CN108510096A - Trade company's attrition prediction method, apparatus, equipment and storage medium - Google Patents
Trade company's attrition prediction method, apparatus, equipment and storage medium Download PDFInfo
- Publication number
- CN108510096A CN108510096A CN201710102115.5A CN201710102115A CN108510096A CN 108510096 A CN108510096 A CN 108510096A CN 201710102115 A CN201710102115 A CN 201710102115A CN 108510096 A CN108510096 A CN 108510096A
- Authority
- CN
- China
- Prior art keywords
- index feature
- trade company
- lost
- sample
- training
- 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
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"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0202—Market predictions or forecasting for commercial activities
Abstract
The invention discloses trade company's attrition prediction method, apparatus, equipment and storage medium, wherein method includes：The current index feature of the index feature for being lost in different time of the trade company before loss and online trade company is obtained respectively, and training sample is generated according to the information got；It is trained to obtain attrition prediction model according to training sample；The input of the index feature of trade company to be predicted is lost in prediction model, obtain whether being lost in and will how long after the prediction result that is lost in.Using scheme of the present invention, the accuracy etc. of prediction result can be improved.
Description
【Technical field】
The present invention relates to Computer Applied Technology, more particularly to trade company's attrition prediction method, apparatus, equipment and storage is situated between
Matter.
【Background technology】
With the development of internet, the services for life such as clothing, food, lodging and transportion -- basic necessities of life, luxurious life by eating, drinking and playing of netizen industry is gradually by all kinds of electric business
It is permeated to (o2o, Online To Offline) product under line on line, monopolization is gradually broken, and there are more cooperations pair in trade company
As it can be selected that and trade company be electric business platform there are lifeblood, it is therefore desirable to control in time is carried out to its situation, such as need and
Early predict which trade company will be lost in and will how long after be lost in etc., corresponding keep measure etc. to take.
But in the prior art, can only calculate roughly out whether trade company can be lost in, it is unpredictable go out by how long after
Be lost in etc., i.e., the content that prediction result is capable of providing is limited, and accuracy is poor, to be not easy to the development of follow-up work.
【Invention content】
In view of this, the present invention provides trade company's attrition prediction method, apparatus, equipment and storage medium, can improve pre-
Survey the accuracy of result.
Specific technical solution is as follows：
A kind of trade company's attrition prediction method, including：
The current index of the index feature for being lost in different time of the trade company before loss and online trade company is obtained respectively
Feature, and training sample is generated according to the information got；
It is trained to obtain attrition prediction model according to training sample；
The index feature of trade company to be predicted is inputted into the attrition prediction model, obtain whether being lost in and will how long
The prediction result being lost in later.
A kind of trade company's attrition prediction device, including：Pretreatment unit and predicting unit；
The pretreatment unit, for obtain respectively be lost in different time of the trade company before loss index feature and
The current index feature of online trade company, and training sample is generated according to the information got；It is trained and is flowed according to training sample
Lose prediction model；The attrition prediction model is sent to the predicting unit；
The predicting unit, for the index feature of trade company to be predicted to be inputted the attrition prediction model, obtain whether
Can be lost in and will how long after be lost in prediction result.
A kind of computer equipment, including memory, processor and be stored on the memory and can be in the processor
The computer program of upper operation, the processor realize method as described above when executing described program.
A kind of computer readable storage medium is stored thereon with computer program, real when described program is executed by processor
Now method as described above.
It can be seen that using scheme of the present invention based on above-mentioned introduction, can obtain respectively and be lost in trade company before loss
Different time the current index feature of index feature and online trade company, and training sample is generated according to the information got
This, obtains attrition prediction model, and then carry out actual prediction, prediction knot according to attrition prediction model according to training sample later
In fruit other than the information for including whether to be lost in, will also include by how long after be lost in information, it is pre- to enrich
The content in result is surveyed, the accuracy of prediction result is improved.
【Description of the drawings】
Fig. 1 is the flow chart of trade company's attrition prediction embodiment of the method for the present invention.
Fig. 2 is the flow chart of the embodiment of the method for the importance measures value of the present invention for obtaining index feature.
Fig. 3 is the schematic diagram of the Drain Causes of the present invention for showing sales force.
Fig. 4 is the composed structure schematic diagram of trade company's attrition prediction device embodiment of the present invention.
Fig. 5 shows the block diagram of the exemplary computer system/server 12 suitable for being used for realizing embodiment of the present invention.
【Specific implementation mode】
In order to keep technical scheme of the present invention clearer, clear, develop simultaneously embodiment referring to the drawings, to institute of the present invention
The scheme of stating is described in further detail.
Fig. 1 is the flow chart of trade company's attrition prediction embodiment of the method for the present invention, as shown in Figure 1, including in detail below
Realization method：
In 101, the index feature for being lost in different time of the trade company before loss is obtained respectively and online trade company works as
Preceding index feature, and training sample is generated according to the information got；
In 102, trained to obtain attrition prediction model according to training sample；
In 103, the input of the index feature of trade company to be predicted is lost in prediction model, obtains whether being lost in and will be
How long later the prediction result of loss.
That is, in scheme of the present invention, needs to obtain training sample first, then be flowed according to training sample
It loses prediction model, and then actual prediction is carried out according to attrition prediction model, in addition to including whether to be lost in prediction result
Except information, will also include by how long after the information that is lost in, to improve the standard of prediction result compared with the prior art
True property.
The specific implementation of each part mentioned above is described in detail individually below.
One) training sample
In scheme of the present invention, it can define in the following way and be lost in trade company and online trade company.
It has been lost in trade company：
If confirm the last one online single group of any trade company downtime be located at [the first scheduled durations of date-,
The second scheduled durations of date-] time range in, then can determine the trade company be lost in trade company；
The downtime of the last one online single group, the loss time of Ji Zhi trade companies.
Online trade company：
If it is determined that there was an online single group in any trade company in the time range of [the second scheduled durations of date-, date], and
There is online single group in the time range of [the first scheduled durations of date-, the second scheduled durations of date-], then can determine the trade company
For online trade company.
Wherein, date indicates that current time, the first scheduled duration are more than the second scheduled duration.
