CN104408641B - The brand identity extracting method and system of ecommerce recommended models - Google Patents
The brand identity extracting method and system of ecommerce recommended models Download PDFInfo
- Publication number
- CN104408641B CN104408641B CN201410593894.XA CN201410593894A CN104408641B CN 104408641 B CN104408641 B CN 104408641B CN 201410593894 A CN201410593894 A CN 201410593894A CN 104408641 B CN104408641 B CN 104408641B
- Authority
- CN
- China
- Prior art keywords
- brand
- mrow
- time
- msub
- purchase
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Abstract
The present invention relates to a kind of brand identity extracting method of ecommerce recommended models, including：The basic data of brand to carrying out ecommerce sale carries out time slicing, so as to construct the brand identity sequence of different time piece；According to the brand identity sequence of the different time piece of above-mentioned construction, temperature and cost analysis are carried out to the transaction data of brand, extract the feature of brand.The invention further relates to a kind of brand identity extraction system of ecommerce recommended models.The present invention carries out dimension enlarging according to basic user journal information and brand operation information, extracts new characteristic set, builds the brand identity system of recommended models.
Description
Technical field
The present invention relates to a kind of brand identity extracting method of ecommerce recommended models and system.
Background technology
The development of internet and information technology has triggered the dramatic change of method of thinking, life style and business model.
Under global commerce linguistic context, " big data epoch " require electric business brand with mass data processing system to consumer from information search
The tracking and search of behavior after even being bought to product purchase, make for consumer demand and more being determined in real time with what is become more meticulous
Plan.Basic personalized recommendation technology is required for extracting the characteristic information of user and brand from the log information of website behavior,
And by feature selecting, unrelated and redundancy feature is eliminated, can just obtain gratifying recommendation effect.But conventional website day
The details of each request are contained to will information all matters, big and small, data characteristics extraction can be really carried out in the information of redundancy
Only user behavior data, including click on, buy, collecting, the operation information such as shopping cart.
Data characteristics is extracted in data management and machine learning field plays an important role, but existing data are special
Signization refers to the scale for reducing initial data in the case of retention data feature.The industrial background of data characterization is with data
It is extensive to increase, the higher-dimension mass data of implicit mass efficient information is produced, in these high value total amounts, low value density
Data in find valuable knowledge, it is necessary to by data characteristics extract retain complex data in effective information, will at a low price
The information of value density is converted into the information of high value density.
Current feature extracting method can extract on known abundant Back ground Information and obtain required recessive character, or
It is background that person obtains the related professional knowledge of business by professional occurrences in human life.However, in the data mining engineering actually faced, wish
Hope that it is that features described above constructing plan institute is irrealizable to go out high-dimensional and orthogonal data characteristics by most rare information structuring.
The content of the invention
In view of this, it is necessary to which the brand identity extracting method and system of a kind of ecommerce recommended models are provided.
The present invention provides a kind of brand identity extracting method of ecommerce recommended models, and this method comprises the following steps：
The basic data of brand to carrying out ecommerce sale carries out time slicing, so as to construct the brand identity sequence of different time piece
Row；According to the brand identity sequence of the different time piece of above-mentioned construction, temperature and cost analysis are carried out to the transaction data of brand,
Extract the feature of brand.
Wherein, this method also includes：Value revision is carried out to the feature of the brand of said extracted.
Described time slicing includes：Conventional time slicing and the time slicing based on buying behavior, wherein described normal
The time slicing of rule includes：According to natural date burst, according to the daily marketing situation of brand and time penalty factor burst, press
According to the date from closely being referred to by short elongated burst, the time slicing mode based on buying behavior user to brand to being far spaced
Time behavior sequence is to buy the date as cut-off.
The feature of described brand includes：Transformation ratio, the marketing cycle of brand, the temperature of brand, the brand of brand are purchased again
Buy probability.
Described value revision refers to the carry out value revision by log functions.
The present invention also provides a kind of brand identity extraction system of ecommerce recommended models, including time slicing module,
Characteristic extracting module, wherein：The basic data that the time slicing module is used for the brand to carrying out ecommerce sale is carried out
Time slicing, so as to construct the brand identity sequence of different time piece；The characteristic extracting module is used for according to above-mentioned construction
The brand identity sequence of different time piece, temperature and cost analysis are carried out to the transaction data of brand, extract the feature of brand.
Wherein, the system also includes value revision module, and the value revision module is used for the brand of said extracted
Feature carries out value revision.
