CN107016460A - User changes planes Forecasting Methodology and device - Google Patents
User changes planes Forecasting Methodology and device Download PDFInfo
- Publication number
- CN107016460A CN107016460A CN201710190104.7A CN201710190104A CN107016460A CN 107016460 A CN107016460 A CN 107016460A CN 201710190104 A CN201710190104 A CN 201710190104A CN 107016460 A CN107016460 A CN 107016460A
- Authority
- CN
- China
- Prior art keywords
- data
- user
- decision
- renewal
- tree model
- 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
- 238000000034 method Methods 0.000 title claims abstract description 70
- 230000000052 comparative effect Effects 0.000 claims abstract description 7
- 238000003066 decision tree Methods 0.000 claims description 113
- 238000012549 training Methods 0.000 claims description 76
- 238000012360 testing method Methods 0.000 claims description 74
- 238000012216 screening Methods 0.000 claims description 44
- 241001269238 Data Species 0.000 claims description 19
- 235000013399 edible fruits Nutrition 0.000 claims description 4
- 230000003542 behavioural effect Effects 0.000 claims description 3
- 238000012545 processing Methods 0.000 description 13
- 238000010586 diagram Methods 0.000 description 8
- 239000000284 extract Substances 0.000 description 4
- 239000000463 material Substances 0.000 description 3
- PEDCQBHIVMGVHV-UHFFFAOYSA-N Glycerine Chemical compound OCC(O)CO PEDCQBHIVMGVHV-UHFFFAOYSA-N 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000011664 signaling Effects 0.000 description 2
- 241000196324 Embryophyta Species 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Marketing (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Changed planes Forecasting Methodology the invention discloses a kind of user, methods described includes：Subscriber's account data and default bill according to getting Forecasting Methodology of changing planes determine first renewal user's result data；Search behavior data and preset search behavior prediction method according to getting determine second renewal user's result data and target terminal model data；The first renewal user result data and the second renewal user result data are compared, and the data that predict the outcome are determined according to comparative result；Data are predicted the outcome according to described and the target terminal model data determines that user changes planes prediction inventory.Changed planes prediction meanss the invention also discloses a kind of user, can more accurately predict the tendency of user's changes terminal.
Description
Technical field
The present invention relates to communication technical field, more particularly to a kind of user changes planes Forecasting Methodology and device.
Background technology
With the popularization of smart mobile phone, how long user changes a new terminal, and can be selected when changing what kind of
Terminal, needs the Important Problems solved when carrying out terminal marketing work as telecom operators at present, manufacturer terminal.It is existing to use
Family Forecasting Methodology of changing planes is mainly operator according to telecommunication bills, according to user's rate ability, the passing terminal usage cycles of user,
Terminal bundling is expired the big datas such as situation, by setting up user terminal forecast model, and prediction user terminal is changed planes tendency, but
It is due to telecommunication bills data limitation, it is impossible to directly reflect user's changes terminal tendency so that it is accurate that prediction user changes planes
Rate is reduced.
Therefore, existing Forecasting Methodology of changing planes, which is existed, can not directly reflect that user's changes terminal is inclined to, and prediction user changes
The problem of accuracy rate of machine is reduced.
The content of the invention
It is a primary object of the present invention to propose that a kind of user changes planes Forecasting Methodology and device, it is intended to solve existing change planes
It can not directly reflect that user's changes terminal is inclined to present in Forecasting Methodology, the problem of accuracy rate that prediction user changes planes is low.
To achieve the above object, a kind of user for providing of the present invention changes planes Forecasting Methodology, and methods described includes：
Subscriber's account data and default bill according to getting Forecasting Methodology of changing planes determine first renewal user's number of results
According to；
Search behavior data and preset search behavior prediction method according to getting determine second renewal user's number of results
According to and target terminal model data；
The first renewal user result data and the second renewal user result data are compared, and according to than
Relatively result determines the data that predict the outcome；
Data are predicted the outcome according to described and the target terminal model data determines that user changes planes prediction inventory.
In addition, to achieve the above object, the present invention also provides a kind of user and changed planes prediction meanss, and described device includes：
First determining module, for determining the according to the subscriber's account data that get and default bill Forecasting Methodology of changing planes
One renewal user's result data；
Second determining module, for determining the according to the search behavior data that get and preset search behavior prediction method
Two renewal user's result datas and target terminal model data；
3rd determining module, for the first renewal user result data and the second renewal user result data
It is compared, and the data that predict the outcome is determined according to comparative result；
4th determining module, for predicting the outcome data according to and the target terminal model data determines that user changes
Machine predicts inventory.
User proposed by the present invention changes planes Forecasting Methodology, is changed planes according to the subscriber's account data and default bill that get pre-
Survey method determines first renewal user's result data, and according to the search behavior data and preset search behavior prediction side got
Method determines second renewal user's result data and target terminal model data, to the first renewal user result data and described
Second renewal user's result data is compared, and determines the data that predict the outcome according to comparative result, is predicted the outcome according to described
Data and the target terminal model data determine that user changes planes prediction inventory.Compared with prior art, the present invention will be according to account
First renewal user's result data that forms data is got and the second renewal user's result arrived according to search behavior data acquisition
Data are compared, so that it is determined that the data that predict the outcome, and by the data that predict the outcome with being determined according to search behavior data
The matching association of target terminal model data, the prediction inventory so that it is determined that user changes planes, so as to more accurately predict user
The tendency of changes terminal.
Brief description of the drawings
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing
There is the accompanying drawing used required in technology description to be briefly described, it should be apparent that, drawings in the following description are only this
Some embodiments of invention, for those skilled in the art, on the premise of not paying creative work, can also basis
These accompanying drawings obtain other accompanying drawings.
Fig. 1 changes planes the schematic flow sheet of Forecasting Methodology for a kind of user that first embodiment of the invention is provided；
Fig. 2 be Fig. 1 in step S101 refinement step schematic flow sheet；
Fig. 3 be Fig. 1 in step S102 refinement step schematic flow sheet；
Fig. 4 be Fig. 1 in step S103 refinement step schematic flow sheet；
Fig. 5 for a kind of user that third embodiment of the invention is provided change planes prediction meanss functional module schematic diagram；
Fig. 6 be third embodiment of the invention in the first determining module 601 refinement functional module schematic diagram；
Fig. 7 be third embodiment of the invention in the second determining module 602 refinement functional module schematic diagram；
Fig. 8 be third embodiment of the invention in the 3rd determining module 603 refinement functional module schematic diagram.