The specific value of first scheduled duration and the second scheduled duration can be decided according to the actual requirements, for example, first is pre-
The long value of timing can be 120 days, and the value of the second scheduled duration can be 30 days.
It has been lost in trade company for each, the index feature of its specified N number of different time before loss can be obtained respectively,
N is the positive integer more than one.
The specific value of N can equally be decided according to the actual requirements, for example, the value of N can be 3, N number of different time can be distinguished
For：It is lost in the last week, be lost in the last fortnight and is lost in preceding surrounding.
Correspondingly, it can obtain respectively and be lost in one week time point of the trade company before being lost in the time, be lost in the last fortnight
Index feature before time point and loss when the time point of surrounding.
For each online trade company, its current index feature can be obtained respectively.
Based on the information got, a series of training samples are produced, may include in each training sample：Sample label with
And index feature.
Sample label may include：It is not lost in, will be lost in after [when loss m- index feature corresponding time] duration.
As previously described, it is assumed that the value of N is 3, and N number of different time is respectively：It is lost in the last week, is lost in the last fortnight and stream
Surrounding before losing, then sample label may include：Be not lost in, will be lost in after one week, will be lost in after the two weeks, will be in surrounding
It is lost in later.
For another example, it may include following information in a certain training sample：[(index is special for index feature when being lost in the last week
Sign), will be lost in after one week (sample label)].
It may include flowing water class index feature, operation class index feature, trade company's Attribute class index feature, basis in index feature
Attribute class index feature and competing to class index feature etc..
Wherein, it can further comprise following index feature again in all kinds of index features：
1) flowing water class index feature
Verification flowing water chow ring ratio, verification several weeks ring ratio, it is upper recently it is single away from number of days today, finally place an order away from number of days today, group
Single online several weeks ring ratio, reimbursement flowing water chow ring ratio, subsidy flowing water chow ring ratio, complains quantity chow ring ratio, visits at operation flowing water chow ring ratio
Visit several weeks ring ratio, flowing water is runed in shop, the equal reimbursement flowing water in shop, shop subsidize flowing water, shop sell number, the equal reimbursement number in shop ....
2) class index feature is operated
Sale id identity belonging to sale id, shops belonging to shops's visit number, shops complains number, upper single number, places an order
Number, shops's single group change price number, shops's visit capacity (PV, Page View), shops independence user sessions (UV, Unique
Visitor), number is claimed by shops's collection number, shops, searching times, businessman log in trade company's end number ....
3) trade company's Attribute class index feature
Whether head, ka states, it is whether online, whether have discount, whether increase newly, whether be lost in, whether reach the standard grade, whether under
Line, whether automatically renewed, whether stored value card is online, whether arrive shop pay online, online single group number, whether prepay guarantor measure ....
4) primary attribute category feature
Commercial circle id, administrative area id, county-level city id, city id, great Qu id, system id, level-one category id, two level category id,
Level-one hang down class id, two level hang down class id, shops's state, shops source ....
5) competing to class index feature
Electric business platform a shares, electric business platform b shares, electric business platform c shares, electric business platform a sales volumes, electric business platform b pins
Online whether amount, electric business platform c sales volumes, electric business platform a, whether the online number of days of electric business platform a, electric business platform b are online, electric business is flat
Whether the online number of days of platform b, electric business platform c online, the online number of days of electric business platform c ....
Above-mentioned chow ring is than the calculation of class index feature：
By taking flowing water chow ring ratio as an example, flowing water chow ring ratio=(this week flowing water summation-last week flowing water summation)/last week flowing water is total
With.
It is specifically included in index feature in training sample in which classification and each classification and which is specifically included respectively
Index feature can be decided according to the actual requirements, and be not limited to illustrated above.
Two) attrition prediction model
After obtaining enough training samples, you can trained to obtain attrition prediction model according to training sample.
In scheme of the present invention, the Method Modeling of random forest can be used, essence is but not to be based on decision tree
One decision tree, but more decision trees.
That is, being trained to obtain M decision tree according to training sample, M is the positive integer more than one, and specific value can root
Depending on actual needs.
Wherein, the mode for obtaining every decision tree is trained to may include：P training sample is extracted from training sample, P is small
The sum of training sample is indicated in Q, Q；It is trained to obtain a decision tree according to the P training sample extracted.
In addition, it is assumed that including X index feature in each training sample, then can also be extracted from X index feature
Go out Y index feature, Y retains the Y index therein respectively less than or equal to X for each training sample extracted
Feature is used for the training of decision tree.
How to carry out extracting and being decided according to the actual requirements, for example can randomly select, it can also be according to preset
Other decimation rules are extracted.
It illustrates：
Assuming that co-existing in 30 training samples, 20 training samples can be randomly selected out from 30 training samples；
Assuming that including 10 index features, respectively 1~index feature of index feature 10 in each training sample, at random
1~index feature of index feature 7 is extracted, i.e., only retains 1~index feature of index feature in the training sample each extracted
7；
Using the training sample obtained after above-mentioned extraction twice, can train to obtain a decision tree.
In the manner described above, it can train to obtain more decision trees.
The training sample that used training sample is all not all of when being trained due to every decision tree, and it is used
Index feature may not be whole index features, therefore can largely improve big data higher-dimension sample characteristics
Modeling speed is practised, while the influence of noise data and missing data can also be weakened, there is relatively good robustness.
Three) wastage is predicted
As previously mentioned, attrition prediction model includes more decision trees, then correspondingly, when needing to wait for any online
When prediction trade company is predicted, can the index feature of trade company to be predicted be inputed into every decision tree respectively, to respectively obtain
The prediction result of every decision tree, and then final required prediction result can be determined by ballot mode.
It illustrates：
Assuming that 3 decision trees are co-existed in, for an online trade company to be predicted, the prediction result difference of each decision tree
For：It is not lost in, will be lost in after one week, will be lost in after one week；
So, final required prediction result is to be lost in after one week.
Since more decision trees carry out by ballot mode the prediction of final prediction result, it is not easy over-fitting occur
Phenomenon, and can ensure global optimum.