Described time slicing includes：Conventional time slicing and the time slicing based on buying behavior, wherein described normal
The time slicing of rule includes：According to natural date burst, according to the daily marketing situation of brand and time penalty factor burst, press
According to the date from closely being referred to by short elongated burst, the time slicing mode based on buying behavior user to brand to being far spaced
Time behavior sequence is to buy the date as cut-off.
The feature of described brand includes：Transformation ratio, the marketing cycle of brand, the temperature of brand, the brand of brand are purchased again
Buy probability.
Described value revision refers to the carry out value revision by log functions.
The brand identity extracting method and system of ecommerce recommended models of the present invention, can make e-commerce website in sea
Measure in data basis, carry out dimension enlarging according to basic user journal information and brand operation information, extract new feature set
Close, build the brand identity system of recommended models.The data value that the present invention extracts is high, and extraction effect is good.
Brief description of the drawings
Fig. 1 is the flow chart of the brand identity extracting method of ecommerce recommended models of the present invention；
Fig. 2 is the time slicing schematic diagram based on buying behavior；
Fig. 3 is the hardware architecture diagram of the brand identity extraction system of ecommerce recommended models of the present invention.
Embodiment
Below in conjunction with the accompanying drawings and specific embodiment the present invention is further detailed explanation.
As shown in fig.1, it is the work of the brand identity extracting method preferred embodiment of ecommerce recommended models of the present invention
Industry flow chart.
Step S401, time slicing is carried out to carrying out the basic data of brand of ecommerce sale, constructs different time
The brand identity sequence of piece.It is specific as follows：
The present embodiment illustrates by taking day cat store as an example.In day cat store, all there can be tens million of users to pass through daily
Brand finds the commodity oneself liked, and brand is to connect the most important tie of consumer and commodity.In existing historical record
In, the feature of brand how is extracted, first key feature is that the time series feature of Brand Marketing situation.
Conventional time slicing has three kinds of modes：
(1) split according to the natural date, split by week, first quarter moon, the moon, directly calculate the point in brand different time span
Situations such as hitting, collecting.With four months summary journal, if by can be divided into 16 week, the click of every, purchase, collection, shopping cart
To click in 7 days, purchase, collection, shopping cart sum；If by first quarter moon 8 can be divided into, the click of every, purchase, collection,
Shopping cart is to click on, buy in 14 days, collecting, shopping cart sum；If can monthly be divided into 4, the click of every, purchase,
Collection, shopping cart are to click on, buy in 30 days, collecting, shopping cart sum.
(2) the daily marketing situation of brand is directly multiplied by a time penalty factor K, brand is purchased than previous recently
Purchased even more important before individual month, so time punishment is inversely proportional with the time, the time, nearlyer penalty value was smaller, time more remote punishment
Value is bigger.
Wherein, x is current date and the distance of last day, and w is the decay factor that can be debugged, if selection 4 months for
Time span, w are test optimum coefficient 20.The change of length over time, w can do flexible adjustment.By daily click,
Purchase, collection, shopping cart is multiplied by time penalty factor, obtain tetra- kinds of behavior k of new brand j (click on purchase collection shopping cart)
Population characteristic value.
(3) by the marketing record of brand according to the date from closely to remote, the interval of burst is progressively longer near to protrude brand
The importance of situation is sold and be clicked to phase.By taking the record in April-July as an example, burst is described in detail below shown in table one, July
For newest behavior record, following 22 are segmented into.Date nearest 7 days, 7 are divided into units of day, then two weeks
With 3 days for time interval, it is divided into 5, ensuing one and a half months is divided into 6 in units of week, then that the remaining date is first
3 are extracted by first quarter moon, last moon is divided into individually a piece of according to the moon for unit.
The time slicing table of one user behavior scheme of table three
Different data set time length, different data set sizes, it can be constructed by three kinds of schemes above
The data characteristics of basal latency burst.Wherein the first scheme is partial to factor of influence less situation of the time to model, the
Two kinds of schemes are adapted to several months user behavior span, and the third scheme is applied to cutting for the time series behavior arbitrarily long to brand
Point, using phase fine granularity recently, the mode of remote date group granularity, build the time series characteristics of different brands.
Different time of the act bursts is provided to allow the forecast models such as logistic regression, decision tree, random forest more
Weight anticipation is carried out to the behavioral implications factor of the different time sections of user well.
In addition to the time slicing scheme of routine, the present embodiment also proposed the time slicing side based on buying behavior
Formula.Time slicing mode based on buying behavior refers to user to the time behavior sequence of brand to buy the date as cutting
Point, cutting are different timeslices.The wherein involved timeslice of ith purchase refers to, terminates from the i-th -1 time purchase latter three days
Three days after terminating to ith purchase.