Embodiment
To enable goal of the invention, feature, the advantage of the present invention more obvious and understandable, below in conjunction with the present invention
Accompanying drawing in embodiment, the technical scheme in the embodiment of the present invention is clearly and completely described, it is clear that described reality
It is only a part of embodiment of the invention to apply example, and not all embodiments.Based on the embodiment in the present invention, people in the art
The every other embodiment that member is obtained under the premise of creative work is not made, belongs to the scope of protection of the invention.
Describe to realize the mobile terminal of each embodiment of the invention referring now to accompanying drawing.In follow-up description, use
For represent element such as " module ", " part " or " unit " suffix only for be conducive to the present invention explanation, itself
Not specific meaning.Therefore, " module " can be used mixedly with " part ".
Changeover device in the embodiment of the present invention can be implemented in a variety of manners.For example, changing planes described in the present invention
Device can include such as mobile phone, smart phone, notebook computer, PAD (tablet personal computer) etc. mobile terminal and
Such as fixed terminal of numeral TV, desktop computer etc..
The flow signal of Forecasting Methodology referring to Fig. 1, Fig. 1 changes planes for a kind of user that first embodiment of the invention is provided
Figure, including：
Step S101, Forecasting Methodology of being changed planes according to the subscriber's account data that get and default bill determine that first changes planes use
Family result data；
In embodiments of the present invention, billing data includes user profile, business essential information, terminal configuration information, contract
Information, terminal change the data such as information, service condition.
Wherein, default bill Forecasting Methodology of changing planes is set for user is self-defined in advance, and user can select according to actual conditions
Select different methods.
Step S102, determine that according to the search behavior data that get and preset search behavior prediction method second changes planes use
Family result data and target terminal model data；
In embodiments of the present invention, search behavior data include user's search total degree, the terminal models of search, search always
The data such as number of days, the sphere of consumption of search.
Wherein, preset search behavior prediction method is user's self-defined setting in advance, and user can select according to actual conditions
Different methods are selected, the preset search behavior prediction method and default bill Forecasting Methodology of changing planes can be that same process also may be used
To be the method differed.
Step S103, first renewal user's result data and second renewal user's result data are compared, and according to
Comparative result determines the data that predict the outcome；
In embodiments of the present invention, the data that predict the outcome represent the user profile that may be changed planes predicted, and this is pre-
Surveying result data includes the desired user number changed planes of prediction, the information such as the geographical position of branch.
Step S104, according to predicting the outcome data and target terminal model data determines that user changes planes prediction inventory.
In embodiments of the present invention, target terminal model data is the terminal models of N before user's searching times ranking
Data, N is positive integer.
It is preferred that, the data of the terminal models of N before predict the outcome data and user's searching times ranking are associated
Matching, determines that user changes planes prediction inventory.
In embodiments of the present invention, is determined according to the subscriber's account data and default bill that get Forecasting Methodology of changing planes
One renewal user's result data, and determine that second changes according to the search behavior data and preset search behavior prediction method got
Machine user result data and target terminal model data, by first renewal user's result data and second renewal user's result data
It is compared, so that it is determined that the data that predict the outcome, the data that predict the outcome are associated with the matching of target terminal model data, so that
Determine that user changes planes prediction inventory, it is possible to more accurately predict the tendency of user's changes terminal.
Referring to Fig. 2, shown in Fig. 1 the refinement step of first embodiment step 101 schematic flow sheet, including：
Step S201, the data target based on subscriber's account data definition prediction renewal user；
In embodiments of the present invention, the data target of prediction renewal user can be defined with actual conditions.
It is preferred that, data target is as shown in table 1：
Table 1
Step S202, using data target and default screening rule subscriber's account data are screened, and will be after screening
Subscriber's account data be divided into training data and test data；
In embodiments of the present invention, subscriber's account data are screened using data target, rejected from billing data
Data target lacks serious data, data and fills in wrong data and the data repeated, obtains the first subscriber's account data, presses
Postsearch screening is carried out to the first subscriber's account data according to default screening rule, wherein, according to default screening rule to first user's account
Forms data carries out postsearch screening and is divided into two steps, and the first step is：Weed out in the first subscriber's account data and searched in certain period
Rope number of times is less than the secondary data of N (N is positive integer), obtains second user billing data, certain period here can be one
The moon, two months etc.；Second step is：Ratio according to default renewal user and not renewal user is entered to second user billing data
Row data pick-up.For example, second user billing data is 1000 users, wherein renewal user is 500 people, and renewal user is not
The ratio of 500 people, default renewal user and not renewal user are 1:5, then 100 renewal users are extracted from 1000 users
With 500 not renewal users, the subscriber's account data one after obtained screening have 600 users, including 100 renewal users
With 500 not renewal users, if the ratio of default renewal user and not renewal user are 1:3, then taken out from 1000 users
The not renewal user of 100 renewal users and 300 is taken, the subscriber's account data one after obtained screening have 400 users, wrap
Include as the not renewal user of 100 renewal users and 300.
Afterwards, training data and test data are divided into the subscriber's account data after screening according to default division proportion.
For example, training data and test data ratio are 1:2, according to 1:2 ratio is drawn to 600 users in previous example
Point, obtained training data be 200 people, test data be 400 people, but in this 200 training datas renewal user with not
The ratio of renewal user is stilled need as 1:5, i.e., this 200 training datas include 33 renewal users (200 be multiplied by six/
One) and 167 not renewal user's (200 are multiplied by 5/6ths), renewal user and not renewal user in this 400 test datas
Ratio still need as 1:5, i.e., include 67 renewal users (400 are multiplied by 1/6th) and 333 in this 400 test datas
Individual not renewal user (400 are multiplied by 5/6ths).
It is preferred that, the ratio of test data and training data is 3:7.
Step S203, using training data default decision-tree model is trained, the first decision-making after being trained
Tree-model；
In embodiments of the present invention, decision tree (Decision Tree) is on the basis of known various situation probability of happening
On, the desired value that net present value (NPV) is asked for by constituting decision tree is more than or equal to zero probability, and assessment item risk judges that its is feasible
Property method of decision analysis, be a kind of diagram method with probability analysis directly perceived.
It is preferred that, the decision-tree model that the present invention is selected is decision tree C5.0 models.
