CN108122116A - A kind of monitoring and managing method and system of product promotion channel - Google Patents
A kind of monitoring and managing method and system of product promotion channel Download PDFInfo
- Publication number
- CN108122116A CN108122116A CN201611070918.9A CN201611070918A CN108122116A CN 108122116 A CN108122116 A CN 108122116A CN 201611070918 A CN201611070918 A CN 201611070918A CN 108122116 A CN108122116 A CN 108122116A
- Authority
- CN
- China
- Prior art keywords
- period
- channel
- sub
- detected
- dimension
- 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
- G06Q30/00—Commerce
- G06Q30/018—Certifying business or products
- G06Q30/0185—Product, service or business identity fraud
Abstract
The invention discloses a kind of monitoring and managing methods and system of product promotion channel.Wherein, method includes：Channel benchmark portrait is established according to the data for each channel collected in the first set period of time；The portrait of the channel to be detected is established according to the data for the channel to be detected collected in the second set period of time；The portrait of the channel to be detected is compared with channel benchmark portrait, and when comparison result instruction diversity factor is more than the first threshold of setting or similarity is less than the second threshold of setting, it is problem channel to determine the channel to be detected.Technical solution in the embodiment of the present invention, is capable of the examination rate of Upgrade Problem channel, and reduces alarm probability by mistake.
Description
Technical field
The present invention relates to Internet technical fields, the particularly a kind of monitoring and managing method and system of product promotion channel.
Background technology
In product promotion, such as in the popularization of cell phone software, in order to promote software installation rate, and then the use of software is promoted
Family is touched up to rate, software developer, the agreement as built in cell phone software developer would generally sign software with each software distributor, so
Software distributor can be installed the cell phone software specified in agreement by modes such as brush machines after the mobile phone newly to dispatch from the factory is taken
Into mobile phone.After user takes mobile phone, start software, software can in the case where network is unimpeded, with the background program of software into
Row communication.
The communication date first time on software and backstage is recorded as the newly-increased date from the background, i.e., the software a user's is new
Increase the date.In addition, also have the definition such as activation date, but these definition can be customized by each cell phone software.For example, than
The more general activation date defines method and is：Second of date started of software in 30 days after the newly-increased date.
Payment rule of the general provision between software developer and software distributor in agreement built in software, such as according to
The activation amount of software is paid.But in practice, it has been found that can there are some software distributors oneself that software is started, with
The payment rule for meeting software is allowed to, then obtains the fraud of interests.
If software distributor is defined as a channel of software popularization, software developer is directed to the fraud row of channel
It is channel supervision for the behavior for being identified and feeding back.Two kinds of channel monitoring and managing methods are explained below.
The first, investigate newly-increased date of the same user of software, the activation date with it is follow-up retain use date when
Between be spaced.
This method assumes that the normal behaviour of user is newly-increased date, activation date and the time interval for retaining use date
It is very short, and channel pretends to be user behavior that can not then embody this rule.Because channel obtains payment in modelling customer behavior
Afterwards, the software would not be reused, causes the real use date of subsequent user and newly-increased and activation date intervals very big.
Problem channel can so be screened by the backstage statistics of software.
Second, investigate international mobile subscriber identity (IMSI, International Mobile Subscriber
Identity) number and international mobile equipment identification code (IMEI, International Mobile Equipment
Identity) the ratio of number.
This method assumes that single SIM card mobile phone is in the majority on the market at present, under normal circumstances the IMSI of each single-card mobile phone and
IMEI is one-to-one, i.e., IMSI the and IMEI numbers that software reports should be close to 1：1.But if if being channel simulation,
A kind of most common operation is to activate multiple mobile phones using a SIM card, i.e. an IMSI corresponds to the situation of multiple IMEI.For
Such case channel supervision, software can be made to report the IMSI of SIM card and the IMEI of mobile phone, from the background in one period this two
A value carries out number statistics it can be found that problem channel.
This channel for reporting to be modeled user behavior using software is supervised, and can be unified for and be dug based on data
The channel of pick and analytical technology is drawn a portrait.As can be seen that above two channel regulation technique can to a certain extent to channel into
Row portrait, and problem channel is screened, but both technical solutions all respectively have deficiency.Under the first scheme, if channel is in software
Continue then be considered as normal users behavior using software with normal users behavior after activation；Second scheme can not then cover
Increasingly increased multiple SIM card mobile telephone in the market, and therefore occur alerting by mistake.
The content of the invention
In view of this, a kind of monitoring and managing method of product promotion channel is on the one hand provided in the embodiment of the present invention, on the other hand
A kind of supervisory systems of product promotion channel is provided, for the examination rate of Upgrade Problem channel, and reduces alarm probability by mistake.
A kind of monitoring and managing method of the product promotion channel provided in the embodiment of the present invention, including：
Channel benchmark portrait is established according to the data for each channel collected in the first set period of time；
The portrait of the channel to be detected is established according to the data for the channel to be detected collected in the second set period of time；
The portrait of the channel to be detected with channel benchmark portrait is compared, and difference is indicated in comparison result
When degree is less than the second threshold of setting more than the first threshold or similarity of setting, it is problem canal to determine the channel to be detected
Road.
A kind of supervisory systems of the product promotion channel provided in the embodiment of the present invention, including：
Channel benchmark portrait establishes module, for establishing canal according to the data for each channel collected in the first set period of time
Road benchmark portrait；
Channel portrait to be detected establishes module, for the data according to the channel to be detected collected in the second set period of time
Establish the portrait of the channel to be detected；
Problem channel evaluation module, for the portrait of the channel to be detected and channel benchmark portrait to be compared
It is right, and when comparison result instruction diversity factor is more than the first threshold of setting or similarity is less than the second threshold of setting, really
The fixed channel to be detected is problem channel.