If but there are collection or shopping cart behavior, ith time buying section in three days after the i-th -1 time purchase terminates
Initial time for the i-th -1 time buy terminate latter three days collection or shopping cart date, while by the i-th -1 time purchase to collect
Or shopping cart behavior directly clicks on behavior according to buying and collecting behavior cutting, specifically as shown in Figure 2.
In Fig. 2, the initial purchase date is on July 15th, 2013, if not collecting and purchasing in three days after this date
Thing garage is, i.e., from July 18th, 2013 to the 18 days July in 2013 of latter three days of second of purchase for second of purchase when
Between piece, but due to having collection behavior in after initial purchase three days, then second of time buying piece is on July 17th, 2013
(collection date) clicks on 10 quantity according to purchase on July 18th, 2013, while by the click behavior after initial purchase
B and collection quantity c cuttings, wherein, 10*b/ (b+c) is included into initial purchase, and 10*c/ (b+c) is included into second of time buying
Piece.After time span t, number of clicks, collection and shopping cart number involved by purchase each to brand j user i, you can with
Brand purchased average required time length, number of clicks and collection, shopping cart number every time are obtained by average value.
Time slicing is carried out based on buying behavior with above-mentioned, the action time bought every time is meticulously analyzed, is advantageous to
Faster more accurately build the feature of brand.
Step S402, according to the brand identity sequence of the different time piece of above-mentioned construction, the transaction data of brand is carried out
Temperature and cost analysis, extract the feature of brand.Specifically：
In offline recommending module, the feature and user characteristics of brand are equally crucial.The sales tactics of different brands, by
Ratings, customer purchase turn-head-rate are completely different, it is impossible to these brands lump together, therefore need to construct description product
The attribute and user's proposed algorithm module of board self-characteristic.Feature of the present embodiment from following direction structure brand：
(1) transformation ratio of brand
As the transformation ratio of user, brand also has the transformation ratio of itself, and the record that brand is operated every time is with being purchased
The transformation ratio relation of record can be hit with branch conversion ratio brandClickRate, collection conversion ratio brandFavoriteRate and
Shopping cart conversion ratio brandCartRate.Define brand and be clicked behavior to effective influence power of purchase and user's click behavior
Identical to the influence power of purchase, user's last time buying behavior terminates to terminate within latter three days to local buying behavior for latter 3rd day
All clicks, will buy that to terminate that the rear date delays three days be because it is that user returns mostly that each buying behavior is clicked in three days last time
It is not the effect of new purchasing demand caused by commodity have been purchased in visit.Such as user i to brand j in dkIt had purchaser record,
In dk+1It has n bar purchaser records, then in dk+ 3 to dk+1Caused click sum in+3 periodDivided by dk+1Record number is bought in reaching for it, as the click conversion ratio of the single purchase of brand.
Wherein, rijdbExpression user i is to brand j in the d days record numbers for operating b, and b=0 is clicks on, and b=1 is buys, b
=2 be collection, and b=3 can similarly be obtained, the collection conversion ratio of this purchase of user to add shopping cart：
With shopping cart conversion ratio：
The click, collection, shopping cart conversion ratio of each purchaser record of user are obtained, averages to obtain that to weigh the brand total
The behavior conversion ratio of body, include brand j click conversion ratio brandClickRatej, collect conversion ratio
brandFavoriteRatejWith shopping cart conversion ratio brandCartRatej, formula is as follows.
(2) the marketing cycle of brand
The marketing cycle of brand is used for weighing the Buying Cycle of brand, and different brands may belong to day consumption category, moon consumption product
Class or durable category.How the marketing cycle that in the present invention with the Buying Cycle of brand characterizes brand is considered.Buying Cycle,
Buy the number of days needed for the average generation single purchase record of user's purchase of the brand.
Buy the user that brand j exceedes once and gather { userid }j,
The Buying Cycle of each user in user's set
Wherein, nijbuyRepresent that user buys brand j number, { dijbuyRepresent purchase date collection of the user i to brand j
Close.
The average value of user's Buying Cycle involved by brand j：
(3) temperature of brand
The temperature of brand refers to that brand is clicked and be purchased involved user's number in magnanimity records offline, is divided into
Click on temperature and purchase two standards of temperature.Wherein, clicking on temperature is：
Buying temperature is：
That is, clickUserjEqual to user's number (not repeating) that brand j carries out clicking operation, buyUserjEqual to brand j
Carry out user's number (not repeating) of clicking operation.