Wherein, when being trained to default decision-tree model, training data can be analyzed, and according to important feelings
Condition automatically excludes the relatively low data target of importance, generates a tables of data, and the tables of data is divided into two kinds：Exclude
Data target and the data target not excluded.
Step S204, based on the first decision-tree model test data is carried out testing determination and predicting the outcome；
In embodiments of the present invention, test data is inputted to the first decision-tree model, can be predicted the outcome.
Whether step S205, the performance for assessing the first decision-tree model using predicting the outcome are satisfied with preparatory condition；In this hair
In bright embodiment, preparatory condition is that user pre-sets, and can be configured and adjust according to actual conditions.For example, default bar
Part is 80 points, then predicts the outcome and represent that the performance of the first decision-tree model is satisfied with preparatory condition if more than or equal to 80, if in advance
It is 70 points to survey result, then needs that the first decision-tree model is carried out to improve processing operation.Here perfect processing operation is divided into two
Walk, the first step is：After step S203 is completed, a tables of data can be obtained, the tables of data is observed, by the data target excluded
Corresponding data, which are rejoined in training data, to be trained, the first decision-tree model after improving, afterwards to complete
The first decision-tree model dealt with problems arising from an accident is tested, and is predicted the outcome, and is checked that this predicts the outcome and whether is more than 80 points or equal to 80
Point, if more than 80 points or equal to 80 points, being met the first decision-tree model of preparatory condition, terminate to improve processing behaviour
Make, if less than 80 points, the second step of the perfect processing operation of progress, adjustment renewal user and the not ratio of renewal user, for example
By renewal user and not, the ratio of renewal user is 1 before:5, now by the renewal user and not renewal user according to actual conditions
Ratio be adjusted to 1：8, the not renewal user of 100 renewal users and 800 is extracted from subscriber's account data, is drawn according to default
This 900 subscriber's account data are divided into training data and test data by point ratio.And be trained and test again, always
Preparatory condition is met to the first obtained decision-tree model.
Wherein, table 2 is shown in default bill changes planes Forecasting Methodology, the key ranking of data target, in the table
Indispensable index when data target is model training.
Table 2
If step S206, meeting, by the decision-tree model of data input first to be predicted, and the first decision-tree model is defeated
The result data gone out is defined as first renewal user's result data.
In embodiments of the present invention, after preparatory condition is met, by the decision-tree model of data input first to be predicted, and will
The result data of first decision-tree model output is defined as first renewal user's result data.For example, the data of training and test
For the data of 2 months 2017, in order to predict first renewal user's result data in April, then the data in March (are treated pre-
Survey data) the first decision-tree model of input, and the result data of output is defined as first renewal user's result data.
In embodiments of the present invention, compared with prior art, the embodiment of the present invention is predicted based on subscriber's account data definition
The data target of renewal user, the data target is trained to model to play an important role, and is determined using training data to default
Plan tree-model is trained, and after the first decision-tree model after being trained, the model is tested, if predicting the outcome satisfaction
The first decision-tree model that data input to be predicted is then met preparatory condition by preparatory condition obtains first renewal user's number of results
According to being unsatisfactory for preparatory condition if predicting the outcome, training data and test data be adjusted, are trained again after adjustment
And test, preparatory condition is met until predicting the outcome, so as to obtain accurate first renewal user result data.
Referring to Fig. 3, shown in Fig. 1 the refinement step of first embodiment step 102 schematic flow sheet, including：
Step S301, the data target based on search behavior data definition user's search behavior；
In embodiments of the present invention, signaling call bill data is screened first, extracts user's search behavior data, enter
One step, keyword match search is carried out, the data for there are the keywords such as mobile phone, terminal in search behavior data are extracted.
Wherein, the data target of user's search behavior can be defined according to actual conditions.
It is preferred that, data target is as shown in table 3：
Table 3
Step S302, using data target and default screening rule search behavior data are screened, and will be after screening
Search behavior data be divided into training data and test data；
In embodiments of the present invention, search behavior data are screened using data target, from search behavior data
Reject the serious data of data target missing, data and fill in wrong data and the data repeated, obtain the first search behavior number
According to, postsearch screening is carried out to the first search behavior data according to default screening rule, wherein, searched according to default screening rule to first
Rope behavioral data carries out postsearch screening and is divided into two steps, and the first step is：Weed out certain period in the first search behavior data
Interior searching times are less than the secondary data of N (N is positive integer), obtain the second search behavior data, and certain period here can be
One month, two months etc.；Second step is：According to the ratio of default renewal user and not renewal user to the second search behavior number
According to progress data pick-up.For example, the second search behavior data are 1000 users, wherein renewal user is 500 people, use of not changing planes
Family is 500 people, and the ratio of default renewal user and not renewal user are 1:5, then extract 100 from 1000 users and change planes
The not renewal user of user and 500, the search behavior data one after obtained screening have 600 users, including 100 are changed planes
The not renewal user of user and 500, if the ratio of default renewal user and not renewal user are 1:3, then from 1000 users
100 renewal users of middle extraction and 300 not renewal users, the search behavior data one after obtained screening have 400 use
Family, including be the not renewal user of 100 renewal users and 300.
Wherein, the first step weeds out in the first search behavior data searching times in certain period and is less than N (N is positive integer)
Secondary data according to actual conditions selection operation or can not operated.
Afterwards, training data and test data are divided into the search behavior data after screening according to default division proportion.
For example, training data and test data ratio are 1:2, according to 1:2 ratio is drawn to 600 users in previous example
Point, obtained training data be 200 people, test data be 400 people, but in this 200 training datas renewal user with not
The ratio of renewal user is stilled need as 1:5, i.e., this 200 training datas include 33 renewal users (200 be multiplied by six/
One) and 167 not renewal user's (200 are multiplied by 5/6ths), renewal user and not renewal user in this 400 test datas
Ratio still need as 1:5, i.e., include 67 renewal users (400 are multiplied by 1/6th) and 333 in this 400 test datas
Individual not renewal user (400 are multiplied by 5/6ths).
It is preferred that, the ratio of test data and training data is 3:7.
Step S303, using training data default decision-tree model is trained, the second decision-making after being trained
Tree-model；
In embodiments of the present invention, when being trained to default decision-tree model, training data can be analyzed,
And automatically excluded the relatively low data target of importance according to material circumstance, a tables of data is generated, the tables of data is divided into
Two kinds：The data target excluded and the data target not excluded.