Four) right assessment
In addition to the above, also, it has been proposed that can be true according to the training sample not being extracted in two) in scheme of the present invention
The weight of each index feature is made, and then weight is determined as the main of influence trade company loss more than the index feature of predetermined threshold
Factor.
Specifically, it after every training obtains a decision tree, can determine to be extracted when training the decision tree respectively
Y index feature importance measures value；
Wherein, the method for determination of the importance measures value of each index feature i includes：
Q-P training sample not being extracted when obtaining the training decision tree only retains and each of is not extracted trained sample
The Y index feature extracted in this will remain with Q-P training sample of Y index feature as assessment sample；
Y index feature in each assessment sample is inputted into the decision tree respectively, obtains prediction result, statistical forecast knot
Fruit correctly assesses sample number R；
Noise disturbance is carried out to the index feature i in each assessment sample respectively, and after each carrying out noise disturbance respectively
Assessment sample in Y index feature input the decision tree, obtain prediction result, statistical forecast result correctly assesses sample
Number R '；
Using R and R ' difference as the importance measures value of index feature i；
The importance measures value corresponding to different decision trees of index feature i is averaged, the power as index feature i
Weight.
Based on above-mentioned introduction, Fig. 2 is the embodiment of the method for the importance measures value of the present invention for obtaining index feature
Flow chart, as shown in Fig. 2, including realization method in detail below.
In 201, P training sample is extracted from Q training sample, and from the X in each training sample extracted
Y index feature is extracted in a index feature respectively.
P<Q, Y≤X.
The index feature extracted from each training sample is identical, for example, extracting verification flowing water chow ring ratio, verification number
The index features such as chow ring ratio.
In 202, it is trained to obtain a decision tree Tb according to the training sample extracted.
How to train to obtain decision tree to be the prior art according to training sample.
In 203, using Q-P training sample not being extracted, remaining with the Y index feature extracted respectively as
Sample is assessed, it is each to assess the outer data Lb of sample composition bag.
That is, which index feature the training sample used in training decision tree includes, assess in sample
To include these same index features.
In 204, data classification is carried out to data Lb outside bag using decision tree Tb, counts the number R correctly to classify.
Y index feature in each assessment sample is inputted into decision tree Tb respectively, obtains prediction result, statistical forecast
As a result sample number R is correctly assessed.
Since the sample label in each assessment sample is known, then sample label and prediction result can be carried out
Compare, for example, the sample label of a certain assessment sample is that will be lost in after one week, and the prediction result of the assessment sample is also
It will be lost in after one week, then then thinking that the prediction result of the assessment sample is correct.
In 205, for each index feature i in Y index feature in assessment sample, respectively according to 206~208 institutes
The mode of stating determines its importance measures value.
In 206, noise disturbance is carried out to the index feature i in each assessment sample respectively, number outside the bag after being disturbed
According to Lb '.
Noise disturbance is carried out to the index feature i in each assessment sample, that is, refers to the random value for changing index feature i.
In 207, data classification is carried out to data Lb ' outside bag using decision tree Tb, counts the number R ' correctly to classify.
Y index feature in the assessment sample after each progress noise disturbance is inputted into decision tree Tb respectively, is obtained
Prediction result, statistical forecast result correctly assess sample number R '.
In 208, using R and R ' difference as the importance measures value of index feature i.
According to mode shown in Fig. 2, the importance measures value of the index feature i corresponding to decision tree Tb can be got, it is right
In other decision trees, the importance measures value of index feature i should can be equally got.
So, after getting the importance measures value corresponding to the index feature i of all decision trees, it is flat that it can be sought
Mean value, and using the average value found out as the weight of index feature i.
In the same way, the weight of each index feature can be got respectively.
As can be seen that for an index feature, what weight embodied is that the value generation of the index feature is slightly disturbed
The average decrement of prediction accuracy and the prediction accuracy before disturbance after dynamic, decrement is bigger, illustrates the index feature
Influence for predicting accuracy is bigger, then importance is also bigger.
After the weight for respectively obtaining each index feature, the weight that value is more than predetermined threshold, institute can be therefrom filtered out
Stating the specific value of threshold value can be decided according to the actual requirements, and then be determined as influencing by the corresponding index feature of the weight filtered out
The principal element that trade company is lost in.
In this way, predict trade company whether can be lost in and will how long after be lost in except, can also further determine that out
The principal element being lost in is influenced, that is, determines Drain Causes.
For an electric business platform, according to mode of the present invention determine to be lost in trade company and Drain Causes it
Afterwards, it can design and draw loss system, result will be lost in and Drain Causes push and show respectively sales force, sales force to receive
After the message that will be lost in trade company, related pages can be opened and intuitively check the reason of trade company will be lost in, so as to
Trade company carries out targetedly follow-up is handled or objective feelings are safeguarded etc. before being lost in.
Showing the Drain Causes of the trade company of sales force can be：The principal element that trade company is lost in is influenced (to filter out
Weight is more than the index feature of predetermined threshold) in there is the content of abnormal factor (index feature).
Fig. 3 is the schematic diagram of the Drain Causes of the present invention for showing sales force, as shown in figure 3, exception therein
Reason is Drain Causes.
For the Drain Causes shown, sales force can take measure of correspondingly keeping, such as, it is seen that the complaint of trade company
Number increase, trade company can be contacted in time with understand trade company sell present situation and complain reason, keep trade company viscosity, it is seen that quotient
The online single group number at family declines, and can contact single group etc. new in trade company in time.
It is the introduction about embodiment of the method above, below by way of device embodiment, to scheme of the present invention into traveling
One step explanation.
Fig. 4 is the composed structure schematic diagram of trade company's attrition prediction device embodiment of the present invention, as shown in figure 4, including：
Pretreatment unit 401 and predicting unit 402.
Pretreatment unit 401, for obtain respectively be lost in different time of the trade company before loss index feature and
The current index feature of online trade company, and training sample is generated according to the information got；It is trained and is flowed according to training sample
Lose prediction model；Attrition prediction model is sent to predicting unit 402.
Whether predicting unit 402 obtains to be lost in for the index feature input of trade company to be predicted to be lost in prediction model
And it will be in the prediction result of how long later loss.