(4) brand purchase probability again
The purchase probability again of brand refers to the Probability p that brand can be purchased again after being purchased i timesbrandId(i).Division product
The unit gap whether board is bought again can be day, the week either moon, after purchased on the day of being in units of day (no matter
Buy number of packages), the probability that can be bought for the second time；It is (no matter to buy number of packages) after this week having purchaser record in units of week,
The purchased probability of the natural weekly assembly of certain later；It is that some later moon can quilt after this month having purchaser record in units of the moon
The probability of purchase.
First, in units of certain unit gap, purchase frequency ns of the statistics different user i to brand jij。
Calculate the purchase probability again of brand：
As shown in above formula, the number of users number of m times was bought to brand j divided by was calculated brand j was bought more than m times
The probability that the business of number of users number, as brand j can be purchased again after being purchased m times.
Step S403, value revision is carried out to the feature of the brand of said extracted, reduces influence of the exceptional value to model.Tool
For body：
The feature of structure includes time slicing, transformation ratio, marketing cycle, the temperature of brand that user operates to brand above
Deng.In mass data, some extreme users be present and the mad of some brands is clicked on or bought, be especially in the presence of some
Low buying rate under the high click of user, this step, which essentially consists in, controls these abnormal data.The data correction of the present embodiment
Mainly act on control to click on, buy, collect, the trend of shopping cart abnormal growth, pass through the correcting action of log functions so that
In the operations such as click, purchase still normal growth after log is taken in low value section, keep relatively steady after log is taken in high level space
It is fixed.
As shown in fig.3, it is the hardware architecture diagram of the brand identity extraction system of ecommerce recommended models of the present invention.Should
System includes time slicing module, characteristic extracting module and value revision module.
The basic data that the time slicing module is used for the brand to carrying out ecommerce sale carries out time slicing, structure
Make the brand identity sequence of different time piece.It is specific as follows：
The present embodiment illustrates by taking day cat store as an example.In day cat store, all there can be tens million of users to pass through daily
Brand finds the commodity oneself liked, and brand is to connect the most important tie of consumer and commodity.In existing historical record
In, the feature of brand how is extracted, first key feature is that the time series feature of Brand Marketing situation.
Conventional time slicing has three kinds of modes：
(1) split according to the natural date, split by week, first quarter moon, the moon, directly calculate the point in brand different time span
Situations such as hitting, collecting.With four months summary journal, if by can be divided into 16 week, the click of every, purchase, collection, shopping cart
To click in 7 days, purchase, collection, shopping cart sum；If by first quarter moon 8 can be divided into, the click of every, purchase, collection,
Shopping cart is to click on, buy in 14 days, collecting, shopping cart sum；If can monthly be divided into 4, the click of every, purchase,
Collection, shopping cart are to click on, buy in 30 days, collecting, shopping cart sum.
(2) the daily marketing situation of brand is directly multiplied by a time penalty factor K, brand is purchased than previous recently
Purchased even more important before individual month, so time punishment is inversely proportional with the time, the time, nearlyer penalty value was smaller, time more remote punishment
Value is bigger.
Wherein, x is current date and the distance of last day, and w is the decay factor that can be debugged, if selection 4 months for
Time span, w are test optimum coefficient 20.The change of length over time, w can do flexible adjustment.By daily click,
Purchase, collection, shopping cart is multiplied by time penalty factor, obtain tetra- kinds of behavior k of new brand j (click on purchase collection shopping cart)
Population characteristic value.
(3) by the marketing record of brand according to the date from closely to remote, the interval of burst is progressively longer near to protrude brand
The importance of situation is sold and be clicked to phase.By taking the record in April-July as an example, burst is described in detail below shown in table one, July
For newest behavior record, following 22 are segmented into.Date nearest 7 days, 7 are divided into units of day, then two weeks
With 3 days for time interval, it is divided into 5, ensuing one and a half months is divided into 6 in units of week, then that the remaining date is first
3 are extracted by first quarter moon, last moon is divided into individually a piece of according to the moon for unit.
The time slicing table of one user behavior scheme of table three
Different data set time length, different data set sizes, it can be constructed by three kinds of schemes above
The data characteristics of basal latency burst.Wherein the first scheme is partial to factor of influence less situation of the time to model, the
Two kinds of schemes are adapted to several months user behavior span, and the third scheme is applied to cutting for the time series behavior arbitrarily long to brand
Point, using phase fine granularity recently, the mode of remote date group granularity, build the time series characteristics of different brands.