Step S304, based on the second decision-tree model test data is carried out testing determination and predicting the outcome；
In embodiments of the present invention, test data is inputted to the second decision-tree model, one can be obtained and predicted the outcome.
Whether step S305, the performance for assessing the second decision-tree model using predicting the outcome are satisfied with preparatory condition；
In embodiments of the present invention, preparatory condition is that user pre-sets, and can be configured and adjust according to actual conditions
It is whole.For example, preparatory condition is 80 points, then predicts the outcome and the performance satisfaction of the second decision-tree model is represented if more than or equal to 80
Preparatory condition, if predicting the outcome as 70 points, needs that the second decision-tree model is carried out to improve processing operation, perfect place here
Reason operation is divided into two steps, and the first step is：After step S303 is completed, a tables of data can be obtained, the tables of data is observed, will be excluded
The data corresponding to data target fallen, which are rejoined in training data, to be trained, the second decision tree mould after improving
Type, tests the second decision-tree model after improving, is predicted the outcome afterwards, checks that this predicts the outcome and whether is more than 80
Divide or equal to 80 points, if more than 80 points or equal to 80 points, being met the second decision-tree model of preparatory condition, terminate
Processing operation is improved, if less than 80 points, improve the second step of processing operation, adjustment renewal user and not renewal user's
Ratio, for example before by renewal user and not the ratio of renewal user be 1:5, now according to actual conditions by renewal user and not
The ratio of renewal user is adjusted to 1：8, the not renewal user of 100 renewal users and 800 is extracted from search behavior data, is pressed
This 900 search behavior data are divided into training data and test data according to default division proportion, and be trained again and
Test, until the second obtained decision-tree model meets preparatory condition.
Wherein, table 4 is shown in preset search behavior prediction method, the key ranking of data target, in the table
Indispensable index when data target is model training.
Table 4
If step S306, meeting, by the decision-tree model of data input second to be predicted, and the second decision-tree model is defeated
The result data gone out is defined as second renewal user's result data.
In embodiments of the present invention, after preparatory condition is met, by the decision-tree model of data input second to be predicted, and will
The result data of second decision-tree model output is defined as second renewal user's result data.For example, the data of training and test
For the data of 2 months 2017, in order to predict second renewal user's result data in April, then the data in March (are treated pre-
Survey data) the second decision-tree model of input, and the result data of output is defined as second renewal user's result data.
Compared with prior art, data of the embodiment of the present invention based on search behavior data definition user's search behavior refer to
Mark, the data target is trained to model to be played an important role, and default decision-tree model is trained using training data,
After the second decision-tree model after being trained, the model is tested, will treat pre- if meeting preparatory condition if predicting the outcome
The second decision-tree model that survey data input meets preparatory condition obtains second renewal user's result data, if it is discontented to predict the outcome
Sufficient preparatory condition, then be adjusted to training data and test data, is trained and tests again after adjustment, until prediction knot
Fruit meets preparatory condition, so as to obtain accurate second renewal user result data.
Referring to Fig. 4, shown in Fig. 1 the refinement step of first embodiment step 103 schematic flow sheet, including：
Step S401, compare first renewal user's result data and whether second renewal user's result data is identical；
If step S402, identical, first renewal user's result data or second renewal user's result data are defined as
Predict the outcome data；
If step S403, differing, by first confidence level relevant with subscriber's account data and with search behavior data
The second relevant confidence level is compared；
In embodiments of the present invention, confidence level is also referred to as reliability, or confidence level, confidence coefficient, i.e., in sampling to total
When body parameter makes an estimate, due to the randomness of sample, its conclusion is always uncertain.Therefore, using a kind of statement of probability
Interval estimation method in method, that is, mathematical statistics, i.e. estimate and population parameter within the error range necessarily allowed,
Its corresponding probability has much, and this corresponding probability is referred to as confidence level.
In embodiments of the present invention, the first decision tree of preparatory condition is met in Forecasting Methodology of being changed planes with default bill
During model, the first decision-tree model can provide the first confidence level, default being met with preset search behavior prediction method
During the first decision-tree model of condition, the first decision-tree model can provide the second confidence level, will be relevant with subscriber's account data
The first confidence level second confidence level relevant with search behavior data be compared.
If step S404, the first confidence level are more than the second confidence level, it is determined that first renewal user's result data is prediction
Result data, if the second confidence level is more than the first confidence level, it is determined that second renewal user's result data is the data that predict the outcome.
In embodiments of the present invention, the order of accuarcy of the higher prediction of confidence level is bigger.For example, confidence level is 0.95 expression：
It is 95% to estimate correct probability, and the probability for mistake occur is 5%, and confidence level represents for 0.85：Estimate that correct probability is
85%, the probability for mistake occur is 15%.
In embodiments of the present invention, differed in first renewal user's result data and second renewal user's result data
When, compare the big of first confidence level relevant with subscriber's account data, second confidence level relevant with search behavior data
It is small, the data that more correctly predict the outcome can relatively be obtained according to the size of the confidence level, it is more accurate so as to obtain
User change planes prediction inventory, so as to more accurately predict the tendency of user's changes terminal.
Referring to Fig. 5, Fig. 5 shows for a kind of the change planes functional module of prediction meanss of user that second embodiment of the invention is provided
It is intended to, including：
First determining module 501, it is true for Forecasting Methodology of being changed planes according to the subscriber's account data and default bill that get
Fixed first renewal user's result data；
In embodiments of the present invention, billing data includes user profile, business essential information, terminal configuration information, contract
Information, terminal change the data such as information, service condition.
Wherein, default bill Forecasting Methodology of changing planes is set for user is self-defined in advance, and user can select according to actual conditions
Select different methods.
Second determining module 502, for true according to the search behavior data and preset search behavior prediction method that get
Fixed second renewal user's result data and target terminal model data；
In embodiments of the present invention, search behavior data include user's search total degree, the terminal models of search, search always
The data such as number of days, the sphere of consumption of search.
Wherein, preset search behavior prediction method is user's self-defined setting in advance, and user can select according to actual conditions
Different methods are selected, the preset search behavior prediction method and default bill Forecasting Methodology of changing planes can be that same process also may be used
To be the method differed.
3rd determining module 503, for first renewal user's result data and the progress of second renewal user's result data
Compare, and the data that predict the outcome are determined according to comparative result；
In embodiments of the present invention, the data that predict the outcome represent the user profile that may be changed planes predicted, and this is pre-
Surveying result data includes the desired user number changed planes of prediction, the information such as the geographical position of branch.