As it can be seen that in the embodiment of the present invention, channel comprehensive modeling is carried out first with the data of each channels, draws out canal
Road benchmark portrait, and then the data of some channel to be detected is recycled to be modeled the channel to be detected, draw out this
Two portraits are compared, calculate departure degree, may determine that according to the departure degree by the portrait of channel to be detected afterwards
The health degree of channel, and then it is supervised.As it can be seen that by initially setting up channel collective model, and then again will be to be detected
The model of channel is compared with channel integration module, can identify the channel that goes wrong, so as to the Zhen of Upgrade Problem channel
Not rate, and reduce alarm probability by mistake.
Description of the drawings
For the clearer technical solution illustrated in the embodiment of the present invention, make required in being described below to embodiment
Attached drawing is briefly described, it should be apparent that, the accompanying drawings in the following description is only some embodiments of the present invention, for
For those of ordinary skill in the art, without creative efforts, it can also be obtained according to these attached drawings other
Attached drawing.Wherein,
Fig. 1 is a kind of exemplary process diagram of the monitoring and managing method of product promotion channel in the embodiment of the present invention.
Fig. 2A is the newly-increased Annual distribution baseline of continuous 24 period in week in the embodiment of the present invention, and Fig. 2 B are implemented for the present invention
Example in a channel to be detected week in continuous 24 period newly-increased time distribution curve.
Benchmark after Fig. 3 A are increased newly for moon junior three day in the embodiment of the present invention retains curve, and Fig. 3 B are in the embodiment of the present invention
The moon junior three day of one channel to be detected increase newly after retention curve.
Fig. 4 is the snapshot of a technical report in the embodiment of the present invention.
Fig. 5 is a kind of structure diagram of the supervisory systems of product promotion channel in the embodiment of the present invention.
Fig. 6 is a kind of structure diagram of the supervisory systems of product promotion channel in another embodiment of the invention.
Fig. 7 is a kind of structure diagram of server in the embodiment of the present invention.
Specific embodiment
In the embodiment of the present invention, it is assumed that the user that most of channels are brought is true normal, their networking
Behavior all meets certain basic law, and the user behavior brought by some problem channels is deviateed then with this deviation by calculating
Degree may determine that the health degree of channel, and then it is supervised.
Based on this, can first with existing data (such as data of last month) to channel there are one whole modeling because silent
Normal channels number is recognized considerably beyond the number of abnormal channel, and the sample of abnormal channel can be almost submerged, therefore so whole
Modeling obtained channel, totally portrait may be considered normal user's portrait.Then to this month, some channel is modeled, such as
The weighted difference of several dimensions of two models of fruit is more than default threshold value, then it is assumed that and the channel is problem channel, otherwise for just
Normal channel.
Fig. 1 is a kind of exemplary process diagram of the monitoring and managing method of software channels in the embodiment of the present invention.Such as Fig. 1 institutes
Show, this method may include following flow.
Step 101, channel benchmark portrait is established according to the data for each channel collected in the first set period of time.
This step 101 is mainly to carry out channel comprehensive modeling using the data of each channel, so as to obtain channel benchmark portrait.
Before channel benchmark portrait is established according to the data gathered, data can be pre-processed as needed：It is clear including data
Wash, data normalization etc., this part can by technical staff according to data characteristic write data processor complete or by
The such as commercial statistical software (SAS, Statistics Analysis System) of ripe software is completed.
In the present embodiment, establish channel benchmark portrait method can there are many.For example, in order to more fully be investigated,
At least two investigation dimensions can be determined according to the data type of channel, then according to identified investigation dimension to set period of time
The data of each channel of interior collection are counted, and obtain what is be made of the benchmark survey data cable (abbreviation baseline) of each investigation dimension
Channel benchmark is drawn a portrait.Wherein, dimension is each investigated to refer to for examining designed by one or more of data data type
Method is examined, which can draw according to ordinary user's behavior and channels violation operation regularity summarization.For example, with mobile phone
Exemplified by software is promoted, the newly-increased, activation that can be reported based on cell phone software the data types such as is retained and determines multiple investigation dimensions, with
Various dimensions investigation is carried out to channel.Wherein, all dimensions substantially use same time period as benchmark.For example, this first
Set period of time can be a few weeks in a calendar month.Specifically, each dimension data can be from first week of calendar month
Start to generate.The beginning of the month, the end of month effect and the weekend effect of data trend variation can so be embodied.Such as user when the beginning of the month
Mobile phone flow it is more, networking behavior it is more frequent；User's unemployed time is more when weekend, and networking behavior is more frequent.
For example, in this step, the first set period of time can be several weeks of the beginning in first week of upper one month.
It is set forth below in the present embodiment for several dimensions determined by cell phone software, it is also can determine in other embodiments
In at least two or other type dimensions, be not limited thereof herein.
Dimension 1：The newly-increased ratio of activation
The ratio between user of channel activation and newly-increased user, for example, the day activation amount of the first sub- period with day new increment it
Than.If the first sub- period was first week all surrounding at the beginning of a calendar month, then this big dimension may be split into one certainly
Right first week moon all surroundings at the beginning, i.e., the dimensions of 28 subdivisions, the activation that each dimension represents daily in 28 days successively are new
Increase ratio.
The activation increase newly than it is represented by curve when be described as follows：
X-axis time span：28 days
Y-axis conversion ratio calculates：The accumulative new increment on the accumulative activation amount/same day on the same day
Specially treated：Because the numerical value in each day is ratio, therefore need not be normalized.
Dimension 2：New increment Annual distribution
Channel according to one day 24 it is small when calculate the quantity that each period increases newly.In order to consider to reduce dimension and all end effects
Should, this dimension can consider first weekly data of calendar month, specifically, the first weekly data can be divided into two parts：
The data of the week are considered as a part, and the data on Sunday Saturday are considered as a part.That is the new increment time point
Cloth can be divided into the new increment Annual distribution of the second sub- period, the new increment Annual distribution of the 3rd sub- period.Specifically retouch
It states as follows：
The week has 24 periods daily, for each period, can using in five days the period it is the sum of newly-increased as
Total newly-increased or new increment, the result so handled of this period can generate 24 period dimensions in all.
Saturday to Sunday generates 24 period dimensions also according to aforesaid way.Can by two days the period increase newly it
With total newly-increased or new increment as this period at weekend.