Wherein, it may particularly include in pretreatment unit 401：Sample acquisition subelement 4011 and model training subelement
4012。
Sample acquisition subelement 4011, is used for：
If confirm the last one online single group of any trade company downtime be located at [the first scheduled durations of date-,
The second scheduled durations of date-] time range in, it is determined that trade company is to be lost in trade company, and obtain respectively and be lost in trade company and exist
The index feature of different time before loss；
If it is determined that there was an online single group in any trade company in the time range of [the second scheduled durations of date-, date], and
Had online single group in the time range of [the first scheduled durations of date-, the second scheduled durations of date-], it is determined that trade company be
Line trade company, and obtain the current index feature of online trade company；
Wherein, date indicates that current time, the first scheduled duration are more than the second scheduled duration；
Training sample is generated according to the information got, is sent to model training subelement 4012.
Model training subelement 4012, for training to obtain attrition prediction model according to training sample, and by attrition prediction
Model is sent to predicting unit 402.
The downtime of the last one online single group, the loss time of Ji Zhi trade companies.
The specific value of first scheduled duration and the second scheduled duration can be decided according to the actual requirements, for example, first is pre-
The long value of timing can be 120 days, and the value of the second scheduled duration can be 30 days.
It is lost in trade company for each, it is specified N number of before loss that sample acquisition subelement 4011 can obtain it respectively
The index feature of different time, N are the positive integer more than one.
The specific value of N can equally be decided according to the actual requirements, for example, the value of N can be 3, N number of different time can be distinguished
For：It is lost in the last week, be lost in the last fortnight and is lost in preceding surrounding.
Based on the information got, sample acquisition subelement 4011 produces a series of training samples, each training sample
In may include：Sample label and index feature.
Sample label includes：It is not lost in, will be lost in after [when loss m- index feature corresponding time] duration, stream
Lose the downtime that the time is the last one online single group.
As previously described, it is assumed that the value of N is 3, and N number of different time is respectively：It is lost in the last week, is lost in the last fortnight and stream
Surrounding before losing, then sample label may include：Be not lost in, will be lost in after one week, will be lost in after the two weeks, will be in surrounding
It is lost in later.
For another example, it may include following information in a certain training sample：[(index is special for index feature when being lost in the last week
Sign), will be lost in after one week (sample label)].
It may include flowing water class index feature, operation class index feature, trade company's Attribute class index feature, basis in index feature
Attribute class index feature and competing to class index feature etc..
After obtaining enough training samples, model training subelement 4012 can be trained according to training sample to be flowed
Prediction model is lost, may include more decision trees in attrition prediction model.
I.e. model training subelement 4012 can train to obtain M decision tree according to training sample, and M is just whole more than one
Number.
Correspondingly, the index feature of trade company to be predicted can be inputted every decision tree by predicting unit 402 respectively, obtain every
The prediction result of decision tree, and final required prediction result is determined by ballot mode.
Wherein, model training subelement 4012 trains the mode for obtaining every decision tree may include：It is taken out from training sample
P training sample is taken out, P is less than Q, and Q indicates the sum of training sample；It trains to obtain one according to the P training sample extracted
Decision tree.
In addition, it is assumed that include X index feature in each training sample, then model training subelement 4012 can also be from
Y index feature is extracted in X index feature, Y retains less than or equal to X for each training sample extracted respectively
The Y index feature therein is used for the training of decision tree.
How to carry out extracting and being decided according to the actual requirements, for example can randomly select, it can also be according to preset
Other decimation rules are extracted.
In addition, model training subelement 4012 can be further used for, determined respectively according to the training sample not being extracted
The weight of index feature, the index feature that weight is more than to predetermined threshold are determined as influencing the principal element that trade company is lost in.
Specifically, model training subelement 4012 can respectively be determined to train after every training obtains a decision tree
The importance measures value of the Y index feature extracted when decision tree；
Wherein, the method for determination of the importance measures value of each index feature i includes：
Q-P training sample not being extracted when training decision tree is obtained, the Y index spy extracted will be remained with respectively
The each training sample of sign not being extracted is as assessment sample；
Y index feature in each assessment sample is inputted into decision tree respectively, obtains prediction result, statistical forecast result
Correctly assessment sample number R；
Noise disturbance is carried out to the index feature i in each assessment sample respectively, and after each carrying out noise disturbance respectively
Assessment sample in Y index feature input decision tree, obtain prediction result, statistical forecast result correctly assesses sample number
R’；
Using R and R ' difference as the importance measures value of index feature i；
The importance measures value corresponding to different decision trees of index feature i is averaged, the power as index feature i
Weight.
As can be seen that for an index feature, what weight embodied is that the value generation of the index feature is slightly disturbed
The average decrement of prediction accuracy and the prediction accuracy before disturbance after dynamic, decrement is bigger, illustrates the index feature
Influence for predicting accuracy is bigger, then importance is also bigger.
After the weight for respectively obtaining each index feature, the weight that value is more than predetermined threshold can be therefrom filtered out, so
The corresponding index feature of the weight filtered out is determined as afterwards to influence the principal element of trade company's loss.
In this way, predict trade company whether can be lost in and will how long after be lost in except, can also further determine that out
The principal element being lost in is influenced, that is, determines Drain Causes.
For an electric business platform, according to mode of the present invention determine to be lost in trade company and Drain Causes it
Afterwards, it can design and draw loss system, result will be lost in and Drain Causes push and show respectively sales force, sales force to receive
After the message that will be lost in trade company, related pages can be opened and intuitively check the reason of trade company will be lost in, so as to
Trade company carries out targetedly follow-up is handled or objective feelings are safeguarded etc. before being lost in.
Showing the Drain Causes of the trade company of sales force can be：The principal element that trade company is lost in is influenced (to filter out
Weight is more than the index feature of predetermined threshold) in there is the content of abnormal factor (index feature).
The specific workflow of Fig. 4 shown device embodiments please refers to the respective description in preceding method embodiment, no longer
It repeats.
Fig. 5 shows the block diagram of the exemplary computer system/server 12 suitable for being used for realizing embodiment of the present invention.
The computer system/server 12 that Fig. 5 is shown is only an example, should not be to the function and use scope of the embodiment of the present invention
Bring any restrictions.
As shown in figure 5, computer system/server 12 is showed in the form of universal computing device.Computer system/service
The component of device 12 can include but is not limited to：One or more processor (processing unit) 16, memory 28 connect not homology
The bus 18 of system component (including memory 28 and processor 16).
Bus 18 indicates one or more in a few class bus structures, including memory bus or Memory Controller,
Peripheral bus, graphics acceleration port, processor or the local bus using the arbitrary bus structures in a variety of bus structures.It lifts
For example, these architectures include but not limited to industry standard architecture (ISA) bus, microchannel architecture (MAC)
Bus, enhanced isa bus, Video Electronics Standards Association (VESA) local bus and peripheral component interconnection (PCI) bus.
Computer system/server 12 typically comprises a variety of computer system readable media.These media can be appointed
What usable medium that can be accessed by computer system/server 12, including volatile and non-volatile media, it is moveable and
Immovable medium.
Memory 28 may include the computer system readable media of form of volatile memory, such as random access memory
Device (RAM) 30 and/or cache memory 32.Computer system/server 12 may further include it is other it is removable/no
Movably, volatile/non-volatile computer system storage medium.Only as an example, storage system 34 can be used for reading and writing
Immovable, non-volatile magnetic media (Fig. 5 do not show, commonly referred to as " hard disk drive ").It, can although being not shown in Fig. 5
To provide for the disc driver to moving non-volatile magnetic disk (such as " floppy disk ") read-write, and to removable non-volatile
Property CD (such as CD-ROM, DVD-ROM or other optical mediums) read and write CD drive.In these cases, each to drive
Dynamic device can be connected by one or more data media interfaces with bus 18.Memory 28 may include at least one program
There is one group of (for example, at least one) program module, these program modules to be configured to perform the present invention for product, the program product
The function of each embodiment.
Program/utility 40 with one group of (at least one) program module 42 can be stored in such as memory 28
In, such program module 42 includes --- but being not limited to --- operating system, one or more application program, other programs
Module and program data may include the realization of network environment in each or certain combination in these examples.Program mould
Block 42 usually executes function and/or method in embodiment described in the invention.
Computer system/server 12 can also be (such as keyboard, sensing equipment, aobvious with one or more external equipments 14
Show device 24 etc.) communication, it is logical that the equipment interacted with the computer system/server 12 can be also enabled a user to one or more
Letter, and/or any set with so that the computer system/server 12 communicated with one or more of the other computing device
Standby (such as network interface card, modem etc.) communicates.This communication can be carried out by input/output (I/O) interface 22.And
And computer system/server 12 can also pass through network adapter 20 and one or more network (such as LAN
(LAN), wide area network (WAN) and/or public network, such as internet) communication.As shown in figure 5, network adapter 20 passes through bus
18 communicate with other modules of computer system/server 12.It should be understood that although not shown in the drawings, computer can be combined
Systems/servers 12 use other hardware and/or software module, including but not limited to：Microcode, device driver, at redundancy
Manage unit, external disk drive array, RAID system, tape drive and data backup storage system etc..
Processor 16 is stored in the program in memory 28 by operation, to perform various functions at application and data
Reason, such as realize the method in embodiment illustrated in fig. 1, i.e., the index for being lost in different time of the trade company before loss is obtained respectively
The current index feature of feature and online trade company, and training sample is generated according to the information got, it is instructed according to training sample
Get attrition prediction model, the input of the index feature of trade company to be predicted be lost in prediction model, obtain whether being lost in and
It will be in the prediction result of how long later loss.
Wherein, the index feature for being lost in specified N number of different time of the trade company before loss can be obtained respectively, and N is big
In one positive integer.
The each training sample generated includes：Sample label and index feature；
Sample label includes：It is not lost in, will be lost in after [when loss m- index feature corresponding time] duration, stream
Lose the downtime that the time is the last one online single group.
It can be trained to obtain M decision tree according to training sample, M is the positive integer more than one, and M decision tree collectively forms stream
Lose prediction model correspondingly can input every decision tree by the index feature of trade company to be predicted respectively, obtain every decision tree
Prediction result, and final required prediction result is determined by ballot mode.
Wherein, the mode for obtaining every decision tree is trained to may include：P training sample is extracted from training sample, P is small
It indicates the sum of training sample in Q, Q, is trained to obtain a decision tree according to the P training sample extracted.
Assuming that including X index feature in each training sample, then the mode that training obtains every decision tree may be used also
Further comprise：Y index feature is extracted from X index feature, Y is less than or equal to X, for each training extracted
Sample retains Y index feature therein respectively, is used for the training of decision tree.
On this basis, the weight of each index feature can be also determined according to the training sample not being extracted, weight is big
It is determined as influencing the principal element of trade company's loss in the index feature of predetermined threshold.
The present invention discloses a kind of computer readable storage mediums, are stored thereon with computer program, the program quilt
Processor will realize the method in embodiment as shown in Figure 1 when executing.
The arbitrary combination of one or more computer-readable media may be used.Computer-readable medium can be calculated
Machine readable signal medium or computer readable storage medium.Computer readable storage medium for example can be --- but it is unlimited
In --- electricity, system, device or the device of magnetic, optical, electromagnetic, infrared ray or semiconductor, or the arbitrary above combination.It calculates
The more specific example (non exhaustive list) of machine readable storage medium storing program for executing includes：Electrical connection with one or more conducting wires, just
It takes formula computer disk, hard disk, random access memory (RAM), read-only memory (ROM), erasable type and may be programmed read-only storage
Device (EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-ROM), light storage device, magnetic memory device,
Or above-mentioned any appropriate combination.In this document, can be any include computer readable storage medium or storage journey
The tangible medium of sequence, the program can be commanded the either device use or in connection of execution system, device.