Different time of the act bursts is provided to allow the forecast models such as logistic regression, decision tree, random forest more
Weight anticipation is carried out to the behavioral implications factor of the different time sections of user well.
In addition to the time slicing scheme of routine, the present embodiment proposes the time slicing mode based on buying behavior.
Time slicing mode based on buying behavior refers to user cut the time behavior sequence of brand to buy the date as cut-off
It is divided into different timeslices.The wherein involved timeslice of ith purchase refers to, three days to i-th after terminating from the i-th -1 time purchase
Secondary purchase terminates latter three days.
If but there are collection or shopping cart behavior, ith time buying section in three days after the i-th -1 time purchase terminates
Initial time for the i-th -1 time buy terminate latter three days collection or shopping cart date, while by the i-th -1 time purchase to collect
Or shopping cart behavior directly clicks on behavior according to buying and collecting behavior cutting, specifically as shown in Figure 2.
In Fig. 2, the initial purchase date is on July 15th, 2013, if not collecting and purchasing in three days after this date
Thing garage is, i.e., from July 18th, 2013 to the 18 days July in 2013 of latter three days of second of purchase for second of purchase when
Between piece, but due to having collection behavior in after initial purchase three days, then second of time buying piece is on July 17th, 2013
(collection date) clicks on 10 quantity according to purchase on July 18th, 2013, while by the click behavior after initial purchase
B and collection quantity c cuttings, wherein, 10*b/ (b+c) is included into initial purchase, and 10*c/ (b+c) is included into second of time buying
Piece.After time span t, number of clicks, collection and shopping cart number involved by purchase each to brand j user i, you can with
Brand purchased average required time length, number of clicks and collection, shopping cart number every time are obtained by average value.
Time slicing is carried out based on buying behavior with above-mentioned, the action time bought every time is meticulously analyzed, is advantageous to
Faster more accurately build the feature of brand.
The characteristic extracting module is used for the brand identity sequence of the different time piece according to above-mentioned construction, the friendship to brand
Easy data carry out temperature and cost analysis, extract the feature of brand.Specifically：
In offline recommending module, the feature and user characteristics of brand are equally crucial.The sales tactics of different brands, by
Ratings, customer purchase turn-head-rate are completely different, it is impossible to these brands lump together, therefore need to construct description product
The attribute and user's proposed algorithm module of board self-characteristic.Feature of the present embodiment from following direction structure brand：
(1) transformation ratio of brand
As the transformation ratio of user, brand also has the transformation ratio of itself, and the record that brand is operated every time is with being purchased
The transformation ratio relation of record can be hit with branch conversion ratio brandClickRate, collection conversion ratio brandFavoriteRate and
Shopping cart conversion ratio brandCartRate.Define brand and be clicked behavior to effective influence power of purchase and user's click behavior
Identical to the influence power of purchase, user's last time buying behavior terminates to terminate within latter three days to local buying behavior for latter 3rd day
All clicks, will buy that to terminate that the rear date delays three days be because it is that user returns mostly that each buying behavior is clicked in three days last time
It is not the effect of new purchasing demand caused by commodity have been purchased in visit.Such as user i to brand j in dkIt had purchaser record,
In dk+1It has n bar purchaser records, then in dk+ 3 to dk+1Caused click sum in+3 periodDivided by dk+1Record number is bought in reaching for it, as the click conversion ratio of the single purchase of brand.
Wherein, rijdbExpression user i is to brand j in the d days record numbers for operating b, and b=0 is clicks on, and b=1 is buys, b
=2 be collection, and b=3 can similarly be obtained, the collection conversion ratio of this purchase of user to add shopping cart：
With shopping cart conversion ratio：
The click, collection, shopping cart conversion ratio of each purchaser record of user are obtained, averages to obtain that to weigh the brand total
The behavior conversion ratio of body, include brand j click conversion ratio brandClickRatej, collect conversion ratio
brandFavoriteRatejWith shopping cart conversion ratio brandCartRatej, formula is as follows.
(2) the marketing cycle of brand
The marketing cycle of brand is used for weighing the Buying Cycle of brand, and different brands may belong to day consumption category, moon consumption product
Class or durable category.How the marketing cycle that in the present invention with the Buying Cycle of brand characterizes brand is considered.Buying Cycle,
Buy the number of days needed for the average generation single purchase record of user's purchase of the brand.