Being described as follows when the newly-increased Annual distribution is represented by curve：
In week (the week)
X-axis time span：24 periods
Y-axis new increment calculates：5 days, add up with period new increment
Weekend (Saturday to Zhou Tian)
X-axis time span：24 periods
Y-axis new increment calculates：2 days, add up with period new increment
Specially treated：Accumulation new increment can press channel number normalization.For example, new increment is added up to channel day part first
It sorts by size, then normalized formula can be installed：Meter is normalized in (time period value-minimum value)/(maximum-minimum value)
It calculates.Wherein, if the denominator in formula is 0, can use final normalized value is 0.5.
Dimension 3：Activation amount Annual distribution
With dimension 2, simply increasing the time newly is changed to activationary time.It is specific as follows：
Channel according to one day 24 it is small when calculate the quantity of each period activation.In order to consider to reduce dimension and all end effects
Should, this dimension can consider first weekly data of calendar month, specifically, the first weekly data can be divided into two parts：
The data of the week are considered as a part, and the data on Sunday Saturday are considered as a part.That is the activation amount time point
Cloth can be divided into the activation amount Annual distribution of the second sub- period, the activation amount Annual distribution of the 3rd sub- period.Specifically retouch
It states as follows：
The week has 24 periods daily, for each period, can using the sum of activation of the period in five days as
Total activation of this period or activation amount, the result so handled can generate 24 period dimensions in all.
Saturday to Sunday generates 24 period dimensions also according to aforesaid way.Can by the activation of the period in two days it
With total activation as this period at weekend or activation amount.
Being described as follows when the newly-increased Annual distribution is represented by curve：
In week (the week)
X-axis time span：24 periods
Y-axis new increment calculates：5 days, add up with period activation amount
Weekend (Saturday to Zhou Tian)
X-axis time span：24 periods
Y-axis new increment calculates：2 days, add up with period activation amount
Specially treated：Accumulation activation amount can press channel number normalization.Specific normalized can be the same as above-mentioned new increment
Annual distribution.
Dimension 4：User's retention ratio
On the day of channel user it is newly-increased after, behind several days retention ratio, i.e. the 4th sub- period user's retention ratio.Example
Such as, this dimension can consider the retention situation of 14 days after daily increase newly, wherein first two weeks that calendar month can be used use as newly-increased
The investigation point at family, the big dimension of this sample may be split into 14*14 dimension altogether.
Being described as follows when the user's retention ratio is represented by curve：
The newly-increased and its retention of 14 days afterwards of Monday first week this month.
X-axis time span：14 days
Y-axis retention ratio calculates：Certain day after whole IMEI duplicate removals that the channel number is reported, belongs to newly-increased IMEI collection
Number divided by new increment are the retention ratio on the same day.
Specially treated：For ratio without normalization, newly-increased collection needs to filter that avoid denominator be 0 problem in advance for 0 situation.
Above-mentioned each statistics line for investigating dimension forms channel benchmark portrait.Wherein, the above-mentioned first sub- period,
When the length range of each sub- period is respectively positioned on the first setting in two sub- periods, the 3rd sub- period and the 4th sub- period
Between in section, and there is overlapping between each other or there is no overlappings.
Step 102, the channel to be detected is established according to the data for the channel to be detected collected in the second set period of time
Portrait.
Portrait method for building up in this step and the portrait method for building up in step 101 are consistent, the difference is that
Used data area is different.It is to be generated using the data of whole channels or most of channel as benchmark in step 101
Each statistics line for investigating dimension is to generate to treat for this only with the data of a certain channel to be detected in this step 102
Detect the statistics line of each investigation dimension of channel.In the present embodiment, the second set period of time can be and the first setting
Identical period period, such as be all several weeks started first week last month；Or or with the first setting time
The period of Duan Butong, for example, the second set period of time can be several weeks started first week this month.Either other times
Section.But the beginning and ending time feature and time duration of the two should be consistent.For example, if the first set period of time is
Since first week of a calendar month, then the second set period of time also should be since first week of a calendar month；Similarly,
If the first set period of time is terminated in the 4th week of a calendar month, the second set period of time also should be in a nature
The 4th week of the moon terminates.
For example, for a variety of situations for investigating dimension are determined in step 101, can be determined in this step according in step 101
The data of to be detected channel of the investigation dimension to being collected in the second set period of time count, obtain by each investigation dimension
The portrait for the channel to be detected that statistics line to be detected is formed.
Step 103, the portrait of the channel to be detected is compared with channel benchmark portrait, and in comparison result
When indicating that similarity is more than second threshold less than first threshold or diversity factor, it is problem channel to determine the channel to be detected.
In this step, when the portrait of the channel to be detected is compared with channel benchmark portrait, it can obtain
The similarity of the two, and in first threshold of the similarity less than setting, the channel to be detected is determined as problem channel, otherwise
For normal channels；Alternatively, can also obtain the diversity factor of the two, and when diversity factor is more than the second threshold of setting, determine described
Channel to be detected is problem channel, is otherwise normal channels.
For example, in the present embodiment, established for the statistics line of each investigation dimension is utilized in step 101 and step 102
The situation of channel portrait, can be directed to it is each investigate dimension, will corresponding statistics line to be detected and benchmark survey data cable into
Row curve similarity calculates, and obtains the corresponding difference angle value for investigating dimension；According to predetermined each investigation dimension pair
The weighted value answered is weighted the corresponding difference angle value of each investigation dimension, obtains comprehensive differences value, poor in the synthesis
When different value is more than the first threshold of setting, it is problem channel to determine the channel to be detected.Specifically it may include：
For dimension 1：The activation of continuous 28 days/newly-increased data
28 day datas are considered as time-serial position, calculate channel curve to be detected (statistics line i.e. to be detected) and base
Dynamic time warping (DTW, Dynamic may be employed in the curve diversity factor of line (benchmark survey data cable), diversity factor computational methods
Time Warping), DTW is a kind of method for the similarity for weighing the identical or different time series of two length.So may be used
To embody the time attribute implied in the dimension.