4th determining module 504, for according to predict the outcome data and target terminal model data to determine that user changes planes pre-
Survey inventory.
In embodiments of the present invention, target terminal model data is the terminal models of N before user's searching times ranking
Data, N is positive integer.
It is preferred that, the data of the terminal models of N before predict the outcome data and user's searching times ranking are associated
Matching, determines that user changes planes prediction inventory.
In embodiments of the present invention, the first determining module 501 is changed according to the subscriber's account data and default bill that get
Machine Forecasting Methodology determines first renewal user's result data, the second determining module 502 and according to the search behavior data got
And preset search behavior prediction method determines second renewal user's result data and target terminal model data, the 3rd determining module
503 are compared first renewal user's result data with second renewal user's result data, so that it is determined that the data that predict the outcome,
4th determining module 504 associates data and target terminal the model data matching that predicts the outcome, so that it is determined that user change planes it is pre-
Survey inventory, it is possible to more accurately predict the tendency of user's changes terminal.
Referring to Fig. 6, in the second embodiment shown in Fig. 5 the first determining module 501 refinement high-level schematic functional block diagram, bag
Include：
First definition unit 601, the data target for predicting renewal user based on subscriber's account data definition；
In embodiments of the present invention, the data target of prediction renewal user can be defined with actual conditions.
It is preferred that, data target is as shown in the table 1 in first embodiment.
First division unit 602, for being screened using data target and default screening rule to subscriber's account data,
And the subscriber's account data after screening are divided into training data and test data；
In embodiments of the present invention, subscriber's account data are screened using data target, rejected from billing data
Data target lacks serious data, data and fills in wrong data and the data repeated, obtains the first subscriber's account data, presses
Postsearch screening is carried out to the first subscriber's account data according to default screening rule, wherein, according to default screening rule to first user's account
Forms data carries out postsearch screening and is divided into two steps, and the first step is：Weed out in the first subscriber's account data and searched in certain period
Rope number of times is less than the secondary data of N (N is positive integer), obtains second user billing data, certain period here can be one
The moon, two months etc.；Second step is：Ratio according to default renewal user and not renewal user is entered to second user billing data
Row data pick-up.For example, second user billing data is 1000 users, wherein renewal user is 500 people, and renewal user is not
The ratio of 500 people, default renewal user and not renewal user are 1:5, then 100 renewal users are extracted from 1000 users
With 500 not renewal users, the subscriber's account data one after obtained screening have 600 users, including 100 renewal users
With 500 not renewal users, if the ratio of default renewal user and not renewal user are 1:3, then taken out from 1000 users
The not renewal user of 100 renewal users and 300 is taken, the subscriber's account data one after obtained screening have 400 users, wrap
Include as the not renewal user of 100 renewal users and 300.
Afterwards, training data and test data are divided into the subscriber's account data after screening according to default division proportion.
For example, training data and test data ratio are 1:2, according to 1:2 ratio is drawn to 600 users in previous example
Point, obtained training data be 200 people, test data be 400 people, but in this 200 training datas renewal user with not
The ratio of renewal user is stilled need as 1:5, i.e., this 200 training datas include 33 renewal users (200 be multiplied by six/
One) and 167 not renewal user's (200 are multiplied by 5/6ths), renewal user and not renewal user in this 400 test datas
Ratio still need as 1:5, i.e., include 67 renewal users (400 are multiplied by 1/6th) and 333 in this 400 test datas
Individual not renewal user (400 are multiplied by 5/6ths).
It is preferred that, the ratio of test data and training data is 3:7.
First training unit 603, for being trained using training data to default decision-tree model, is obtained after training
The first decision-tree model；
In embodiments of the present invention, when being trained to default decision-tree model, training data can be analyzed,
And automatically excluded the relatively low data target of importance according to material circumstance, and generate a tables of data, the tables of data point
For two kinds：The data target excluded and the data target not excluded.
First test cell 604, predicts the outcome for testing determination to test data progress based on the first decision-tree model；
First assessment unit 605, it is default whether the performance for assessing the first decision-tree model using predicting the outcome is satisfied with
Condition；
In embodiments of the present invention, preparatory condition is that user pre-sets, and can be configured and adjust according to actual conditions
It is whole.For example, preparatory condition is 80 points, then predicts the outcome and the performance satisfaction of the first decision-tree model is represented if more than or equal to 80
Preparatory condition, if predicting the outcome as 70 points, needs that the first decision-tree model is carried out to improve processing operation.Here perfect place
Reason operation is divided into two steps, and the first step is：After being trained to default decision-tree model, a tables of data can be obtained, is observed
The tables of data, the data corresponding to the data target excluded are rejoined in training data and are trained, are improved
The first decision-tree model afterwards, tests the first decision-tree model after improving, is predicted the outcome, check that this is pre- afterwards
Survey whether result is more than 80 points or equal to 80 points, if more than 80 points or equal to 80 points, being met the of preparatory condition
One decision-tree model, terminates to improve processing operation, if less than 80 points, improve the second step of processing operation, adjustment is changed planes
The ratio of user and not renewal user, for example before by renewal user and not the ratio of renewal user be 1:5, now according to reality
By renewal user and not, the ratio of renewal user is adjusted to 1 to situation：8, extracted from subscriber's account data 100 renewal users and
This 900 subscriber's account data are divided into training data and test number by 800 not renewal users according to default division proportion
According to, and be trained and test again, until the first obtained decision-tree model meets preparatory condition.
Wherein, the table 2 in first embodiment is shown in default bill changes planes Forecasting Methodology, data target it is key
Indispensable index when data target in ranking, the table is model training.
First determining unit 606, if for meeting, by the decision-tree model of data input first to be predicted, and by first
The result data of decision-tree model output is defined as first renewal user's result data.
In embodiments of the present invention, after preparatory condition is met, by the decision-tree model of data input first to be predicted, and will
The result data of first decision-tree model output is defined as first renewal user's result data.For example, the data of training and test
For the data of 2 months 2017, in order to predict first renewal user's result data in April, then the data in March (are treated pre-
Survey data) the first decision-tree model of input, and the result data of output is defined as first renewal user's result data.