Buy the user that brand j exceedes once and gather { userid }j,
The Buying Cycle of each user in user's set
Wherein, nijbuyRepresent that user buys brand j number, { dijbuyRepresent purchase date collection of the user i to brand j
Close.
The average value of user's Buying Cycle involved by brand j：
(3) temperature of brand
The temperature of brand refers to that brand is clicked and be purchased involved user's number in magnanimity records offline, is divided into
Click on temperature and purchase two standards of temperature.Wherein, clicking on temperature is：
Buying temperature is：
That is, clickUserjEqual to user's number (not repeating) that brand j carries out clicking operation, buyUserjEqual to brand j
Carry out user's number (not repeating) of clicking operation.
(4) brand purchase probability again
The purchase probability again of brand refers to the Probability p that brand can be purchased again after being purchased i timesbrandId(i).Division product
The unit gap whether board is bought again can be day, the week either moon, after purchased on the day of being in units of day (no matter
Buy number of packages), the probability that can be bought for the second time；It is (no matter to buy number of packages) after this week having purchaser record in units of week,
The purchased probability of the natural weekly assembly of certain later；It is that some later moon can quilt after this month having purchaser record in units of the moon
The probability of purchase.
First, in units of certain unit gap, purchase frequency ns of the statistics different user i to brand jij。
Calculate the purchase probability again of brand：
As shown in above formula, the number of users number of m times was bought to brand j divided by was calculated brand j was bought more than m times
The probability that the business of number of users number, as brand j can be purchased again after being purchased m times.
The value revision module is used to carry out value revision to the feature of the brand of said extracted, reduces exceptional value to mould
The influence of type.Specifically：
The feature of structure includes time slicing, transformation ratio, marketing cycle, the temperature of brand that user operates to brand above
Deng.In mass data, some extreme users be present and the mad of some brands is clicked on or bought, be especially in the presence of some
Low buying rate under the high click of user, this step, which essentially consists in, controls these abnormal data.The data correction of the present embodiment
Mainly act on control to click on, buy, collect, the trend of shopping cart abnormal growth, pass through the correcting action of log functions so that
In the operations such as click, purchase still normal growth after log is taken in low value section, keep relatively steady after log is taken in high level space
It is fixed.
Although the present invention is described with reference to current better embodiment, those skilled in the art should be able to manage
Solution, above-mentioned better embodiment is only used for illustrating the present invention, is not used for limiting protection scope of the present invention, any in the present invention
Spirit and spirit within, any modification for being done, equivalence replacement, improvements etc., should be included in the right guarantor of the present invention
Within the scope of shield.
Claims (8)
1. a kind of brand identity extracting method of ecommerce recommended models, it is characterised in that this method comprises the following steps：
The basic data of brand to carrying out ecommerce sale carries out time slicing, so as to construct the brand of different time piece spy
Levy sequence；
According to the brand identity sequence of the different time piece of above-mentioned construction, temperature is carried out to the transaction data of brand and into one's duty
Analysis, extract the feature of brand；
Wherein：Described time slicing includes：Conventional time slicing and the time slicing based on buying behavior, wherein described normal
The time slicing of rule includes：According to natural date burst, according to the daily marketing situation of brand and time penalty factor burst, press
According to the date from closely being referred to by short elongated burst, the time slicing mode based on buying behavior user to brand to being far spaced
Time behavior sequence to buy the date as cut-off,
The marketing situation daily according to brand and time penalty factor burst include：By daily click, purchase, collection,
Time penalty factor is multiplied by four kinds of behaviors of shopping cart, obtains the above-mentioned four kinds of behavior k of new brand j population characteristic value：
<mrow>
<msub>
<mi>u</mi>
<mrow>
<mi>j</mi>
<mi>k</mi>
</mrow>
</msub>
<mo>=</mo>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mi>d</mi>
<mo>=</mo>
<mn>1</mn>
<mo>:</mo>
<mi>n</mi>
<mi>u</mi>
<mi>m</mi>
<mrow>
<mo>(</mo>
<mi>d</mi>
<mi>a</mi>
<mi>y</mi>
<mo>)</mo>
</mrow>
</mrow>
</munder>
<msub>
<mi>u</mi>
<mrow>
<mi>j</mi>
<mi>d</mi>
</mrow>
</msub>
<msub>
<mi>k</mi>
<mi>d</mi>
</msub>
<mo>=</mo>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mi>d</mi>
<mo>=</mo>
<mn>1</mn>
<mo>:</mo>
<mi>n</mi>
<mi>u</mi>
<mi>m</mi>
<mrow>
<mo>(</mo>
<mi>d</mi>
<mi>a</mi>
<mi>y</mi>
<mo>)</mo>
</mrow>
</mrow>
</munder>
<msub>
<mi>u</mi>
<mrow>
<mi>j</mi>
<mi>d</mi>
</mrow>
</msub>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<mfrac>
<mrow>
<mi>n</mi>
<mi>u</mi>
<mi>m</mi>
<mrow>
<mo>(</mo>
<mi>d</mi>
<mi>a</mi>
<mi>y</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mi>d</mi>
</mrow>
<mi>w</mi>
</mfrac>
</mrow>
</msup>
<mo>.</mo>
</mrow>
2. the method as described in claim 1, it is characterised in that this method also includes：The feature of the brand of said extracted is entered
Row value revision.