For dimension 2：It is newly-increased with continuous 24 period at weekend in all
Value after 24 periods were normalized is considered as time-serial position, calculates the newly-increased time distribution curve of channel to be detected
With the diversity factor of newly-increased Annual distribution baseline, Euclidean distance may be employed in diversity factor computational methods.
Wherein, shown in the formula of Euclidean distance such as following formula (1)：
For example, Fig. 2A shows the newly-increased Annual distribution baseline of continuous 24 period in week, Fig. 2 B show a canal to be detected
The newly-increased time distribution curve of continuous 24 period in the week in road.By calculating the Euclidean distance of two curves, the difference found
Different degree is larger, it can be considered that in the investigation dimension, which does not conform to rule, problematic channel suspicion.
For dimension 3：The activation of all interior and weekend continuous 24 periods
Value after 24 periods were normalized is considered as time-serial position, calculates the activationary time distribution curve of channel to be detected
With the diversity factor of activationary time distribution baseline, Euclidean distance equally may be employed in diversity factor computational methods.
For dimension 4：The retention ratio of 14 days after increasing newly for continuous 14 days
Continuous 14 days it is newly-increased after retention, the data being so unfolded have 14*14=196 small dimensions, can be formed one it is huge
Big wide table, is not suitable for carrying out curve fitting, it is necessary to carry out appropriate dimension-reduction treatment.The dimension-reduction treatment scheme that this programme uses
For：Merge the retention ratio curve after first three day (i.e. preceding several days of the 4th sub- period) of the beginning of the month increases newly, merge after the middle of the month three days
Retention ratio curve after (i.e. rear several days of the 4th sub- period) are newly-increased, calculates the diversity factor of this two curves and baseline, poor
Different degree computational methods are DTW.Wherein, three days are trial value, could alternatively be other number of days.Therefore this 196 dimensions, are dropped
Dimension processing is 2 dimensions.
For example, Fig. 3 A show moon junior three day increase newly after benchmark retain curve, Fig. 3 B show a channel to be detected
Month junior three day increase newly after retention curve.Using DTW algorithms carry out similarity calculation after, find difference very greatly to get to
Diversity factor it is larger, it is believed that the channel in the investigation dimension, operation do not conform to rule.
In the present embodiment, by experimental verification, excessively high weights should not be endowed by finding the diversity factor of dimension 1, because real
Verify the unfixed trend of performance of bright continuous 28 days, therefore the dimension is only used for lower limit monitoring.I.e. the ratio is less than certain
After a default threshold value, just need to draw attention.Therefore 28 dimensions of the group by the way of average value, can be reduced to
One dimension.In addition, it is demonstrated experimentally that the diversity factor of dimension 2 and dimension 3 can be endowed higher weights, and the channel of dimension 4 is drawn
The degree of fitting highest of picture, therefore its diversity factor should be given highest weights.
In the present embodiment, it is assumed that the first weighted value W1 is previously determined for dimension 1, in the week of dimension 2 and dimension 3
Diversity factor and weekend diversity factor are previously determined the second identical weighted value W2, and the 3rd weighted value is previously determined for dimension 4
W3.Wherein, the first weighted value is less than the second weighted value, and the second weighted value is less than the 3rd weighted value.
Assuming that the difference angle value being calculated for dimension 1 is known as the first difference angle value D1, by what is obtained for dimension 2
Difference angle value and weekend difference angle value are referred to as the second difference angle value D2 and the 3rd difference angle value D3 in week, will be directed to dimension 3
Difference angle value and weekend difference angle value are referred to as the 4th difference angle value D4 and the 5th difference angle value D5 in obtained week, will be directed to
The beginning of the month difference angle value and middle of the month difference angle value that dimension 4 obtains are referred to as the 6th difference angle value D6 and the 7th difference angle value D7.
Then in this step, comprehensive differences angle value SD can be represented such as following formula (2)：
SD=W1 × D1+W2 × (D2+D3)+W2 × (D4+D5)+W3 × (D6+D7) (2)
As an example, wherein, the first weighted value W1 can be 0.05, and the second weighted value W2 can be 0.1, and the 3rd adds
Weights W3 can be 0.275.Here each weighted value for investigating dimension can draw by experiment, and using linear regression algorithm into
Row verification, when verification result and experimental result are basically identical, it is believed that be that effective channel investigates calculation formula.
Further, in the present embodiment, after obtaining assessment result in step 103, a skill can also further be generated
Art report is supplied to corresponding personnel, such as commercial affairs contact person.
Fig. 4 shows the snapshot of a technical report.As shown in figure 4, the technical report is divided into three parts, first
Be divided into conclusion, directly give and the larger channel number of gap is thought by channel portrait comparison, second portion for modeling process (i.e.
Establish channel portrait process) data statistics, Part III is assessment data statistics, what to carrying out portrait mathematic interpolation when used
Data statistics.In other embodiments, which can also only include a part therein or two parts as needed,
Or further comprise other contents.It is not limited thereof herein.
Fig. 5 is a kind of structure diagram of the supervisory systems of product promotion channel in the embodiment of the present invention.As shown in figure 5,
The system may include：Channel benchmark portrait establishes module 510, channel to be detected portrait establishes module 520 and the assessment of problem channel
Module 530.
Wherein, channel benchmark portrait establishes module 510 for the number according to each channel collected in the first set period of time
It draws a portrait according to channel benchmark is established.
Channel portrait to be detected establishes module 520 for the number according to the channel to be detected collected in the second set period of time
According to the portrait for establishing the channel to be detected.Wherein, the beginning and ending time feature of the second set period of time and the first set period of time
It is consistent with duration.
Problem channel evaluation module 530 is used to compare in the portrait of the channel to be detected and channel benchmark portrait
It is right, and when comparison result instruction diversity factor is more than the first threshold of setting or similarity is less than the second threshold of setting, really
The fixed channel to be detected is problem channel.