Compared with prior art, the embodiment of the present invention predicts that the data of renewal user refer to based on subscriber's account data definition
Mark, the data target is trained to model to play an important role.Wherein, default decision-tree model is carried out using training data
After training, the first decision-tree model after being trained, the model is tested, will if meeting preparatory condition if predicting the outcome
The first decision-tree model that data input to be predicted meets preparatory condition obtains first renewal user's result data, if predicting the outcome
Preparatory condition is unsatisfactory for, then training data and test data are adjusted, is trained and tests again after adjustment, until pre-
Survey result and meet preparatory condition, so as to obtain accurate first renewal user result data.
Referring to Fig. 7, in the second embodiment shown in Fig. 5 the second determining unit 502 refinement high-level schematic functional block diagram, bag
Include：
Second definition unit 701, for the data target based on search behavior data definition user's search behavior；
In embodiments of the present invention, the second definition unit 701 is screened to signaling call bill data first, extracts user
Search behavior data, further, carry out keyword match search, will have the keywords such as mobile phone, terminal in search behavior data
Data extract.
In embodiments of the present invention, the data target of user's search behavior can be defined with actual conditions.
It is preferred that, data target is as shown in the table 3 in first embodiment.
Second division unit 702, for being screened using data target and default screening rule to search behavior data,
And the search behavior data after screening are divided into training data and test data；
In embodiments of the present invention, the second division unit 702 is screened using data target to search behavior data, from
The serious data of data target missing, data are rejected in search behavior data and fill in wrong data and the data repeated, are obtained
First search behavior data are carried out postsearch screening by the first search behavior data according to default screening rule, wherein, according to default
Screening rule carry out postsearch screening to the first search behavior data and are divided into two steps, and the first step is：Weed out the first search behavior
Searching times are less than the secondary data of N (N is positive integer) in the period of certain in data, obtain the second search behavior data, here
Certain period can be one month, two months etc.；Second step is：According to the ratio pair of default renewal user and not renewal user
Second search behavior data carry out data pick-up.For example, the second search behavior data are 1000 users, wherein renewal user is
500 people, renewal user is not 500 people, and the ratio of default renewal user and not renewal user are 1:5, then from 1000 users
100 renewal users of middle extraction and 500 not renewal users, the search behavior data one after obtained screening have 600 use
Family, including the not renewal user of 100 renewal users and 500, if the ratio of default renewal user and not renewal user are 1:
3, then 100 renewal users and 300 not renewal user, the search behavior number after obtained screening are extracted from 1000 users
400 users are had according to one, including are the not renewal user of 100 renewal users and 300.
Afterwards, training data and test data are divided into the search behavior data after screening according to default division proportion.
For example, training data and test data ratio are 1:2, according to 1:2 ratio is drawn to 600 users in previous example
Point, obtained training data be 200 people, test data be 400 people, but in this 200 training datas renewal user with not
The ratio of renewal user is stilled need as 1:5, i.e., this 200 training datas include 33 renewal users (200 be multiplied by six/
One) and 167 not renewal user's (200 are multiplied by 5/6ths), renewal user and not renewal user in this 400 test datas
Ratio still need as 1:5, i.e., include 67 renewal users (400 are multiplied by 1/6th) and 333 in this 400 test datas
Individual not renewal user (400 are multiplied by 5/6ths).
It is preferred that, the ratio of test data and training data is 3:7.
Second training unit 703, for being trained using training data to default decision-tree model, is obtained after training
The second decision-tree model；
In embodiments of the present invention, when being trained to default decision-tree model, training data can be analyzed,
And automatically excluded the relatively low data target of importance according to material circumstance, and generate a tables of data, the tables of data point
For two kinds：The data target excluded and the data target not excluded.
Second test cell 704, predicts the outcome for testing determination to test data progress based on the second decision-tree model；
Second assessment unit 705, it is default whether the performance for assessing the second decision-tree model using predicting the outcome is satisfied with
Condition；
In embodiments of the present invention, preparatory condition is that user pre-sets, and can be configured and adjust according to actual conditions
It is whole.For example, preparatory condition is 80 points, then predicts the outcome and the performance satisfaction of the second decision-tree model is represented if more than or equal to 80
Preparatory condition, if predicting the outcome as 70 points, needs that the second decision-tree model is carried out to improve processing operation.Here perfect place
Reason operation is divided into two steps, and the first step is：After step S303 is completed, a tables of data can be obtained, the tables of data is observed, will be excluded
The data corresponding to data target fallen, which are rejoined in training data, to be trained, the second decision tree mould after improving
Type, tests the second decision-tree model after improving, is predicted the outcome, check this predict the outcome whether more than 80 points or
Equal to 80 points, if more than 80 points or equal to 80 points, being met the second decision-tree model of preparatory condition, terminate perfect
Processing operation, if less than 80 points, improve the second step of processing operation, not adjustment renewal user and the ratio of renewal user
Example, for example before by renewal user and not the ratio of renewal user be 1:5, now changed according to actual conditions by renewal user and not
The ratio of machine user is adjusted to 1：8, the not renewal user of 100 renewal users and 800 is extracted from search behavior data, according to
This 900 search behavior data are divided into training data and test data by default division proportion, are trained and are tested again,
Until the second obtained decision-tree model meets preparatory condition.
Wherein, the table 4 in first embodiment is shown in preset search behavior prediction method, data target it is key
Indispensable index when data target in ranking, the table is model training.
Second determining unit 706, if for meeting, by the decision-tree model of data input second to be predicted, and by second
The result data of decision-tree model output is defined as second renewal user's result data.
In embodiments of the present invention, after preparatory condition is met, by the decision-tree model of data input second to be predicted, and will
The result data of second decision-tree model output is defined as second renewal user's result data.For example, the data of training and test
For the data of 2 months 2017, in order to predict second renewal user's result data in April, then the data in March (are treated pre-
Survey data) the second decision-tree model of input, and the result data of output is defined as second renewal user's result data.
Compared with prior art, data of the embodiment of the present invention based on search behavior data definition user's search behavior refer to
Mark, the data target is trained to model to play an important role.Wherein, default decision-tree model is carried out using training data
After training, the second decision-tree model after being trained, the model is tested, will if meeting preparatory condition if predicting the outcome
The second decision-tree model that data input to be predicted meets preparatory condition obtains second renewal user's result data, if predicting the outcome
Preparatory condition is unsatisfactory for, then training data and test data are adjusted, is trained and tests again after adjustment, until pre-
Survey result and meet preparatory condition, so as to obtain accurate second renewal user result data.