3. the method as described in claim 1, it is characterised in that the feature of described brand includes：Transformation ratio, the brand of brand
The marketing cycle, the temperature of brand, brand purchase probability again.
4. method as claimed in claim 2, it is characterised in that described value revision refers to the carry out numerical value by log functions
Amendment.
5. a kind of brand identity extraction system of ecommerce recommended models, it is characterised in that the system includes time slicing mould
Block, characteristic extracting module, wherein：
The basic data that the time slicing module is used for the brand to carrying out ecommerce sale carries out time slicing, so as to structure
Make the brand identity sequence of different time piece；
The characteristic extracting module is used for the brand identity sequence of the different time piece according to above-mentioned construction, to the number of deals of brand
According to temperature and cost analysis is carried out, the feature of brand is extracted；
Wherein：Described time slicing includes：Conventional time slicing and the time slicing based on buying behavior, wherein described normal
The time slicing of rule includes：According to natural date burst, according to the daily marketing situation of brand and time penalty factor burst, press
According to the date from closely being referred to by short elongated burst, the time slicing mode based on buying behavior user to brand to being far spaced
Time behavior sequence to buy the date as cut-off,
The marketing situation daily according to brand and time penalty factor burst include：By daily click, purchase, collection,
Time penalty factor is multiplied by four kinds of behaviors of shopping cart, obtains the above-mentioned four kinds of behavior k of new brand j population characteristic value：
<mrow>
<msub>
<mi>u</mi>
<mrow>
<mi>j</mi>
<mi>k</mi>
</mrow>
</msub>
<mo>=</mo>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mi>d</mi>
<mo>=</mo>
<mn>1</mn>
<mo>:</mo>
<mi>n</mi>
<mi>u</mi>
<mi>m</mi>
<mrow>
<mo>(</mo>
<mi>d</mi>
<mi>a</mi>
<mi>y</mi>
<mo>)</mo>
</mrow>
</mrow>
</munder>
<msub>
<mi>u</mi>
<mrow>
<mi>j</mi>
<mi>d</mi>
</mrow>
</msub>
<msub>
<mi>k</mi>
<mi>d</mi>
</msub>
<mo>=</mo>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mi>d</mi>
<mo>=</mo>
<mn>1</mn>
<mo>:</mo>
<mi>n</mi>
<mi>u</mi>
<mi>m</mi>
<mrow>
<mo>(</mo>
<mi>d</mi>
<mi>a</mi>
<mi>y</mi>
<mo>)</mo>
</mrow>
</mrow>
</munder>
<msub>
<mi>u</mi>
<mrow>
<msub>
<mi>jd</mi>
<mi>o</mi>
</msub>
</mrow>
</msub>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<mfrac>
<mrow>
<mi>n</mi>
<mi>u</mi>
<mi>m</mi>
<mrow>
<mo>(</mo>
<mi>d</mi>
<mi>a</mi>
<mi>y</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mi>d</mi>
</mrow>
<mi>w</mi>
</mfrac>
</mrow>
</msup>
<mo>.</mo>
</mrow>
6. system as claimed in claim 5, it is characterised in that the system also includes value revision module, the value revision
Module is used to carry out value revision to the feature of the brand of said extracted.
7. system as claimed in claim 5, it is characterised in that the feature of described brand includes：Transformation ratio, the brand of brand
The marketing cycle, the temperature of brand, brand purchase probability again.