Further, technical report generation module 540 is can further include in the present embodiment, for according to problem canal
The assessment result of road evaluation module 530 generates a technical report.The technical report can as shown in the figure, including three parts,
A part is conclusion, directly gives and thinks the larger channel number of gap by channel portrait comparison, and second portion is to model
The data statistics of journey (establish channel portrait process), Part III are assessment data statistics, during to carrying out portrait mathematic interpolation
The data statistics used.In other embodiments, which can also only include a part therein or two as needed
Part further comprises other contents.It is not limited thereof herein.
Fig. 6 is a kind of structure diagram of the supervisory systems of product promotion channel in another embodiment of the invention.Such as Fig. 6
Shown, which further comprises dimension determining module 550, for basis in addition to including the modules shown in Fig. 5
The data type of channel determines at least two investigation dimensions.It is investigated for example, dimension determining module 550 may be determined as follows in dimension
At least two：The ratio between the day activation amount of first sub- period and day new increment；The new increment Annual distribution of second sub- period,
The new increment Annual distribution of three sub- periods；The activation amount Annual distribution of second sub- period, the activation amount of the 3rd sub- period
Annual distribution；With user's retention ratio of the 4th sub- period.Wherein, the described first sub- period, the second sub- period, the 3rd son
The length range of each sub- period is respectively positioned on first set period of time and second and sets in period and the 4th sub- period
It fixes time in section, and presence is overlapped between each other or there is no overlappings.As an example, the first sub- period can be one
28 days of first week all surrounding at the beginning of calendar month；Second sub- period can be the calendar month Monday of first week a to week
Five 24 periods；3rd sub- period can be 24 periods on the calendar month Saturday of first week a to Sunday；4th sub- time
First week Monday of Duan Weiyi calendar month occurs 14 days started after increasing newly.
Correspondingly, channel benchmark portrait is established module 510 and can be set according to described at least two investigation dimensions to described first
The data for each channel collected in section of fixing time are counted, and obtain the canal being made of the benchmark survey data cable of each investigation dimension
Road benchmark portrait；When channel to be detected portrait is established module 520 and can be set according to the definite investigation dimension to described second
Between the data of channel to be detected collected in section counted, obtain what is be made of the statistics line to be detected of each investigation dimension
The portrait of channel to be detected.Problem channel evaluation module 530 may include：Diversity factor computing module 531, weighted calculation module 532
With result judgement module 533.
Wherein, diversity factor computing module 531 is used to investigate dimension for each, will corresponding statistics line to be detected and
Benchmark survey data cable carries out curve similarity calculation, obtains the corresponding difference angle value for investigating dimension.
Weighted calculation module 532 is used to, according to the corresponding weighted value of predetermined each investigation dimension, tie up each investigate
It spends corresponding difference angle value to be weighted, obtains comprehensive differences value.
Result judgement module 533 is used to, when the comprehensive differences value is more than the first threshold of setting, determine described to be checked
Survey channel is problem channel.
Wherein, diversity factor computing module 531 may include following moulds in the block at least two：
First diversity factor computational submodule 5311, for being directed to the day activation amount of the described first sub- period and day new increment
The ratio between, using the diversity factor of dynamic time warping calculating statistics line to be detected and benchmark survey data cable；
Second diversity factor calculate submodule 5312, for be directed to the described second sub- period new increment Annual distribution, the 3rd
When the new increment Annual distribution of sub- period, the activation amount Annual distribution of the second sub- period, the activation amount of the 3rd sub- period
Between be distributed in any one, the sub- period is normalized first, then using Euclidean distance calculate normalize
The diversity factor of treated statistics line to be detected and benchmark survey data cable；
3rd diversity factor computational submodule 5313 is right first for being directed to user's retention ratio of the 4th sub- period
User's retention ratio of the 4th sub- period carries out dimension-reduction treatment, after then calculating dimension-reduction treatment using dynamic time warping
The diversity factor of statistics line to be detected and benchmark survey data cable.
Wherein, weighted calculation module 532 can be directed to the ratio between day activation amount and day new increment of the described first sub- period, adopt
With predetermined first weighted value；New increment Annual distribution, the 3rd sub- period for the described second sub- period it is new
Appointing in Delta Time distribution, the activation amount Annual distribution of the second sub- period, the activation amount Annual distribution of the 3rd sub- period
Predetermined second weighted value of one use；For user's retention ratio of the 4th sub- period, use is predetermined
3rd weighted value；Comprehensive differences value is obtained after being weighted.Wherein, the first weighted value is less than the second weighted value, and second adds
Weights are less than the 3rd weighted value.
Fig. 7 is a kind of structure diagram of server in the embodiment of the present invention.The server can be used for realizing side shown in Fig. 1
Method and Fig. 5 and Fig. 6 shown devices.As shown in fig. 7, the server may include：Processor 701, non-volatile computer are readable
Memory 702, display unit 703 and network communication interface 704.These components are communicated by bus 705.
In the present embodiment, multiple program modules are stored in memory 702, such as：Operating system 706, network communication mould
Block 707 and application program 708.
Processor 701 can read the various module (not shown)s included by the application program 708 in memory 702
Carry out various function application and the data processing of execute server.Processor 701 in the present embodiment can be one, can also
To be multiple, can be CPU, processing unit/module, ASIC, logic module or programmable gate array etc..
Wherein, operating system 706 includes but not limited to：Android operation system, Windows operating system, apple iOS
Operating system, apple Mac OS operating systems etc..
Application program 708 may include each function module in Fig. 5 or Fig. 6 shown devices, and form corresponding computer
Executable instruction set 709 and corresponding metadata and heuritic approach 710.These set of computer-executable instructions can be by described
Processor 701 performs and completes method shown in Fig. 1 or the function of Fig. 5 or Fig. 6 shown devices.
In the present embodiment, network communication interface 704 is engaged with network communication module 707 completes the various networks of server
The transmitting-receiving of signal, transmission and network data exchange including the data between each channels etc..
Display unit 703 has a display panel, for completing the input of relevant information and display, including technology report is presented
Accuse etc..