Computer-readable signal media may include in a base band or as the data-signal that a carrier wave part is propagated,
Wherein carry computer-readable program code.Diversified forms may be used in the data-signal of this propagation, including --- but
It is not limited to --- electromagnetic signal, optical signal or above-mentioned any appropriate combination.Computer-readable signal media can also be
Any computer-readable medium other than computer readable storage medium, which can send, propagate or
Transmission for by instruction execution system, device either device use or program in connection.
The program code for including on computer-readable medium can transmit with any suitable medium, including --- but it is unlimited
In --- wireless, electric wire, optical cable, RF etc. or above-mentioned any appropriate combination.
It can be write with one or more programming languages or combinations thereof for executing the computer that operates of the present invention
Program code, described program design language include object oriented program language-such as Java, Smalltalk, C++,
Further include conventional procedural programming language-such as " C " language or similar programming language.Program code can be with
It fully executes, partly execute on the user computer on the user computer, being executed as an independent software package, portion
Divide and partly executes or executed on a remote computer or server completely on the remote computer on the user computer.
Be related in the situation of remote computer, remote computer can pass through the network of any kind --- including LAN (LAN) or
Wide area network (WAN)-be connected to subscriber computer, or, it may be connected to outer computer (such as carried using Internet service
It is connected by internet for quotient).
In several embodiments provided by the present invention, it should be understood that disclosed device and method etc. can pass through
Other modes are realized.For example, the apparatus embodiments described above are merely exemplary, for example, the division of the unit,
Only a kind of division of logic function, formula that in actual implementation, there may be another division manner.
The unit illustrated as separating component may or may not be physically separated, aobvious as unit
The component shown may or may not be physical unit, you can be located at a place, or may be distributed over multiple
In network element.Some or all of unit therein can be selected according to the actual needs to realize the mesh of this embodiment scheme
's.
In addition, each functional unit in each embodiment of the present invention can be integrated in a processing unit, it can also
It is that each unit physically exists alone, it can also be during two or more units be integrated in one unit.Above-mentioned integrated list
The form that hardware had both may be used in member is realized, can also be realized in the form of hardware adds SFU software functional unit.
The above-mentioned integrated unit being realized in the form of SFU software functional unit can be stored in one and computer-readable deposit
In storage media.Above-mentioned SFU software functional unit is stored in a storage medium, including some instructions are used so that a computer
It is each that equipment (can be personal computer, server or the network equipment etc.) or processor (processor) execute the present invention
The part steps of embodiment the method.And storage medium above-mentioned includes：USB flash disk, mobile hard disk, read-only memory (ROM), with
Machine accesses the various media that can store program code such as memory (RAM), magnetic disc or CD.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all essences in the present invention
With within principle, any modification, equivalent substitution, improvement and etc. done should be included within the scope of protection of the invention god.
Claims (20)
1. a kind of trade company's attrition prediction method, which is characterized in that including：
The current index feature of the index feature for being lost in different time of the trade company before loss and online trade company is obtained respectively,
And training sample is generated according to the information got；
It is trained to obtain attrition prediction model according to training sample；
The index feature of trade company to be predicted is inputted into the attrition prediction model, obtain whether being lost in and will how long after
The prediction result of loss.
2. according to the method described in claim 1, it is characterized in that,
It is described to obtain the current index of the index feature for being lost in different time of the trade company before loss and online trade company respectively
Feature includes：
If confirming, the downtime of the last one online single group of any trade company is located at [the first scheduled durations of date-, date- the
Two scheduled durations] time range in, it is determined that the trade company is to be lost in trade company, and be lost in trade company described in obtaining respectively and existed
The index feature of different time before loss；
If it is determined that there was an online single group in any trade company in the time range of [the second scheduled durations of date-, date], and
There is online single group in the time range of [the first scheduled durations of date-, the second scheduled durations of date-], it is determined that the trade company
For online trade company, and obtain the current index feature of the online trade company；
Wherein, the date indicates that current time, first scheduled duration are more than second scheduled duration.
3. according to the method described in claim 1, it is characterized in that,
Described obtain respectively has been lost in the index feature of different time of the trade company before loss and includes：
The index feature for being lost in specified N number of different time of the trade company before loss is obtained respectively, and N is just whole more than one
Number；
The each training sample generated includes：Sample label and index feature；
The sample label includes：It is not lost in, will be lost in after [when loss m- index feature corresponding time] duration, stream
Lose the downtime that the time is the last one online single group.
4. according to the method described in claim 1, it is characterized in that,
The index feature includes：Flowing water class index feature, operation class index feature, trade company's Attribute class index feature, basis belong to
Property class index feature and competing to class index feature.
5. according to the method described in claim 1, it is characterized in that,
It is described train to obtain attrition prediction model according to training sample include：
It is trained to obtain M decision tree according to training sample, M is the positive integer more than one；
The index feature by trade company to be predicted inputs the attrition prediction model, obtain whether being lost in and will how long
The prediction result being lost in later includes：
The index feature of the trade company to be predicted is inputted into every decision tree respectively, obtains the prediction result of every decision tree, and
Final required prediction result is determined by ballot mode.
6. according to the method described in claim 5, it is characterized in that,
Training obtains the mode of every decision tree：
P training sample is extracted from training sample, P is less than Q, and Q indicates the sum of training sample；
It is trained to obtain a decision tree according to the P training sample extracted.
7. according to the method described in claim 6, it is characterized in that,
Include X index feature in each training sample；
The mode that the training obtains every decision tree further comprises：
Y index feature is extracted from X index feature, Y is less than or equal to X, for each training sample extracted, divides
Do not retain the Y index feature therein, is used for the training of decision tree.
8. the method according to the description of claim 7 is characterized in that
This method further comprises：
The weight of each index feature is determined according to the training sample not being extracted；
The index feature that weight is more than to predetermined threshold is determined as influencing the principal element that trade company is lost in.
9. according to the method described in claim 8, it is characterized in that,
The training sample that the basis is not extracted determines that the weight of each index feature includes：
After every training obtains a decision tree, determine to train the Y index extracted when the decision tree special respectively
The importance measures value of sign；
Wherein, the method for determination of the importance measures value of each index feature i includes：
Q-P training sample not being extracted when obtaining the training decision tree, will remain with respectively described in extract Y refer to
The each training sample of mark feature not being extracted is as assessment sample；
Y index feature in each assessment sample is inputted into the decision tree respectively, obtains prediction result, statistical forecast result
Correctly assessment sample number R；
Noise disturbance is carried out to the index feature i in each assessment sample respectively, and respectively by commenting after each progress noise disturbance
The Y index feature estimated in sample inputs the decision tree, obtains prediction result, statistical forecast result correctly assesses sample number
R’；
Using R and R ' difference as the importance measures value of index feature i；
The importance measures value corresponding to different decision trees of index feature i is averaged, the weight as index feature i.
10. a kind of trade company's attrition prediction device, which is characterized in that including：Pretreatment unit and predicting unit；
The pretreatment unit has been lost in the index feature of different time of the trade company before loss and online for obtaining respectively
The current index feature of trade company, and training sample is generated according to the information got；Train to obtain loss in advance according to training sample
Survey model；The attrition prediction model is sent to the predicting unit；
Whether the predicting unit obtains to flow for the index feature of trade company to be predicted to be inputted the attrition prediction model
Lose and will how long after be lost in prediction result.
11. device according to claim 10, which is characterized in that
The pretreatment unit includes：Sample acquisition subelement and model training subelement；
The sample acquisition subelement, is used for：
If confirming, the downtime of the last one online single group of any trade company is located at [the first scheduled durations of date-, date- the
Two scheduled durations] time range in, it is determined that the trade company is to be lost in trade company, and be lost in trade company described in obtaining respectively and existed
The index feature of different time before loss；
If it is determined that there was an online single group in any trade company in the time range of [the second scheduled durations of date-, date], and
There is online single group in the time range of [the first scheduled durations of date-, the second scheduled durations of date-], it is determined that the trade company
For online trade company, and obtain the current index feature of the online trade company；
Wherein, the date indicates that current time, first scheduled duration are more than second scheduled duration；
Training sample is generated according to the information got, is sent to the model training subelement；
The model training subelement, for training to obtain attrition prediction model according to training sample, and by the attrition prediction
Model is sent to the predicting unit.
12. according to the devices described in claim 11, which is characterized in that
The sample acquisition subelement obtains the index spy for being lost in specified N number of different time of the trade company before loss respectively
Sign, N are the positive integer more than one；
The each training sample generated includes：Sample label and index feature；
The sample label includes：It is not lost in, will be lost in after [when loss m- index feature corresponding time] duration, stream
Lose the downtime that the time is the last one online single group.
13. device according to claim 10, which is characterized in that
The index feature includes：Flowing water class index feature, operation class index feature, trade company's Attribute class index feature, basis belong to
Property class index feature and competing to class index feature.
14. according to the devices described in claim 11, which is characterized in that
The model training subelement trains to obtain M decision tree according to training sample, and M is the positive integer more than one；
The index feature of the trade company to be predicted is inputted every decision tree by the predicting unit respectively, obtains every decision tree
Prediction result, and final required prediction result is determined by ballot mode.
15. device according to claim 14, which is characterized in that
The model training subelement extracts P training sample from training sample, and P is less than Q, and Q indicates the total of training sample
Number, trains to obtain a decision tree according to the P training sample extracted.
16. device according to claim 15, which is characterized in that
Include X index feature in each training sample；
The model training subelement is further used for,
When training obtains every decision tree, Y index feature is extracted from X index feature, Y is less than or equal to X, for
The each training sample extracted retains the Y index feature therein respectively, is used for the training of decision tree.
17. device according to claim 16, which is characterized in that
The model training subelement is further used for,
The weight of each index feature is determined according to the training sample not being extracted；
The index feature that weight is more than to predetermined threshold is determined as influencing the principal element that trade company is lost in.
18. device according to claim 17, which is characterized in that
The model training subelement is determined to train decision tree when institute respectively after every training obtains a decision tree
The importance measures value of the Y index feature extracted；
Wherein, the method for determination of the importance measures value of each index feature i includes：
Q-P training sample not being extracted when obtaining the training decision tree, will remain with respectively described in extract Y refer to
The each training sample of mark feature not being extracted is as assessment sample；
Y index feature in each assessment sample is inputted into the decision tree respectively, obtains prediction result, statistical forecast result
Correctly assessment sample number R；
Noise disturbance is carried out to the index feature i in each assessment sample respectively, and respectively by commenting after each progress noise disturbance
The Y index feature estimated in sample inputs the decision tree, obtains prediction result, statistical forecast result correctly assesses sample number
R’；
Using R and R ' difference as the importance measures value of index feature i；
The importance measures value corresponding to different decision trees of index feature i is averaged, the weight as index feature i.
19. a kind of computer equipment, including memory, processor and it is stored on the memory and can be on the processor
The computer program of operation, which is characterized in that the processor is realized when executing described program as any in claim 1~9
Method described in.
20. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that described program is handled
Such as method according to any one of claims 1 to 9 is realized when device executes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710102115.5A CN108510096A (en) | 2017-02-24 | 2017-02-24 | Trade company's attrition prediction method, apparatus, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710102115.5A CN108510096A (en) | 2017-02-24 | 2017-02-24 | Trade company's attrition prediction method, apparatus, equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108510096A true CN108510096A (en) | 2018-09-07 |
Family
ID=63373706
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710102115.5A Pending CN108510096A (en) | 2017-02-24 | 2017-02-24 | Trade company's attrition prediction method, apparatus, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108510096A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109559172A (en) * | 2018-11-29 | 2019-04-02 | 北京车联天下信息技术有限公司 | Data processing method, device, electronic equipment and computer readable storage medium |
| CN110009154A (en) * | 2019-04-03 | 2019-07-12 | 深圳威狮物流网络科技有限公司 | A kind of reimbursement prediction technique, device, terminal device and storage medium |
| CN111629216A (en) * | 2020-04-20 | 2020-09-04 | 南京邮电大学 | VOD service cache replacement method based on random forest algorithm under edge network environment |
| CN111724185A (en) * | 2019-03-21 | 2020-09-29 | 北京沃东天骏信息技术有限公司 | User maintenance method and device |
| CN111738610A (en) * | 2020-06-29 | 2020-10-02 | 金电联行（北京）信息技术有限公司 | Public opinion data-based enterprise loss risk early warning system and method |
| CN111861588A (en) * | 2020-08-06 | 2020-10-30 | 网易（杭州）网络有限公司 | Training method of loss prediction model, player loss reason analysis method and player loss reason analysis device |
| CN112101397A (en) * | 2019-06-18 | 2020-12-18 | 北京京东振世信息技术有限公司 | Method and device for predicting book weight interval |
| CN112686718A (en) * | 2021-03-19 | 2021-04-20 | 深圳索信达数据技术有限公司 | Method and device for acquiring user loss reason, computer equipment and storage medium |
| CN113223224A (en) * | 2021-05-12 | 2021-08-06 | 支付宝(杭州)信息技术有限公司 | Traffic processing method and device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104486350A (en) * | 2014-12-24 | 2015-04-01 | 电子科技大学 | Network content acceleration method based on user behavior |
| CN105279691A (en) * | 2014-07-25 | 2016-01-27 | 中国银联股份有限公司 | Financial transaction detection method and equipment based on random forest model |
| CN106203679A (en) * | 2016-06-27 | 2016-12-07 | 武汉斗鱼网络科技有限公司 | A kind of customer loss Forecasting Methodology and system |
-
2017
- 2017-02-24 CN CN201710102115.5A patent/CN108510096A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105279691A (en) * | 2014-07-25 | 2016-01-27 | 中国银联股份有限公司 | Financial transaction detection method and equipment based on random forest model |
| CN104486350A (en) * | 2014-12-24 | 2015-04-01 | 电子科技大学 | Network content acceleration method based on user behavior |
| CN106203679A (en) * | 2016-06-27 | 2016-12-07 | 武汉斗鱼网络科技有限公司 | A kind of customer loss Forecasting Methodology and system |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109559172A (en) * | 2018-11-29 | 2019-04-02 | 北京车联天下信息技术有限公司 | Data processing method, device, electronic equipment and computer readable storage medium |
| CN111724185A (en) * | 2019-03-21 | 2020-09-29 | 北京沃东天骏信息技术有限公司 | User maintenance method and device |
| CN110009154A (en) * | 2019-04-03 | 2019-07-12 | 深圳威狮物流网络科技有限公司 | A kind of reimbursement prediction technique, device, terminal device and storage medium |
| CN112101397A (en) * | 2019-06-18 | 2020-12-18 | 北京京东振世信息技术有限公司 | Method and device for predicting book weight interval |
| CN111629216A (en) * | 2020-04-20 | 2020-09-04 | 南京邮电大学 | VOD service cache replacement method based on random forest algorithm under edge network environment |
| CN111629216B (en) * | 2020-04-20 | 2021-04-06 | 南京邮电大学 | VOD service cache replacement method based on random forest algorithm under edge network environment |
| CN111738610A (en) * | 2020-06-29 | 2020-10-02 | 金电联行（北京）信息技术有限公司 | Public opinion data-based enterprise loss risk early warning system and method |
| CN111861588A (en) * | 2020-08-06 | 2020-10-30 | 网易（杭州）网络有限公司 | Training method of loss prediction model, player loss reason analysis method and player loss reason analysis device |
| CN111861588B (en) * | 2020-08-06 | 2023-10-31 | 网易（杭州）网络有限公司 | Training method of loss prediction model, player loss reason analysis method and player loss reason analysis device |
| CN112686718A (en) * | 2021-03-19 | 2021-04-20 | 深圳索信达数据技术有限公司 | Method and device for acquiring user loss reason, computer equipment and storage medium |
| CN112686718B (en) * | 2021-03-19 | 2021-06-29 | 深圳索信达数据技术有限公司 | Method and device for acquiring user loss reason, computer equipment and storage medium |
| CN113223224A (en) * | 2021-05-12 | 2021-08-06 | 支付宝(杭州)信息技术有限公司 | Traffic processing method and device |
| CN113223224B (en) * | 2021-05-12 | 2023-10-24 | 支付宝(杭州)信息技术有限公司 | Traffic processing method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108510096A (en) | Trade company's attrition prediction method, apparatus, equipment and storage medium | |
| CN104915879B (en) | The method and device that social relationships based on finance data are excavated | |
| CN107194419A (en) | Video classification methods and device, computer equipment and computer-readable recording medium | |
| US10025980B2 (en) | Assisting people with understanding charts | |
| WO2022116420A1 (en) | Speech event detection method and apparatus, electronic device, and computer storage medium | |
| CN107807941B (en) | Information processing method and device | |
| CN109189666A (en) | Interface test method, device and computer equipment | |
| CN110020938A (en) | Exchange information processing method, device, equipment and storage medium | |
| CN105023165A (en) | Method, device and system for controlling release tasks in social networking platform | |
| US10459934B2 (en) | Re-sizing data partitions for ensemble models in a mapreduce framework | |
| CN107291672A (en) | The treating method and apparatus of tables of data | |
| CN103905532A (en) | Microblog marketing account recognition method and system | |
| CN110633991A (en) | Risk identification method and device and electronic equipment | |
| WO2017062194A1 (en) | Method and system for providing a housing recommendation | |
| AU2021218244A1 (en) | Machine learning modeling for protection against online disclosure of sensitive data | |
| CN112037038A (en) | Bank credit risk prediction method and device | |
| CN110298508A (en) | Behavior prediction method, device and equipment | |
| CN110913354A (en) | Short message classification method and device and electronic equipment | |
| CN108197795A (en) | The account recognition methods of malice group, device, terminal and storage medium | |
| US11257222B2 (en) | Iterative approach for weakly-supervised action localization | |
| CN109711849A (en) | Ether house address portrait generation method and device, electronic equipment and storage medium | |
| CN107767155A (en) | A kind of method and system for assessing user's representation data | |
| CN109657148A (en) | For abnormal operation recognition methods, device, server and the medium for reporting POI | |
| JP7288062B2 (en) | Methods and devices for outputting information, electronic devices, storage media, and computer programs | |
| CN105528618A (en) | Short image text identification method and device based on social network |
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 |