Referring to Fig. 8, in the second embodiment shown in Fig. 5 the 3rd determining module refinement high-level schematic functional block diagram, including：
First comparing unit 801, be for comparing first renewal user's result data and second renewal user's result data
It is no identical；
3rd determining unit 802, if for identical, by first renewal user's result data or second renewal user's result
Data are defined as the data that predict the outcome；
Second comparing unit 803, if for differing, by first confidence level relevant with subscriber's account data and with searching
The second relevant confidence level of rope behavioral data is compared；
In embodiments of the present invention, the first decision tree of preparatory condition is met in Forecasting Methodology of being changed planes with default bill
During model, the first decision-tree model can provide the first confidence level, default being met with preset search behavior prediction method
During the first decision-tree model of condition, the first decision-tree model can provide the second confidence level, will be relevant with subscriber's account data
The first confidence level second confidence level relevant with search behavior data be compared.
4th determining unit 804, if being more than the second confidence level for the first confidence level, it is determined that first renewal user's result
Data are the data that predict the outcome, if the second confidence level is more than the first confidence level, it is determined that second renewal user's result data is pre-
Survey result data.
In embodiments of the present invention, differed in first renewal user's result data and second renewal user's result data
When, it is relevant with search behavior data that the second comparing unit 803 compares first confidence level relevant with subscriber's account data
The size of second confidence level, the data that more correctly predict the outcome can be relatively obtained according to the size of the confidence level, so as to
Changed planes prediction inventory with obtaining more accurate user, so as to more accurately predict the tendency of user's changes terminal.
Claims (10)
- The Forecasting Methodology 1. a kind of user changes planes, it is characterised in that methods described includes：Subscriber's account data and default bill according to getting Forecasting Methodology of changing planes determine first renewal user's result data；Search behavior data and preset search behavior prediction method according to getting determine second renewal user's result data and Target terminal model data；The first renewal user result data and the second renewal user result data are compared, and according to comparing knot Fruit determines the data that predict the outcome；Data are predicted the outcome according to described and the target terminal model data determines that user changes planes prediction inventory.
- 2. method according to claim 2, it is characterised in that the target terminal model data is arranged for user's searching times The data of the terminal models of N before name, N is positive integer, the data that predicted the outcome described in the basis and the target terminal model Data determine user change planes prediction inventory the step of include：The data of the terminal models of N before predict the outcome data and user's searching times ranking are associated Match somebody with somebody, determine that the user changes planes prediction inventory.
- 3. method according to claim 2, it is characterised in that subscriber's account data and default account that the basis is got The step of monodromy machine Forecasting Methodology determines first renewal user's result data includes：The data target of renewal user is predicted based on the subscriber's account data definition；The subscriber's account data are screened using the data target and default screening rule, and by the user after screening Billing data is divided into training data and test data；Default decision-tree model is trained using the training data, the first decision-tree model after being trained；Determination is tested based on first decision-tree model to test data progress to predict the outcome；Assess the performance of first decision-tree model using described predict the outcome and whether be satisfied with preparatory condition；If meeting, by the first decision-tree model described in data input to be predicted, and first decision-tree model is exported Result data is defined as the first renewal user result data.
- 4. method according to claim 2, it is characterised in that search behavior data that the basis is got and default search The step of rope behavior prediction method determines second renewal user's result data includes：Data target based on search behavior data definition user's search behavior；The search behavior data are screened using the data target and default screening rule, and by the search after screening Behavioral data is divided into training data and test data；Default decision-tree model is trained using the training data, the second decision-tree model after being trained；Determination is tested based on second decision-tree model to test data progress to predict the outcome；Assess the performance of second decision-tree model using described predict the outcome and whether be satisfied with preparatory condition；If meeting, by the second decision-tree model described in data input to be predicted, and second decision-tree model is exported Result data is defined as the second renewal user result data.
- 5. the method according to Claims 1-4 any one, it is characterised in that described to be tied to first renewal user Fruit data and the second renewal user result data are compared, and according to comparative result determine predict the outcome data the step of Including：Compare the first renewal user result data and whether the second renewal user result data is identical；If identical, the first renewal user result data or the second renewal user result data are defined as described pre- Survey result data；If differing, by first confidence level relevant with subscriber's account data, second confidence relevant with search behavior data Degree is compared；If first confidence level is more than second confidence level, it is determined that the first renewal user result data is described pre- Result data is surveyed, if second confidence level is more than first confidence level, it is determined that the second renewal user result data For the data that predict the outcome.
- The prediction meanss 6. a kind of user changes planes, it is characterised in that described device includes：First determining module, determines that first changes for Forecasting Methodology of being changed planes according to the subscriber's account data and default bill that get Machine user's result data；Second determining module, for determining that second changes according to the search behavior data and preset search behavior prediction method that get Machine user result data and target terminal model data；3rd determining module, for the first renewal user result data and the second renewal user result data progress Compare, and the data that predict the outcome are determined according to comparative result；4th determining module, for predict the outcome data according to and the target terminal model data to determine that user changes planes pre- Survey inventory.
- 7. device according to claim 6, it is characterised in that the target terminal model data is arranged for user's searching times Name before N terminal models data, N is positive integer, the 4th determining module specifically for：The data of the terminal models of N before predict the outcome data and user's searching times ranking are associated Match somebody with somebody, determine that the user changes planes prediction inventory.
- 8. device according to claim 7, it is characterised in that first determining module includes：First definition unit, the data target for predicting renewal user based on the subscriber's account data definition；First division unit, for being sieved using the data target and default screening rule to the subscriber's account data Choosing, and the subscriber's account data after screening are divided into training data and test data；First training unit, for being trained using the training data to default decision-tree model, after being trained First decision-tree model；First test cell, prediction knot is determined for carrying out test to the test data based on first decision-tree model Really；First assessment unit, for using described in predict the outcome assess first decision-tree model performance whether be satisfied with it is default Condition；First determining unit, if for meeting, by the first decision-tree model described in data input to be predicted, and by described first The result data of decision-tree model output is defined as the first renewal user result data.
- 9. device according to claim 7, it is characterised in that second determining module includes：Second definition unit, for the data target based on search behavior data definition user's search behavior；Second division unit, for being sieved using the data target and default screening rule to the search behavior data Choosing, and the search behavior data after screening are divided into training data and test data；Second training unit, for being trained using the training data to default decision-tree model, after being trained Second decision-tree model；Second test cell, prediction knot is determined for carrying out test to the test data based on second decision-tree model Really；Second assessment unit, for using described in predict the outcome assess second decision-tree model performance whether be satisfied with it is default Condition；Second determining unit, if for meeting, by the second decision-tree model described in data input to be predicted, and by described second The result data of decision-tree model output is defined as the second renewal user result data.
- 10. the device according to claim 6 to 9 any one, it is characterised in that the 3rd determining module includes：First comparing unit, be for comparing the first renewal user result data and the second renewal user result data It is no identical；3rd determining unit, if for identical, the first renewal user result data or second renewal user are tied Fruit data are defined as the data that predict the outcome；Second comparing unit, if for differing, by first confidence level relevant with subscriber's account data and with search behavior The second relevant confidence level of data is compared；4th determining unit, if being more than second confidence level for first confidence level, it is determined that described first changes planes use Family result data is the data that predict the outcome, if second confidence level is more than first confidence level, it is determined that described the Two renewal user's result datas are the data that predict the outcome.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710190104.7A CN107016460A (en) | 2017-03-27 | 2017-03-27 | User changes planes Forecasting Methodology and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710190104.7A CN107016460A (en) | 2017-03-27 | 2017-03-27 | User changes planes Forecasting Methodology and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107016460A true CN107016460A (en) | 2017-08-04 |
Family
ID=59445197
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710190104.7A Pending CN107016460A (en) | 2017-03-27 | 2017-03-27 | User changes planes Forecasting Methodology and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107016460A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109583949A (en) * | 2018-11-22 | 2019-04-05 | 中国联合网络通信集团有限公司 | A kind of user changes planes prediction technique and system |
| CN110648153A (en) * | 2018-06-11 | 2020-01-03 | 中国移动通信集团浙江有限公司 | Change tendency prediction method and device, electronic equipment and storage medium |
| CN113905078A (en) * | 2021-09-27 | 2022-01-07 | 中国联合网络通信集团有限公司 | Information pushing method, device, equipment and readable storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103577432A (en) * | 2012-07-26 | 2014-02-12 | 阿里巴巴集团控股有限公司 | Method and system for searching commodity information |
| CN104794636A (en) * | 2015-04-17 | 2015-07-22 | 南京大学 | Mobile phone model recommendation method based on user display scoring |
| CN104794195A (en) * | 2015-04-17 | 2015-07-22 | 南京大学 | Data mining method for finding potential telecommunication users changing cell phones |
| CN105678596A (en) * | 2016-03-23 | 2016-06-15 | 中国联合网络通信集团有限公司 | Mobile terminal replacement prediction method, and replacement prediction system thereof |
| CN105824833A (en) * | 2015-01-07 | 2016-08-03 | 苏宁云商集团股份有限公司 | Keyword recommendation method and system based on user behavior feedback |
-
2017
- 2017-03-27 CN CN201710190104.7A patent/CN107016460A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103577432A (en) * | 2012-07-26 | 2014-02-12 | 阿里巴巴集团控股有限公司 | Method and system for searching commodity information |
| CN105824833A (en) * | 2015-01-07 | 2016-08-03 | 苏宁云商集团股份有限公司 | Keyword recommendation method and system based on user behavior feedback |
| CN104794636A (en) * | 2015-04-17 | 2015-07-22 | 南京大学 | Mobile phone model recommendation method based on user display scoring |
| CN104794195A (en) * | 2015-04-17 | 2015-07-22 | 南京大学 | Data mining method for finding potential telecommunication users changing cell phones |
| CN105678596A (en) * | 2016-03-23 | 2016-06-15 | 中国联合网络通信集团有限公司 | Mobile terminal replacement prediction method, and replacement prediction system thereof |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110648153A (en) * | 2018-06-11 | 2020-01-03 | 中国移动通信集团浙江有限公司 | Change tendency prediction method and device, electronic equipment and storage medium |
| CN109583949A (en) * | 2018-11-22 | 2019-04-05 | 中国联合网络通信集团有限公司 | A kind of user changes planes prediction technique and system |
| CN113905078A (en) * | 2021-09-27 | 2022-01-07 | 中国联合网络通信集团有限公司 | Information pushing method, device, equipment and readable storage medium |
| CN113905078B (en) * | 2021-09-27 | 2023-07-04 | 中国联合网络通信集团有限公司 | Information pushing method, device, equipment and readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109670940A (en) | Credit Risk Assessment Model generation method and relevant device based on machine learning | |
| CN109934704A (en) | Information recommendation method, device, equipment and storage medium | |
| CN105894359A (en) | Order pushing method, device and system | |
| CN101951617B (en) | Mobile network communication quality evaluation method based on analytic hierarchy process | |
| CN103150696A (en) | Method and device for selecting potential customer of target value-added service | |
| CN106919579A (en) | A kind of information processing method and device, equipment | |
| CN109670962A (en) | Finance product method for pushing, device, equipment and storage medium based on big data | |
| CN102158879A (en) | Essential factor lost score data processing method and equipment | |
| CN109982367A (en) | Mobile terminal Internet access customer complaint prediction technique, device, equipment and storage medium | |
| CN108399565A (en) | Financial product recommendation apparatus, method and computer readable storage medium | |
| CN109063000A (en) | Question sentence recommended method, customer service system and computer readable storage medium | |
| CN110288350A (en) | User's Value Prediction Methods, device, equipment and storage medium | |
| CN111311030B (en) | User credit risk prediction method and device based on influence factor detection | |
| CN107016460A (en) | User changes planes Forecasting Methodology and device | |
| CN111815169A (en) | Business approval parameter configuration method and device | |
| CN107798615A (en) | Declaration form renews charge difficulty Forecasting Methodology and device | |
| CN110162958A (en) | For calculating the method, apparatus and recording medium of the synthesis credit score of equipment | |
| CN116915710A (en) | Traffic early warning method, device, equipment and readable storage medium | |
| CN109145932A (en) | User's gender prediction's method, device and equipment | |
| CN116362359A (en) | User satisfaction prediction method, device, equipment and medium based on AI big data | |
| CN113850669A (en) | User grouping method and device, computer equipment and computer readable storage medium | |
| CN109670964A (en) | Investment risk evaluating method, equipment, storage medium and device | |
| CN113361778A (en) | Information processing method and device applied to business continuity management platform | |
| CN110517132A (en) | Credit-graded approach, system, terminal and computer readable storage medium | |
| CN106126541B (en) | Personalized solution recommended method and system based on big data knowledge base |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170804 |
|
| RJ01 | Rejection of invention patent application after publication |