The present invention also provides a kind of storage medium, such as non-volatile computer readable storage medium storing program for executing, wherein being stored with number
According to processing routine, which is used to perform any embodiment of the above method of the present invention.
As it can be seen that in the embodiment of the present invention, channel comprehensive modeling is carried out first with the data of each channels, draws out canal
Road benchmark portrait, and then the data of some channel to be detected is recycled to be modeled the channel to be detected, draw out this
Two portraits are compared, calculate departure degree, may determine that according to the departure degree by the portrait of channel to be detected afterwards
The health degree of channel, and then it is supervised.
Further, data mining and the similarity of curves measurement technology in analysis have been used, channels have been reported
The data such as newly-increased, activation, retention carry out various dimensions investigation, and correlation is small between each dimension, so as to right in all directions
Channel carries out relief painting picture；It is more difficult to cause erroneous judgement or omission compared to considering for single dimension.By establishing comparison model
Data to be assessed are analyzed, the user behavior portrait of the cell phone software of certain channels is finally drawn, and draws a portrait with benchmark
It is compared, achievees the purpose that supervise channels.It realizes comprehensive, full automatic channel and considers system, save
Manpower and time cost；And the examination rate of problem channel is improved, reduce alarm probability by mistake.
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 modifications, equivalent replacements and improvements are made should all be included in the protection scope of the present invention god.
Claims (12)
1. a kind of monitoring and managing method of product promotion channel, which is characterized in that including：
Channel benchmark portrait is established according to the data for each channel collected in the first set period of time；
The portrait of the channel to be detected is established according to the data for the channel to be detected collected in the second set period of time；
The portrait of the channel to be detected is compared with channel benchmark portrait, and it is big in comparison result instruction diversity factor
When the first threshold or similarity of setting are less than the second threshold set, it is problem channel to determine the channel to be detected.
It is 2. according to the method described in claim 1, it is characterized in that, described according to each channel collected in the first set period of time
Data establish channel benchmark portrait before, further comprise：At least two investigation dimensions are determined according to the data type of channel；
The data according to each channel collected in the first set period of time, which establish channel benchmark portrait, to be included：According to determining
The data of each channel of the investigation dimension to being collected in first set period of time count, obtain by each investigation dimension
The channel benchmark portrait that benchmark survey data cable is formed；
The data according to the channel to be detected collected in the second set period of time establish the portrait bag of the channel to be detected
It includes：It is united according to the data of to be detected channel of the definite investigation dimension to being collected in second set period of time
Meter obtains the portrait for the channel to be detected being made of the statistics line to be detected of each investigation dimension；
The portrait by the channel to be detected is compared with channel benchmark portrait, and similar in comparison result instruction
When degree is more than the second threshold of setting less than the first threshold or diversity factor of setting, it is problem canal to determine the channel to be detected
Road includes：It is for each investigation dimension, corresponding statistics line to be detected is similar to benchmark survey data cable progress curve
Degree calculates, and obtains the corresponding difference angle value for investigating dimension；According to the corresponding weighted value of predetermined each investigation dimension,
The corresponding difference angle value of each investigation dimension is weighted, comprehensive differences value is obtained, is more than in the comprehensive differences value and sets
During fixed first threshold, it is problem channel to determine the channel to be detected.
3. according to the method described in claim 2, it is characterized in that, described at least two investigate dimension including investigating dimension as follows
In at least two：
The ratio between the day activation amount of first sub- period and day new increment；
New increment Annual distribution, the new increment Annual distribution of the 3rd sub- period of second sub- period；
Activation amount Annual distribution, the activation amount Annual distribution of the 3rd sub- period of second sub- period；
User's retention ratio of 4th sub- period；
Wherein, each sub- time in the described first sub- period, the second sub- period, the 3rd sub- period and the 4th sub- period
The length range of section is respectively positioned in first set period of time and the second set period of time, and between each other in the presence of overlapping or not
There are overlappings.
4. according to the method described in claim 3, it is characterized in that, the first sub- period is first week week an of calendar month
28 days of surrounding at the beginning；
The second sub- period is 24 periods of the calendar month Mon-Fri of first week；
The 3rd sub- period is 24 periods on the calendar month Saturday of first week a to Sunday；
The 4th sub- period is to occur first week Monday an of calendar month 14 days started after increasing newly.
5. the method according to claim 3 or 4, which is characterized in that it is described to investigate dimension for each, it will be corresponding to be checked
Survey statistics line and benchmark survey data cable carry out curve similarity calculation include it is following at least two；
The ratio between day activation amount and day new increment for the described first sub- period calculate system to be detected using dynamic time warping
Count the diversity factor of line and benchmark survey data cable；
For the new increment Annual distribution of the described second sub- period, the new increment Annual distribution of the 3rd sub- period, the second son
Any one in the activation amount Annual distribution of period, the activation amount Annual distribution of the 3rd sub- period, first to the period of the day from 11 p.m. to 1 a.m
Between section be normalized, statistics line to be detected and benchmark after normalized are then calculated using Euclidean distance and is united
Count the diversity factor of line；
For user's retention ratio of the 4th sub- period, user's retention ratio of the 4th sub- period is dropped first
Then dimension processing calculates the statistics line to be detected after dimension-reduction treatment and benchmark survey data cable using dynamic time warping
Diversity factor.
6. the according to the method described in claim 5, it is characterized in that, corresponding weighting of predetermined each investigation dimension
Value includes at least two in following weighted values；
The ratio between the first definite weighted value of day activation amount and day new increment for the described first sub- period；
For the new increment Annual distribution of the described second sub- period, the new increment Annual distribution of the 3rd sub- period, the second son
Any one second definite weighting in the activation amount Annual distribution of period, the activation amount Annual distribution of the 3rd sub- period
Value；
The 3rd weighted value determined for user's retention ratio of the 4th sub- period；
Wherein, first weighted value is less than second weighted value, and second weighted value is less than the 3rd weighted value.
7. according to the method described in claim 6, it is characterized in that, first weighted value be 0.05, second weighted value
For 0.1, the 3rd weighted value is 0.275.
8. a kind of supervisory systems of product promotion channel, which is characterized in that including：
Channel benchmark portrait establishes module, for establishing Canal Base according to the data for each channel collected in the first set period of time
Quasi- portrait；
Channel portrait to be detected establishes module, for being established according to the data for the channel to be detected collected in the second set period of time
The portrait of the channel to be detected；
Problem channel evaluation module, for the portrait of the channel to be detected to be compared with channel benchmark portrait, and
When comparison result instruction diversity factor is more than the first threshold of setting or similarity is less than the second threshold of setting, determine described
Channel to be detected is problem channel.
9. system according to claim 8, which is characterized in that further comprise：Dimension determining module, for according to channel
Data type determine at least two investigation dimensions；
The channel benchmark portrait establishes module and investigates dimension to being received in first set period of time according to described at least two
The data of each channel of collection are counted, and obtain being drawn a portrait by the channel benchmark that the benchmark survey data cable of each investigation dimension is formed；
The channel portrait to be detected establishes module according to the definite investigation dimension to being received in second set period of time
The data of the channel to be detected of collection are counted, and obtain the canal to be detected being made of the statistics line to be detected of each investigation dimension
The portrait in road；
Described problem channel evaluation module includes：
Diversity factor computing module, for being directed to each investigation dimension, by corresponding statistics line to be detected and benchmark survey number
Curve similarity calculation is carried out according to line, obtains the corresponding difference angle value for investigating dimension；
Weighted calculation module, for according to the corresponding weighted value of predetermined each investigation dimension, being corresponded to each investigation dimension
Difference angle value be weighted, obtain comprehensive differences value；With
Result judgement module, for when the comprehensive differences value is more than the first threshold set, determining the channel to be detected
For problem channel.
10. system according to claim 9, which is characterized in that the dimension determining module determines to investigate in dimension as follows
At least two：
The ratio between the day activation amount of first sub- period and day new increment；
New increment Annual distribution, the new increment Annual distribution of the 3rd sub- period of second sub- period；
Activation amount Annual distribution, the activation amount Annual distribution of the 3rd sub- period of second sub- period；
User's retention ratio of 4th sub- period；
Wherein, each sub- time in the described first sub- period, the second sub- period, the 3rd sub- period and the 4th sub- period
The length range of section is respectively positioned in first set period of time and the second set period of time, and between each other in the presence of overlapping or not
There are overlappings.
11. system according to claim 10, which is characterized in that it is in the block that the diversity factor computing module includes following moulds
At least two：
The ratio between first diversity factor computational submodule, day activation amount and day new increment for being directed to for the described first sub- period, is adopted
The diversity factor of statistics line to be detected and benchmark survey data cable is calculated with dynamic time warping；
Second diversity factor computational submodule, for being directed to new increment Annual distribution, the 3rd sub- time of the described second sub- period
New increment Annual distribution, the activation amount Annual distribution of the second sub- period, the activation amount Annual distribution of the 3rd sub- period of section
In any one, the sub- period is normalized first, then using Euclidean distance calculate normalized after
Statistics line to be detected and benchmark survey data cable diversity factor；
3rd diversity factor computational submodule, for being directed to user's retention ratio of the 4th sub- period, first to the described 4th
User's retention ratio of sub- period carries out dimension-reduction treatment, and the system to be detected after dimension-reduction treatment is then calculated using dynamic time warping
Count the diversity factor of line and benchmark survey data cable.
12. system according to claim 11, which is characterized in that the weighted calculation module is used for for the described first son
The ratio between the day activation amount of period and day new increment, using predetermined first weighted value；For the described second sub- period
New increment Annual distribution, the new increment Annual distribution of the 3rd sub- period, the activation amount Annual distribution of the second sub- period,
Any one in the activation amount Annual distribution of three sub- periods uses predetermined second weighted value；For the 4th period of the day from 11 p.m. to 1 a.m
Between section user's retention ratio, using predetermined 3rd weighted value；Comprehensive differences value is obtained after being weighted；
Wherein, first weighted value is less than second weighted value, and second weighted value is less than the 3rd weighted value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611070918.9A CN108122116A (en) | 2016-11-29 | 2016-11-29 | A kind of monitoring and managing method and system of product promotion channel |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611070918.9A CN108122116A (en) | 2016-11-29 | 2016-11-29 | A kind of monitoring and managing method and system of product promotion channel |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108122116A true CN108122116A (en) | 2018-06-05 |
Family
ID=62224296
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611070918.9A Pending CN108122116A (en) | 2016-11-29 | 2016-11-29 | A kind of monitoring and managing method and system of product promotion channel |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108122116A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109325779A (en) * | 2018-08-20 | 2019-02-12 | 北京数美时代科技有限公司 | A kind of read-write portrait method, system and portrait processing system cheated for counter |
| CN110348715A (en) * | 2019-06-28 | 2019-10-18 | 北京淇瑀信息科技有限公司 | Channel method for detecting abnormality, device and electronic equipment based on flow indicator monitoring |
| CN110875856A (en) * | 2018-08-31 | 2020-03-10 | 北京京东尚科信息技术有限公司 | Method and apparatus for activation data anomaly detection and analysis |
| CN111353796A (en) * | 2018-12-20 | 2020-06-30 | 北京搜狗科技发展有限公司 | Quality judgment method and device for flow channel |
| CN111797141A (en) * | 2020-07-06 | 2020-10-20 | 深圳市活力天汇科技股份有限公司 | Method for detecting abnormal flow of air ticket search |
| CN116385080A (en) * | 2023-04-17 | 2023-07-04 | 云洞(上海)科技股份有限公司 | Mobile internet user data statistics popularization system based on artificial intelligence |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8347394B1 (en) * | 2009-07-15 | 2013-01-01 | Trend Micro, Inc. | Detection of downloaded malware using DNS information |
| CN103593465A (en) * | 2013-11-26 | 2014-02-19 | 北京网秦天下科技有限公司 | Method and device for diagnosing abnormality of application popularization channel |
| CN104731914A (en) * | 2015-03-24 | 2015-06-24 | 浪潮集团有限公司 | Method for detecting user abnormal behavior based on behavior similarity |
| CN104915455A (en) * | 2015-07-02 | 2015-09-16 | 焦点科技股份有限公司 | Website exception access identification method and system based on user behaviors |
| CN105208040A (en) * | 2015-10-12 | 2015-12-30 | 北京神州绿盟信息安全科技股份有限公司 | Network attack detection method and device |
| CN105205394A (en) * | 2014-06-12 | 2015-12-30 | 腾讯科技（深圳）有限公司 | Data detection method and device for invasion detection |
| CN105744043A (en) * | 2014-12-08 | 2016-07-06 | 东莞宇龙通信科技有限公司 | Identification application activation condition method and device |
-
2016
- 2016-11-29 CN CN201611070918.9A patent/CN108122116A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8347394B1 (en) * | 2009-07-15 | 2013-01-01 | Trend Micro, Inc. | Detection of downloaded malware using DNS information |
| CN103593465A (en) * | 2013-11-26 | 2014-02-19 | 北京网秦天下科技有限公司 | Method and device for diagnosing abnormality of application popularization channel |
| CN105205394A (en) * | 2014-06-12 | 2015-12-30 | 腾讯科技（深圳）有限公司 | Data detection method and device for invasion detection |
| CN105744043A (en) * | 2014-12-08 | 2016-07-06 | 东莞宇龙通信科技有限公司 | Identification application activation condition method and device |
| CN104731914A (en) * | 2015-03-24 | 2015-06-24 | 浪潮集团有限公司 | Method for detecting user abnormal behavior based on behavior similarity |
| CN104915455A (en) * | 2015-07-02 | 2015-09-16 | 焦点科技股份有限公司 | Website exception access identification method and system based on user behaviors |
| CN105208040A (en) * | 2015-10-12 | 2015-12-30 | 北京神州绿盟信息安全科技股份有限公司 | Network attack detection method and device |
Non-Patent Citations (1)
| Title |
|---|
| 黄志洪等: "《现代计算机信息安全技术 2004年7月第1版》", 30 July 2004, 冶金工业出版社 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109325779A (en) * | 2018-08-20 | 2019-02-12 | 北京数美时代科技有限公司 | A kind of read-write portrait method, system and portrait processing system cheated for counter |
| CN110875856A (en) * | 2018-08-31 | 2020-03-10 | 北京京东尚科信息技术有限公司 | Method and apparatus for activation data anomaly detection and analysis |
| CN111353796A (en) * | 2018-12-20 | 2020-06-30 | 北京搜狗科技发展有限公司 | Quality judgment method and device for flow channel |
| CN111353796B (en) * | 2018-12-20 | 2024-03-26 | 北京搜狗科技发展有限公司 | Quality judgment method and device for flow channel |
| CN110348715A (en) * | 2019-06-28 | 2019-10-18 | 北京淇瑀信息科技有限公司 | Channel method for detecting abnormality, device and electronic equipment based on flow indicator monitoring |
| CN110348715B (en) * | 2019-06-28 | 2024-01-23 | 北京淇瑀信息科技有限公司 | Channel anomaly detection method and device based on flow index monitoring and electronic equipment |
| CN111797141A (en) * | 2020-07-06 | 2020-10-20 | 深圳市活力天汇科技股份有限公司 | Method for detecting abnormal flow of air ticket search |
| CN111797141B (en) * | 2020-07-06 | 2023-09-22 | 深圳市活力天汇科技股份有限公司 | Method for detecting air ticket searching flow abnormality |
| CN116385080A (en) * | 2023-04-17 | 2023-07-04 | 云洞(上海)科技股份有限公司 | Mobile internet user data statistics popularization system based on artificial intelligence |
| CN116385080B (en) * | 2023-04-17 | 2024-01-26 | 云洞(上海)科技股份有限公司 | Mobile internet user data statistics popularization system based on artificial intelligence |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108122116A (en) | A kind of monitoring and managing method and system of product promotion channel | |
| US20150371163A1 (en) | Churn prediction in a broadband network | |
| US10690512B2 (en) | Analysis of smart meter data based on frequency content | |
| CN108182515B (en) | Intelligent rule engine rule output method, equipment and computer readable storage medium | |
| US20170371757A1 (en) | System monitoring method and apparatus | |
| CN110866820A (en) | Real-time monitoring system, method, equipment and storage medium for banking business | |
| CN109918279B (en) | Electronic device, method for identifying abnormal operation of user based on log data and storage medium | |
| CN105721187A (en) | Service fault diagnosis method and apparatus | |
| WO2014126576A2 (en) | Churn prediction in a broadband network | |
| CN109255486A (en) | A kind of optimization method and device of strategy configuration | |
| CN112598513B (en) | Method and device for identifying stockholder risk transaction behaviors | |
| CN106959925A (en) | A kind of version method of testing and device | |
| CN111694718A (en) | Method and device for identifying abnormal behavior of intranet user, computer equipment and readable storage medium | |
| CN106611023A (en) | Detection method and device for website access exception | |
| CN107484134A (en) | resident position analysis method and device | |
| CN107391571A (en) | The processing method and processing device of sensing data | |
| CN107590016A (en) | Power-down rebooting recognition methods and device | |
| CN110807050B (en) | Performance analysis method, device, computer equipment and storage medium | |
| CN105429792A (en) | User behavior flow obtaining method and device and user behavior analysis method and system | |
| CN109559218A (en) | A kind of determination method, apparatus traded extremely and storage medium | |
| CN113342939B (en) | Data quality monitoring method and device and related equipment | |
| CN111768287A (en) | Period identification method, period identification device, server and readable storage medium | |
| CN112765463B (en) | Data management method for big data and user requirements and cloud computing server | |
| CN109586990B (en) | Method and device for identifying cheating flow | |
| CN108491213A (en) | Software continuous integrated approach, software continuous integrating device and electronic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180605 |
|
| RJ01 | Rejection of invention patent application after publication |