8. system as claimed in claim 6, it is characterised in that described value revision refers to the carry out numerical value by log functions
Amendment.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410593894.XA CN104408641B (en) | 2014-10-29 | 2014-10-29 | The brand identity extracting method and system of ecommerce recommended models |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410593894.XA CN104408641B (en) | 2014-10-29 | 2014-10-29 | The brand identity extracting method and system of ecommerce recommended models |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104408641A CN104408641A (en) | 2015-03-11 |
| CN104408641B true CN104408641B (en) | 2018-02-06 |
Family
ID=52646270
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410593894.XA Active CN104408641B (en) | 2014-10-29 | 2014-10-29 | The brand identity extracting method and system of ecommerce recommended models |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104408641B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106294483A (en) * | 2015-06-04 | 2017-01-04 | 北京国双科技有限公司 | For mating data processing method and the device of mouthpiece for product brand |
| CN108665290A (en) * | 2017-03-27 | 2018-10-16 | 增长引擎（北京）信息技术有限公司 | The method and system of trade company's conversion ratio is improved based on user access path |
| CN109493101A (en) * | 2017-09-11 | 2019-03-19 | 北京京东尚科信息技术有限公司 | Target brand message determines method, apparatus, electronic equipment and storage medium |
| CN108898425A (en) * | 2018-06-14 | 2018-11-27 | 口碑(上海)信息技术有限公司 | The evaluation method and device of shop quality |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1623343A2 (en) * | 2003-04-25 | 2006-02-08 | Sap Ag | Automated data mining runs |
| CN102800006A (en) * | 2012-07-23 | 2012-11-28 | 姚明东 | Real-time goods recommendation method based on customer shopping intention exploration |
| CN102956009A (en) * | 2011-08-16 | 2013-03-06 | 阿里巴巴集团控股有限公司 | Electronic commerce information recommending method and electronic commerce information recommending device on basis of user behaviors |
| CN103942330A (en) * | 2014-04-30 | 2014-07-23 | 深圳先进技术研究院 | Method and system for processing big data |
-
2014
- 2014-10-29 CN CN201410593894.XA patent/CN104408641B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1623343A2 (en) * | 2003-04-25 | 2006-02-08 | Sap Ag | Automated data mining runs |
| CN102956009A (en) * | 2011-08-16 | 2013-03-06 | 阿里巴巴集团控股有限公司 | Electronic commerce information recommending method and electronic commerce information recommending device on basis of user behaviors |
| CN102800006A (en) * | 2012-07-23 | 2012-11-28 | 姚明东 | Real-time goods recommendation method based on customer shopping intention exploration |
| CN103942330A (en) * | 2014-04-30 | 2014-07-23 | 深圳先进技术研究院 | Method and system for processing big data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104408641A (en) | 2015-03-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104331817B (en) | The user characteristics extracting method and system of e-commerce recommended models | |
| US20170262899A1 (en) | Computing Mathematically-Optimized Properties for Paid Search | |
| CN102254028A (en) | Personalized commodity recommending method and system which integrate attributes and structural similarity | |
| US8620746B2 (en) | Scoring quality of traffic to network sites | |
| CN104866474A (en) | Personalized data searching method and device | |
| Ahmed et al. | Establishing standard rules for choosing best KPIs for an e-commerce business based on google analytics and machine learning technique | |
| CN104408641B (en) | The brand identity extracting method and system of ecommerce recommended models | |
| CN106447463A (en) | Commodity recommendation method based on Markov decision-making process model | |
| CN107066476A (en) | A kind of real-time recommendation method based on article similarity | |
| CN102385601A (en) | Product information recommendation method and system | |
| US20030036890A1 (en) | Predictive method | |
| CN105096152A (en) | Commodity popularity-based operation execution method and device | |
| CN102629360A (en) | Effective dynamic commodity recommendation method and commodity recommendation system | |
| JP6679451B2 (en) | Selection device, selection method, and selection program | |
| US20140180799A1 (en) | Techniques for optimizing the impact of video content on electronic commerce sales | |
| CN107292713A (en) | A kind of rule-based individual character merged with level recommends method | |
| CN101853282A (en) | System and method for extracting information of cross-site shopping mode of user | |
| US20020194148A1 (en) | Predictive method | |
| Li | Accurate digital marketing communication based on intelligent data analysis | |
| CN105303447A (en) | Method and device for carrying out credit rating through network information | |
| CN103412865A (en) | Method and system for carrying out notification on website projects | |
| Kupfer et al. | Valuable information in early sales proxies: The use of Google search ranks in portfolio optimization | |
| JP7011552B2 (en) | Ad management system, ad management method, and ad management program | |
| CN107622409A (en) | Purchase the Forecasting Methodology and prediction meanss of car ability | |
| Deng et al. | Multiple kernel learning on time series data and social networks for stock price prediction |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |