement, it is excellent to the testing accuracy progress of each individual in initial population for passing through neural network algorithm
Change, obtain transitional population;
Judgment sub-unit, for judging to be more than the individual of default testing accuracy in transitional population with the presence or absence of testing accuracy,
And when being more than the individual for presetting testing accuracy there are testing accuracy in transitional population, trigger the first determination subelement;
First determination subelement, for the triggering according to judgment sub-unit, using the individual as training result.
Wherein, device further includes:
First operation module, for when testing accuracy being not present in transitional population being more than the individual of default testing accuracy,
According to default testing accuracy, transitional population is selected by genetic algorithm, is intersected and mutation operation, obtains new group;
Decoder module for being decoded to new group, and passes through neural network algorithm to decoded each individual
Testing accuracy optimize, obtain new transitional population;
First judgment module, for judging to be more than default testing accuracy with the presence or absence of testing accuracy in new transitional population
Individual, and when being more than the individual for presetting testing accuracy there are testing accuracy in new transitional population, determining module is triggered, when new
Transitional population in there is no during the individual that testing accuracy is more than default testing accuracy, trigger the second operation module;
Determining module, for the triggering according to the first judgment module, using the individual as training result;
Second operation module for the triggering according to the first judgment module, continues through genetic algorithm to new intermediate group
Body selected, is intersected and mutation operation, obtains new group, and continue through neural network algorithm to every in new group
The testing accuracy of individual optimizes, and obtains new transitional population, to be more than until there are testing accuracies in new transitional population
The individual of default testing accuracy, and using the individual as training result.
Wherein, device further includes:
Whether the second judgment module, the interative computation number for judging genetic algorithm reach preset times, and when heredity
When the interative computation number of algorithm reaches preset times, stopping modular is triggered;
Stopping modular, for the triggering according to the second judgment module, deconditioning, and last time operation obtained new
Transitional population in testing accuracy it is highest individual be used as training result.
In the fourth embodiment of the present invention, the recommendation apparatus of telecommunication service can according to the historical information of target user and
Be previously obtained for training result that user is examined whether to receive telecommunication service to be recommended, obtain a telecommunication service matching
The result of prediction.And then judge whether the target user is that can receive to treat according to the obtained result of telecommunication service matching prediction
Recommend the user of telecommunication service, when it is that can receive the user of telecommunication service to be recommended, be sent to telecommunication service recommendation
Breath.Accurate business marketing pattern just can be realized in this way, achieve the purpose that accurately to recommend telecommunication service to user.
It should be noted that the recommendation apparatus of telecommunication service provided in an embodiment of the present invention is the dress using the above method
It puts, i.e., all embodiments of the above method are suitable for the device, and can reach the same or similar advantageous effect.
The above is the preferred embodiment of the present invention, it is noted that for those skilled in the art
For, without departing from the principles of the present invention, several improvements and modifications can also be made, these improvements and modifications
It should be regarded as protection scope of the present invention.
Claims (10)
1. a kind of recommendation method of telecommunication service, which is characterized in that the method includes:
It obtains to examine whether user can receive the training result of telecommunication service to be recommended;
According to the historical information of target user and be previously obtained for examining whether user can receive telecommunication service to be recommended
Training result carries out the target user telecommunication service matching prediction, obtains the result of telecommunication service matching prediction;
According to telecommunication service matching prediction as a result, identifying the user that can receive telecommunication service to be recommended;
Telecommunication service recommendation information is sent to the user that can receive telecommunication service to be recommended;
Wherein described acquisition specifically includes for examining whether user can receive the training result of telecommunication service to be recommended:
Obtain the required basic data of training;
The basic data is pre-processed, obtains intermediate data;
Operation is iterated to the intermediate data by genetic algorithm, obtains initial population;
The initial population is decoded, obtains each individual in the initial population;
The testing accuracy of each individual in the initial population is optimized by neural network algorithm, obtains intermediate group
Body;
Judge the individual for being more than default testing accuracy in the transitional population with the presence or absence of testing accuracy;
If it is tied in the transitional population there are the individual that testing accuracy is more than default testing accuracy using the individual as training
Fruit.
2. recommend method as described in claim 1, which is characterized in that the knot according to telecommunication service matching prediction
Fruit identifies the user that can receive telecommunication service to be recommended, specifically includes:
Judge whether the result of the telecommunication service matching prediction is more than preset value;
If the result of the telecommunication service matching prediction is more than preset value, it is determined that the target user is can receive telecommunications to be recommended
The user of business.
3. recommend method as described in claim 1, which is characterized in that it is described that the basic data is pre-processed, it obtains
Intermediate data specifically includes:
The basic data is screened according to preset screening requirement, the data after being screened;
According to preset data format, the data after the screening are ranked up;
Data after sequence are normalized, obtain the mediant for the data structure that there is the genetic algorithm can identify
According to.
4. recommend method as described in claim 1, which is characterized in that the method further includes:
If there is no the individuals that testing accuracy is more than default testing accuracy, basis in the transitional population to preset testing accuracy,
The transitional population is selected by genetic algorithm, is intersected and mutation operation, obtains new group;
The new group is decoded, and passes through neural network algorithm and the testing accuracy of decoded each individual is carried out
Optimization, obtains new transitional population;
Judge the individual for being more than default testing accuracy in new transitional population with the presence or absence of testing accuracy;
If there are the individual that testing accuracy is more than default testing accuracy in the new transitional population, using the individual as training
As a result;
If hereditary calculation is continued through there is no the individual that testing accuracy is more than default testing accuracy in the new transitional population
Method selects new transitional population, intersects and mutation operation, obtains new group, and continue through neural network algorithm pair
The testing accuracy of each individual in new group optimizes, and obtains new transitional population, until being deposited in new transitional population
It is more than the individual of default testing accuracy, and using the individual as training result in testing accuracy.
5. recommend method as claimed in claim 4, which is characterized in that in the neural network algorithm that continues through to new group
The testing accuracy of each individual in body optimizes, and after obtaining new transitional population, the method further includes:
Judge whether the interative computation number of genetic algorithm reaches preset times;
If the interative computation number of genetic algorithm reaches preset times, deconditioning, and last time operation is obtained new
Transitional population in testing accuracy it is highest individual be used as training result.
6. a kind of recommendation apparatus of telecommunication service, which is characterized in that described device includes:
Acquisition module, for obtaining the training result for user to be examined whether to receive telecommunication service to be recommended;
Prediction module, for the historical information according to target user and be previously obtained for examining whether user can receive to wait to push away
The training result of telecommunication service is recommended, telecommunication service matching prediction is carried out to the target user, obtains telecommunication service matching prediction
Result;
Identification module, for according to the telecommunication service matching prediction as a result, telecommunication service to be recommended can be received by identifying
User;
Sending module, for sending telecommunication service recommendation information to the user that can receive telecommunication service to be recommended;
Wherein described acquisition module includes:
Acquiring unit, for obtaining the required basic data of training;
Processing unit for being pre-processed to the basic data, obtains intermediate data;
Training unit is trained the intermediate data for passing through genetic algorithm and neural network algorithm, obtains the instruction
Practice result;
Wherein described training unit includes:
Operation subelement is iterated operation to the intermediate data for passing through genetic algorithm, obtains initial population;
Decoding subunit for being decoded to the initial population, obtains each individual in the initial population;
Optimize subelement, it is excellent to the testing accuracy progress of each individual in the initial population for passing through neural network algorithm
Change, obtain transitional population;
Judgment sub-unit, for judging to be more than the individual of default testing accuracy in the transitional population with the presence or absence of testing accuracy,
And when being more than the individual for presetting testing accuracy there are testing accuracy in the transitional population, trigger the first determination subelement;
First determination subelement, for the triggering according to the judgment sub-unit, using the individual as training result.
7. recommendation apparatus as claimed in claim 6, which is characterized in that the identification module includes:
Whether judging unit, the result for judging the telecommunication service matching prediction are more than preset value, and when the telecommunications industry
When the result of business matching prediction is more than preset value, determination unit is triggered;
Determination unit for the triggering according to the judging unit, determines the target user for that can receive telecommunication service to be recommended
User.
8. recommendation apparatus as claimed in claim 6, which is characterized in that the processing unit includes:
Subelement is screened, for being screened according to preset screening requirement to the basic data, the data after being screened;
Sort subelement, for according to preset data format, being ranked up to the data after the screening;
Normalizing unit for the data after sequence to be normalized, obtains the number that there is the genetic algorithm can identify
According to the intermediate data of structure.
9. recommendation apparatus as claimed in claim 6, which is characterized in that described device further includes:
First operation module, for when testing accuracy being not present in the transitional population being more than the individual of default testing accuracy,
According to default testing accuracy, the transitional population is selected by genetic algorithm, is intersected and mutation operation, obtains new group
Body;
Decoder module for being decoded to the new group, and passes through neural network algorithm to decoded each individual
Testing accuracy optimize, obtain new transitional population;
First judgment module, for judging to be more than of default testing accuracy in new transitional population with the presence or absence of testing accuracy
Body, and when being more than the individual for presetting testing accuracy there are testing accuracy in the new transitional population, determining module is triggered, when
There is no during the individual that testing accuracy is more than default testing accuracy in the new transitional population, the second operation module is triggered;
Determining module, for the triggering according to first judgment module, using the individual as training result;
Second operation module for the triggering according to first judgment module, continues through genetic algorithm to new intermediate group
Body selected, is intersected and mutation operation, obtains new group, and continue through neural network algorithm to every in new group
The testing accuracy of individual optimizes, and obtains new transitional population, to be more than until there are testing accuracies in new transitional population
The individual of default testing accuracy, and using the individual as training result.
10. recommendation apparatus as claimed in claim 9, which is characterized in that described device further includes:
Whether the second judgment module, the interative computation number for judging genetic algorithm reach preset times, and work as genetic algorithm
Interative computation number when reaching preset times, trigger stopping modular;
Stopping modular, for the triggering according to second judgment module, deconditioning, and last time operation obtained new
Transitional population in testing accuracy it is highest individual be used as training result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510329668.5A CN104866626B (en) | 2015-06-15 | 2015-06-15 | A kind of recommendation method and device of telecommunication service |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510329668.5A CN104866626B (en) | 2015-06-15 | 2015-06-15 | A kind of recommendation method and device of telecommunication service |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104866626A CN104866626A (en) | 2015-08-26 |
CN104866626B true CN104866626B (en) | 2018-06-26 |
Family
ID=53912452
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510329668.5A Active CN104866626B (en) | 2015-06-15 | 2015-06-15 | A kind of recommendation method and device of telecommunication service |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104866626B (en) |
Families Citing this family (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106815221B (en) * | 2015-11-27 | 2020-02-14 | 华为软件技术有限公司 | Telecommunication service relation judging method and device |
CN105894028B (en) * | 2016-03-31 | 2020-01-10 | 百度在线网络技术(北京)有限公司 | User identification method and device |
CN106600369A (en) * | 2016-12-09 | 2017-04-26 | 广东奡风科技股份有限公司 | Real-time recommendation system and method of financial products of banks based on Naive Bayesian classification |
CN107766556B (en) * | 2017-11-03 | 2021-07-30 | 福建工程学院 | Interactive ontology matching method based on evolutionary algorithm and computer equipment |
CN109600757B (en) * | 2018-11-29 | 2022-01-18 | 南京亚信软件有限公司 | Prediction method and device for base station capacity expansion, computer equipment and storage medium |
CN109919675A (en) * | 2019-03-04 | 2019-06-21 | 深圳微品致远信息科技有限公司 | Communication user upshift prediction probability recognition methods neural network based and system |
CN110378739B (en) * | 2019-07-23 | 2022-03-29 | 中国联合网络通信集团有限公司 | Data traffic matching method and device |
TWI726398B (en) * | 2019-08-27 | 2021-05-01 | 中華電信股份有限公司 | Self-adapted telecommunication service recommend system and method |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7013238B1 (en) * | 2003-02-24 | 2006-03-14 | Microsoft Corporation | System for delivering recommendations |
US8380548B2 (en) * | 2007-02-15 | 2013-02-19 | So-ling Carmen Ng | Method for managing intellectual property |
CN103107936A (en) * | 2011-11-11 | 2013-05-15 | 中国移动通信集团上海有限公司 | Method and device for sending information |
-
2015
- 2015-06-15 CN CN201510329668.5A patent/CN104866626B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7013238B1 (en) * | 2003-02-24 | 2006-03-14 | Microsoft Corporation | System for delivering recommendations |
US8380548B2 (en) * | 2007-02-15 | 2013-02-19 | So-ling Carmen Ng | Method for managing intellectual property |
CN103107936A (en) * | 2011-11-11 | 2013-05-15 | 中国移动通信集团上海有限公司 | Method and device for sending information |
Non-Patent Citations (4)
Title |
---|
Color recommendation system combining design concepts with interactive customers preference modeling from context changes;Ladys Rodriguez.etc;《Evolutionary Computation (CEC)》;20100927;第1-8页 * |
基于数值计算方法的BP神经网络及遗传算法的优化研究;吴仕勇;《中国优秀硕士学位论文全文数据库》;20061215;I140-67 * |
基于消费行为的电信套餐推荐模型研究;兰慧;《中国优秀硕士学位论文全文数据库》;20140115;第23-24页和第33-44页 * |
遗传算法BP神经网络的预报研究和应用;吴建生等;《数学的实践与认识》;20050131;第35卷(第1期);第83-88页 * |
Also Published As
Publication number | Publication date |
---|---|
CN104866626A (en) | 2015-08-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104866626B (en) | A kind of recommendation method and device of telecommunication service | |
CN109803295B (en) | Method and device for evaluating communication cell rectification priority | |
CN108052639A (en) | Industry user based on carrier data recommends method and device | |
CN104750760B (en) | A kind of implementation method and device for recommending application software | |
CN105975479B (en) | A kind of telecommunication user interest-degree analysis method and system based on tag library | |
CN105007171A (en) | User data analysis system and method based on big data in communication field | |
CN102591917A (en) | Data processing method and system and related device | |
CN108319585A (en) | Data processing method and device, electronic equipment, computer-readable medium | |
CN107527240A (en) | A kind of operator's industry product Praise effect identification system and method | |
CN110267288A (en) | Mobile network complains localization method and device | |
CN110781256B (en) | Method and device for determining POI matched with Wi-Fi based on sending position data | |
CN108230040B (en) | Store arrival prediction method and device | |
US8250002B2 (en) | Method and system for positioning | |
CN104468764B (en) | A kind of tactful dispatching method, apparatus and system | |
CN105740434A (en) | Network information scoring method and device | |
CN114022196A (en) | Advertisement putting method, device, electronic device and storage medium | |
CN107291860B (en) | Seed user determination method | |
CN109474755A (en) | Abnormal phone active predicting method and system based on sequence study and integrated study | |
CN112153636A (en) | Method for predicting number portability and roll-out of telecommunication industry user based on machine learning | |
CN110690982B (en) | Method and system for correlation analysis of management performance data of telecommunication network | |
CN112307075B (en) | User relationship identification method and device | |
CN106682030A (en) | Method and device for information processing | |
CN114970495A (en) | Name disambiguation method and device, electronic equipment and storage medium | |
CN111405464B (en) | Base station position detection method and device | |
CN107231334A (en) | A kind of short message monitoring method and device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
EXSB | Decision made by sipo to initiate substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |