CN108629665A - A kind of individual commodity recommendation method and system - Google Patents
A kind of individual commodity recommendation method and system Download PDFInfo
- Publication number
- CN108629665A CN108629665A CN201810433175.XA CN201810433175A CN108629665A CN 108629665 A CN108629665 A CN 108629665A CN 201810433175 A CN201810433175 A CN 201810433175A CN 108629665 A CN108629665 A CN 108629665A
- Authority
- CN
- China
- Prior art keywords
- commodity
- user
- historical behavior
- behavior data
- training sample
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 66
- 230000006399 behavior Effects 0.000 claims abstract description 99
- 238000012549 training Methods 0.000 claims abstract description 60
- 238000013136 deep learning model Methods 0.000 claims abstract description 40
- 238000012163 sequencing technique Methods 0.000 claims abstract description 7
- 230000000306 recurrent effect Effects 0.000 claims description 21
- 230000003993 interaction Effects 0.000 claims description 18
- 239000013065 commercial product Substances 0.000 claims description 10
- 238000001914 filtration Methods 0.000 claims description 10
- 238000012216 screening Methods 0.000 claims description 6
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 230000001537 neural effect Effects 0.000 claims description 4
- 238000003062 neural network model Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 6
- 238000013528 artificial neural network Methods 0.000 description 18
- 230000003542 behavioural effect Effects 0.000 description 16
- 239000000047 product Substances 0.000 description 6
- 230000002452 interceptive effect Effects 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 3
- 206010034719 Personality change Diseases 0.000 description 2
- 238000013459 approach Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 230000001154 acute effect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 208000015181 infectious disease Diseases 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 230000015654 memory Effects 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000002203 pretreatment Methods 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 239000000516 sunscreening agent Substances 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
Abstract
The present invention discloses a kind of individual commodity recommendation method and system, and method includes:The historical behavior data for obtaining multiple users in preset time period, obtain the first training sample after being arranged according to pre-defined rule;Impact factor is obtained as the second training sample based on cosine similarity method;First training sample and the second training sample are trained as the training sample of deep learning model, the deep learning model trained;The interested items list of user that output model predicts.The present invention effectively utilizes the timing information of commodity in user's history behavior, make the commodity in historical behavior that there is different weighted values according to the time sequencing that its interbehavior occurs in commending system calculating, commodity impact factor embody commodity global characteristics and the user to the level of interest of the commodity, it is effectively increased the characteristic quantity that deep learning model is got, effectively promotes the personalized recommendation effect to cold start-up user.
Description
Technical field
The present invention relates to a kind of individual commodity recommendation method and systems, belong to recommended technology field.
Background technology
With universal and e-commerce the fast development of internet, more and more users are clear by e-commerce platform
Look at and buy commodity.Type of merchandize is various in e-commerce platform, if the help of recommended engine with no personalization, it is difficult to by user
Interested commodity accurately recommend user.
One of the main stream approach of personalized recommendation method that current e-commerce platform uses:Collaborative filtering method leads to
Cross look for user similar in the historical viewings of active user or buying behavior, the interested commodity of the similar user of behavior are pushed away
It recommends to active user.Based on the personalized recommendation method of collaborative filtering for all historical viewings of active user or the quotient of purchase
Product give same weight when calculating the similarity with other commodity, such as the commodity first that active user browsed before ten days,
Commodity second was browsed before one hour, commodity first and commodity second are to calculating active user commodity interested in collaborative filtering method
When the effect that is played be identical.But from the point of view of experience, the commodity second that was browsed before one hour often than ten days before the quotient that browses
Product first is to the reference value bigger of current individual commodity recommendation, therefore the personalized recommendation method based on collaborative filtering cannot
Embodying weight difference of the commodity of different order in historical behavior in current Personalized recommendation causes recommendation effect not accurate enough
Really.In addition, user's history browse or buy data it is very little when collaborative filtering it is ineffective, that is, there is cold start-up and ask
Topic.
The two of main stream approach:Method based on recurrent neural network, by the way that the historical behavior of user is sequentially input recurrence
The interested commodity of user are trained and predicted in neural network.This method can utilize the commodity in user's history behavior
Temporal aspect, thus it is suitable in training sample abundance and recurrent neural network parameter regulation, and effect is generally better than association
Same filter algorithm.Based on the personalized recommendation method of recurrent neural network in the case where user's history behavioral data is less, by
It predicts that the historical viewings of interest-degree or purchase commodity data are very little in can be used for Recursive Networks, therefore there are problems that cold start-up.
There is cold start-up in above two method, i.e., cannot be very when retrievable managing operation history is less
Recommendations well, affect user experience.Cold start-up problem is personalized recommendation field common problem, by being slapped
The historical behavior data of the current user to be recommended held are very little, it is difficult to which accurately providing it to the user may interested quotient
Product.
About one of main solution of cold start-up problem in existing scheme:Method based on user information, according to
The essential informations such as age, gender, the residence at family are recommended.But since certain customers may not fill in detailed
People's information causes the user information that commending system can obtain limited.And the recommendation done according to information such as user's gender, ages
The common interest of the user group can only be represented, is not the personalized recommendation for the user.
In existing scheme about the main solution of cold start-up problem two:Based on the method for label selection, in user
A variety of tag along sorts are provided when registration uses for the first time to select for user, are selected user in the case that user behavior data is insufficient
Labeling under commercial product recommending to user.Based on label selection method the problem is that:Certain customers may not
Tag along sort is filled at the beginning of use, it is likely that selection " skipping the step " chooses at random, and causes selected label can not be true
Just represent user interest.And the interested commodity of user may can be changed at any time, therefore originally selection interest tags can with work as
There are bigger differences between preceding commodity interested, such as certain user commodity interested at summer may be " suncream ", winter
" down jackets " are then changing into, will produce very large deviation if selecting label to be recommended originally by user.Further, since commodity kind
Class is varied, and all product features cannot be completely covered in label.
Invention content
In view of the foregoing drawbacks, the present invention provides a kind of individual commodity recommendation method and systems, obtain user's history row
After data, analysis and arrangement is carried out to data, is corresponded to using the commodity in the historical behavior information and historical behavior of active user
Global characteristics etc. be calculated the impact factors of the corresponding each commodity of active user, and by the commodity in user's history behavior
And its corresponding impact factor is sequentially input by the time sequencing that user interacts with commodity in preset deep learning model, is obtained
The interested commodity of active user of prediction, and the level of interest according to prediction is ranked up, and it is by institute's alignment sequence that user is emerging
For the interesting highest several commercial product recommendings of degree to user, deep learning model uses recurrent neural networks model.
Present invention may effectively utilize the timing information of commodity in user's history behavior, the commodity in historical behavior is made to push away
Recommend in system-computed has different weighted values according to the time sequencing that its interbehavior occurs, and the commodity in the present invention influence
The factor embody commodity global characteristics and the user to the level of interest of the commodity, can in user's history behavioral data deficiency
To be effectively increased the characteristic quantity that deep learning model is got, the personalized recommendation effect to cold start-up user is effectively promoted.
In order to achieve the above objectives, the present invention implements by the following technical programs:
The present invention provides a kind of individual commodity recommendation method, this method includes:
The historical behavior data for obtaining multiple users in preset time period, obtain the first training after being arranged according to pre-defined rule
Sample;
Using the historical behavior data after arrangement, the historical behavior number of each user is calculated based on cosine similarity method
The corresponding impact factor of each commodity is as the second training sample in;
First training sample and the second training sample are trained as the training sample of deep learning model, obtained
Trained deep learning model;
The first training sample of pre- recommended user and the second training sample are input in the deep learning model trained,
The interested items list of user that output model predicts is recommended pre- recommendation according to the commodity sequence in items list and is used
Family.
Further, it is described arranged according to pre-defined rule after the step of obtaining the first training sample include:
The historical behavior data of multiple users, filter out the interbehavior of particular category in preset time period based on acquisition
Data;
Convert the information in the interbehavior data of particular category to the storage of unique number form;
The interbehavior data of each user in the interbehavior data of particular category are sorted in chronological order.
Further, each commodity in the historical behavior data that each user is calculated based on cosine similarity method
The specific steps of corresponding impact factor include:
It is calculated after the N-dimensional vector of particular commodity based on cosine similarity method and maximum value normalizing is carried out to the vector
Change, make in vector each numerical value between [0,1], N-dimensional vector is the corresponding impact factor of commodity, and N is pretreated
The total quantity of commodity or in advance after in pretreated historical behavior data by certain Rules Filterings in historical behavior data
Candidate commodity amount, the value Value of i-th of number in the impact factor that the corresponding N-dimensional vector of commodity indicatesiSpecific calculating
Method includes:
Wherein, Count (i-th of commodity of particular commodity &) represents particular commodity and i-th of commodity appears in pretreatment jointly
The number in historical behavior data afterwards, i >=1 and i≤N, Count (particular commodity) represent particular commodity according to pre-defined rule
The interaction total degree in historical behavior data after arrangement, Count (i-th of commodity) represent i-th of commodity according to pre- set pattern
The interaction total degree in historical behavior data after then arranging, Max (Values) are represented in the historical behavior data of each user
Maximum value in the corresponding impact factor of all commodity.
Further, the deep learning model includes but not limited to recurrent neural networks model.
Further, the commodity sequence according in items list recommends pre- recommended user, including:
Several commercial product recommendings before being come according to the sequence of the commercial productainterests degree predicted in items list from high to low
To pre- recommended user.
The present invention also provides a kind of individual commodity recommendation systems, including:
Acquisition module, the historical behavior data for obtaining multiple users in preset time period, arranges according to pre-defined rule
After obtain the first training sample;
Computing module, for using the historical behavior data after arranging, each use to be calculated based on cosine similarity method
Each corresponding impact factor of commodity is as the second training sample in the historical behavior data at family;
Training module, for using the first training sample and the second training sample as the training sample of deep learning model into
Row training, the deep learning model trained;
Recommending module, for the first training sample of pre- recommended user and the second training sample to be input to the depth trained
It spends in learning model, the interested items list of user that output model predicts, is pushed away according to the commodity sequence in items list
It recommends to pre- recommended user.
The acquisition module, including:
Screening unit filters out specific for the historical behavior data of multiple users in the preset time period based on acquisition
The interbehavior data of classification;
Conversion unit is converted into the storage of unique number form for the information in the interbehavior data by particular category;
Sequencing unit is used for the interbehavior data of each user in the interbehavior data of particular category temporally
Sequence sorts.
The computing module, including:
Computing unit, for based on cosine similarity method be calculated after the N-dimensional vector of particular commodity to the vector into
Row maximum value normalizes, and makes in vector that for each numerical value between [0,1], N-dimensional vector is the corresponding impact factor of commodity, N
Pass through certain for the total quantity of commodity in pretreated historical behavior data or in advance from pretreated historical behavior data
Candidate commodity amount after a little Rules Filterings, the value of i-th of number in the impact factor that the corresponding N-dimensional vector of commodity indicates
ValueiCircular include:
Wherein, Count (i-th of commodity of particular commodity &) represents particular commodity and i-th of commodity appears in pretreatment jointly
The number in historical behavior data afterwards, i >=1 and i≤N, Count (particular commodity) represent particular commodity according to pre-defined rule
The interaction total degree in historical behavior data after arrangement, Count (i-th of commodity) represent i-th of commodity according to pre- set pattern
The interaction total degree in historical behavior data after then arranging, Max (Values) are represented in the historical behavior data of each user
Maximum value in the corresponding impact factor of all commodity.
The deep learning model includes but not limited to recurrent neural networks model.
The commodity sequence according in items list recommends pre- recommended user, including:
Several commercial product recommendings before being come according to the sequence of the commercial productainterests degree predicted in items list from high to low
To pre- recommended user.
The beneficial effects of the invention are as follows:
The technical solution provided through the invention can handle the user's history behavioral data comprising timing information, due to
Correlation between the commodity that different time interacts and the interested commodity of active user is different, in deep learning model
The timing information for including in user's history behavior is extracted and personalized using more accurately commodity can be carried out to user
Recommend.The present invention creatively proposes commodity impact factor, finds that each commodity are had from the historical behavior of multiple users
Global characteristics and thus calculate the corresponding impact factor of commodity, by pretreated user's history behavior commodity vector
The vectorial expression-form of expression-form and its corresponding impact factor is input to jointly in deep learning model, due to impact factor
Contain more information so that the prediction commodity that deep learning model obtains are more acurrate, waiting for carrying out personalized recommendation to it
User's history behavioral data it is insufficient in the case of can more embody its superiority, in cold start-up problem be better than original individual character
Change recommendation method.
Description of the drawings
Fig. 1 show a kind of one flow chart of embodiment of individual commodity recommendation method provided by the invention.
Specific implementation mode
Technical scheme of the present invention is specifically addressed below, it should be pointed out that technical scheme of the present invention is unlimited
Embodiment described in embodiment, those skilled in the art's reference and the content for using for reference technical solution of the present invention, in this hair
The improvement and design carried out on the basis of bright, should belong to the scope of protection of the present invention.
Embodiment one
The embodiment of the present invention one provides a kind of individual commodity recommendation method, the method comprising the steps of S110-S140:
In step s 110, the historical behavior data for obtaining multiple users in preset time period, arrange according to pre-defined rule
After obtain the first training sample.
Preset time segment occurrence and the number of users of extraction can be set according to actual conditions, such as the period is arranged
It it is one month, extraction number of users is Num, then is used from randomly selected Num in current all extracting datas one month
The historical behavior data at family." down jackets " have been browsed before usually there is timeliness, such as user's half a year due to personalized recommendation, and
" one-piece dress " is browsed in the recent period, recommendation effect may run counter to desire if the browsing data of user before still considering half a year, therefore
The occurrence of preset time period need to be set according to practical situations.
Further, it is described arranged according to pre-defined rule after the step of obtaining the first training sample include:Based on acquisition
The historical behavior data of multiple users in preset time period, filter out the interbehavior data of particular category;By particular category
Information in interbehavior data is converted into the storage of unique number form;By each use in the interbehavior data of particular category
The interbehavior data at family sort in chronological order.
Wherein, the step of being arranged according to pre-defined rule is the preprocessing process of the present invention, for according to being carried in preset time period
The historical behavior data got carry out data preparation.Since specific implementation scene is different, behavioral data type may be different, such as
Behavioral data may be the diversified forms such as to browse, pay close attention to, listening to, watching, therefore " interaction " acute pyogenic infection of finger tip user and the commodity is used to generate
Behavioral data.Preprocessing process includes specifically that the information in the interbehavior data by particular category is converted into unique volume
The storage of number form;The interbehavior data of each user in the interbehavior data of particular category are sorted in chronological order
Operation.
The historical behavior data of multiple users, filter out the interbehavior of particular category in preset time period based on acquisition
Data are to select the behavioral data of particular category from all interbehavior data of user according to the needs of concrete application.Example
The behavioral data of user can have browsing, folder of puting into collection, shopping cart, purchase behavior is added in such as shopping website, if drafting
Personalized recommendation only is carried out using the browsing data of user, then extracts the browsing commodity data of user, and ignores user's addition
The interbehaviors such as collection.
The information in the interbehavior data of particular category is converted to the storage of unique number form that is, specifying information is turned
The replacement of number form is turned to, is specially replaced user in the form of numbering, commodity are replaced in the form of numbering, and different user is handed over
The considerations of mutual same goods number is identical, this step is primarily for protection privacy of user.Goods number is then switched into vector
Form indicates that each commodity have unique vector representation, convenient for the input as deep learning model.By goods number
There are many kinds of the methods for switching to vector form expression, as long as ensureing after conversion that each commodity have unique vector representation
It will not obscure with other commodity.For example, a kind of method that goods number is switched to vector form expression can be:N is pre-
The total quantity of commodity or from pretreated historical behavior data pass through certain rule in advance in historical behavior data that treated
Candidate commodity amount after then screening, a N-dimensional vector is built, commodity first is expressed as [1,0,0 ..., 0,0], i.e., the vector removes
First is outside 1, other N-1 is 0, commodity second is represented by [0,1,0 ..., 0,0], i.e., the vector is in addition to second
Outside 1, other N-1 is 0, and converting N number of commodity in commodity library to N number of vector form by this method indicates.
The interbehavior data of each user in the interbehavior data of particular category are sorted in chronological order and are
By being ranked sequentially for time order and function that the commodity of same user filtered out after specific interbehavior are occurred by interbehavior, first send out
The commodity of raw interaction occur after coming before interactive commodity.
In the step s 120, using the historical behavior data after arrangement, each use is calculated based on cosine similarity method
Each corresponding impact factor of commodity is as the second training sample in the historical behavior data at family.
Further, each commodity in the historical behavior data that each user is calculated based on cosine similarity method
The specific steps of corresponding impact factor include:
It is calculated after the N-dimensional vector of particular commodity based on cosine similarity method and maximum value normalizing is carried out to the vector
Change, make in vector each numerical value between [0,1], N-dimensional vector is the corresponding impact factor of commodity, and N is pretreated
The total quantity of commodity or in advance after in pretreated historical behavior data by certain Rules Filterings in historical behavior data
Candidate commodity amount, the value Value of i-th of number in the impact factor that the corresponding N-dimensional vector of commodity indicatesiSpecific calculating
Method includes:
Wherein, Count (i-th of commodity of particular commodity &) represents particular commodity and i-th of commodity appears in pretreatment jointly
The number in historical behavior data afterwards, i >=1 and i≤N, Count (particular commodity) represent particular commodity according to pre-defined rule
The interaction total degree in historical behavior data after arrangement, Count (i-th of commodity) represent i-th of commodity according to pre- set pattern
The interaction total degree in historical behavior data after then arranging, Max (Values) are represented in the historical behavior data of each user
Maximum value in the corresponding impact factor of all commodity.The N of particular commodity after maximum value normalization is calculated by the method
Dimensional vector, and in vector each numerical value between [0,1], N-dimensional vector at this time be the corresponding influence of particular commodity because
Son.The correlation of the bigger commodity for indicating that the particular commodity is corresponding with i-th of number of the numerical value of i-th of number is bigger in vector.
Using pretreated historical behavior data, the number occurred in historical behavior data according to commodity may determine that
Go out the temperature of commodity, certain two commodity is by information such as the interested numbers of same user.In general, it is interacted by more users
Commodity are more likely to active user can be caused to interact, and two commodity often appear in the historical behavior data of same user simultaneously
It can be shown that the relevance of the two commodity is stronger when middle, if active user interacts with the generation of one of commodity, have larger
Probability can generate interest to another commodity.
In step s 130, using the first training sample and the second training sample as the training sample of deep learning model into
Row training, the deep learning model trained.
In the historical behavior data and step S120 that are obtained after being arranged what is obtained in step S110 according to pre-defined rule
To each commodity impact factor as training sample input deep learning model in be trained.The depth that this method uses
It is recurrent neural networks model to practise model, which can be RNN models and its improved model, such as LSTM
Deng.
For each parameter elder generation random initializtion in deep learning model, then according to pretreated historical behavior data
In Customs Assigned Number successively will by the interaction time of origin sequencing of the corresponding interactive commodity of the user for each user
The vector expression of the corresponding vector expression of interactive commodity and impact factor, which inputs, carries out god in recurrent neural networks model
Training through network.
For recurrent neural networks model, the input of ith is that the user interacts in pretreated behavioral data
I-th of commodity vectorial expression-form and its impact factor vectorial expression-form, at this time model obtain one prediction it is defeated
Go out, the i+1 commodity which interacts with the user in pretreated behavioral data are compared, calculates recurrent neural
The deviation of network and the parameter that neural network model is constantly corrected according to deviation.When the behavioral data of a user after the pre-treatment
The storewide vectorial expression-form of middle interaction and its vectorial expression-form of impact factor sequentially input deep learning mould
After type is trained, by the vectorial expression-form of the interaction commodity of next user in pretreated behavioral data and its influence because
The vectorial expression-form of son is sequentially input in chronological order in the recurrent neural networks model trained, and continues to train,
Until the vector table of the vectorial expression-form of the interaction commodity of whole users and its impact factor in pretreated behavioral data
Deep learning model training is fully entered up to form to finish.The deep learning model trained at this time.Due to recurrent neural
Network model generally uses in deep learning field, therefore builds this method no longer to the specific of recurrent neural networks model
It repeats.
The model can be by sequentially inputting mould by the pretreated historical behavior data of the user of pending personalized recommendation
The next commodity predicted in type, specific output are the prediction score of all commodity in selected commodity library, and press
Score value arranges from high to low.Selected commodity library can be the entire service in the platform of recommendation method application, or according to certain
Candidate commodity library after the screening in advance of a little principles.Such as shopping platform is want to carry out new commodity recommendation to user, from 2000 it is new on
Select several commercial product recommendings to user in line commodity, then what candidate commodity library at this time included is this 2000 new commodity of reaching the standard grade.
In step S140, the first training sample of pre- recommended user and the second training sample are input to the depth trained
It spends in learning model, the interested items list of user that output model predicts, is pushed away according to the commodity sequence in items list
It recommends to pre- recommended user.
When carrying out personalized recommendation to pre- recommended user or active user, pre- recommended user or active user are got in institute
Select the historical behavior data in the period, and by obtaining pre- recommended user or active user in step S110 and step S120
Historical behavior data and the corresponding impact factor of each commodity after being arranged according to pre-defined rule are input in step S130 and have trained
Deep learning model in calculated.The score of each commodity in the commodity library that the deep learning model trained is predicted,
And sort from high to low by score, the as interested items list of active user, and from high to low according to the interest-degree of prediction
Sequence will come before several commercial product recommendings to active user.The higher commodity of the level of interest of prediction come more forward
Position.
If only to the commodity in user's recommended candidate commodity library, such as 2000 new commodity of reaching the standard grade, then only to this 2000
The prediction score of commodity is calculated and is sorted.The specific commodity amount then recommended as needed to active user, from working as
Several commercial product recommendings are completed to active user to active user's before selection comes in the preceding interested items list of user
Personalized recommendation.
Commodity of the present invention are not limited to the physical commodities such as clothes, daily necessities, for multimedia platform, according to user
Listen song list or other users and the interactive information of music platform to carry out personalized recommendation music to user, or according to user's viewing or
With other interactive information of film by film personalized recommendation to user when, specific music and film can be considered quotient
Product.
Interbehavior of multiple users in special time period is collected, includes on internet platform according to user
The interbehavior of timing information predicts that this can be used in the interested commodity of user using improved recurrent neural networks model
Method carries out personalized recommendation.Improved recurrent neural networks model may be the neuronal quantity of network model, network model
The number of plies, addition threshold function etc., if the recurrent neural networks model after improved structure do not directly affect this method proposition by quotient
Product timing information and impact factor are added to the mode of recurrent neural networks model, then can be considered this method implementation method it
One.
Used deep learning model can be recurrent neural networks model or make improvements after model, such as
LSTM models etc. if deep learning model for recursiveness is ranked up commodity by certain rule, and are sequentially sequentially input
Its corresponding deep learning output is the feature of next commodity after the corresponding feature of current commodity, then can be considered recurrent neural net
Network model and its improved model.
Interbehavior can be user browse commodity, watch or listen to, folder of puting into collection, that shopping cart, purchase etc. is added is more
One or a combination set of kind behavior behavior.
If the coefficient etc. for only having changed impact factor calculating method still falls within this method protection domain.
Embodiment two
The embodiment of the present invention two additionally provides a kind of individual commodity recommendation system, including:
Acquisition module, the historical behavior data for obtaining multiple users in preset time period, arranges according to pre-defined rule
After obtain the first training sample;
Computing module, for using the historical behavior data after arranging, each use to be calculated based on cosine similarity method
Each corresponding impact factor of commodity is as the second training sample in the historical behavior data at family;
Training module, for using the first training sample and the second training sample as the training sample of deep learning model into
Row training, the deep learning model trained;
Recommending module, for the first training sample of pre- recommended user and the second training sample to be input to the depth trained
It spends in learning model, the interested items list of user that output model predicts, is pushed away according to the commodity sequence in items list
It recommends to pre- recommended user.
The acquisition module, including:
Screening unit filters out specific for the historical behavior data of multiple users in the preset time period based on acquisition
The interbehavior data of classification;
Conversion unit is converted into the storage of unique number form for the information in the interbehavior data by particular category;
Sequencing unit is used for the interbehavior data of each user in the interbehavior data of particular category temporally
Sequence sorts.
The computing module, including:
Computing unit, for based on cosine similarity method be calculated after the N-dimensional vector of particular commodity to the vector into
Row maximum value normalizes, and makes in vector that for each numerical value between [0,1], N-dimensional vector is the corresponding impact factor of commodity, N
Pass through certain for the total quantity of commodity in pretreated historical behavior data or in advance from pretreated historical behavior data
Candidate commodity amount after a little Rules Filterings, the value of i-th of number in the impact factor that the corresponding N-dimensional vector of commodity indicates
ValueiCircular include:
Wherein, Count (i-th of commodity of particular commodity &) represents particular commodity and i-th of commodity appears in pretreatment jointly
The number in historical behavior data afterwards, i >=1 and i≤N, Count (particular commodity) represent particular commodity according to pre-defined rule
The interaction total degree in historical behavior data after arrangement, Count (i-th of commodity) represent i-th of commodity according to pre- set pattern
The interaction total degree in historical behavior data after then arranging, Max (Values) are represented in the historical behavior data of each user
Maximum value in the corresponding impact factor of all commodity.
The deep learning model includes but not limited to recurrent neural networks model.
The commodity sequence according in items list recommends pre- recommended user, including:
Several commercial product recommendings before being come according to the sequence of the commercial productainterests degree predicted in items list from high to low
To pre- recommended user.
The specific steps that the function and processing mode of specific implementation are described referring to embodiment of the method one.
The processing and function realized by the system of the present embodiment two essentially correspond to the reality of aforementioned method shown in FIG. 1
Apply example, principle and example, therefore not detailed place in the description of the present embodiment, the related description in previous embodiment is may refer to,
This will not be repeated here.
The beneficial effects of the invention are as follows:
The technical solution provided through the invention can handle the user's history behavioral data comprising timing information, due to
Correlation between the commodity that different time interacts and the interested commodity of active user is different, in deep learning model
The timing information for including in user's history behavior is extracted and personalized using more accurately commodity can be carried out to user
Recommend.The present invention creatively proposes commodity impact factor, finds that each commodity are had from the historical behavior of multiple users
Global characteristics and thus calculate the corresponding impact factor of commodity, by pretreated user's history behavior commodity vector
The vectorial expression-form of expression-form and its corresponding impact factor is input to jointly in deep learning model, due to impact factor
Contain more information so that the prediction commodity that deep learning model obtains are more acurrate, waiting for carrying out personalized recommendation to it
User's history behavioral data it is insufficient in the case of can more embody its superiority, in cold start-up problem be better than original individual character
Change recommendation method.
Through the above description of the embodiments, those skilled in the art can be understood that the present invention can lead to
Hardware realization is crossed, the mode of necessary general hardware platform can also be added to realize by software.Based on this understanding, this hair
Bright technical solution can be expressed in the form of software products, which can be stored in a non-volatile memories
In medium (can be CD-ROM, USB flash disk, mobile hard disk etc.), including some instructions are used so that a computer equipment (can be
Personal computer, server or network equipment etc.) execute method described in each embodiment of the present invention.
Disclosed above is only several specific embodiments of the present invention, and still, the present invention is not limited to above-described embodiment,
The changes that any person skilled in the art can think of should all fall into protection scope of the present invention.
Claims (10)
1. a kind of individual commodity recommendation method, which is characterized in that this method includes:
The historical behavior data for obtaining multiple users in the preset time period, obtain the first training after being arranged according to pre-defined rule
Sample;
Using the historical behavior data after arrangement, in the historical behavior data that each user is calculated based on cosine similarity method
Each corresponding impact factor of commodity is as the second training sample;
First training sample and the second training sample are trained as the training sample of deep learning model, trained
Deep learning model;
The first training sample of pre- recommended user and the second training sample are input in the deep learning model trained, exported
The interested items list of user that model prediction goes out recommends pre- recommended user according to the commodity sequence in items list.
2. the method as described in claim 1, which is characterized in that it is described arranged according to pre-defined rule after obtain the first training sample
The step of include:
The historical behavior data of multiple users in preset time period based on acquisition, filter out the interbehavior number of particular category
According to;
Convert the information in the interbehavior data of particular category to the storage of unique number form;
The interbehavior data of each user in the interbehavior data of particular category are sorted in chronological order.
3. the method as described in claim 1, which is characterized in that described to calculate each user's based on cosine similarity method
The specific steps of the corresponding impact factor of each commodity include in historical behavior data:
It is calculated after the N-dimensional vector of particular commodity based on cosine similarity method and maximum value normalization is carried out to the vector, made
For each numerical value between [0,1], N-dimensional vector is the corresponding impact factor of commodity in vector, and N is pretreated history row
Pass through the candidate after certain Rules Filterings for the total quantity of commodity in data or in advance from pretreated historical behavior data
Commodity amount, the value Value of i-th of number in the impact factor that the corresponding N-dimensional vector of commodity indicatesiCircular packet
It includes:
Wherein, Count (i-th of commodity of particular commodity &) represent particular commodity and i-th of commodity appear in jointly it is pretreated
Number in historical behavior data, i >=1 and i≤N, Count (particular commodity) represent particular commodity and are arranged according to pre-defined rule
The interaction total degree in historical behavior data afterwards, Count (i-th of commodity) represent i-th of commodity whole according to pre-defined rule
The interaction total degree in historical behavior data after reason, Max (Values), which is represented in the historical behavior data of each user, to be owned
Maximum value in the corresponding impact factor of commodity.
4. the method as described in claim 1, which is characterized in that preferred, the deep learning model includes but not limited to pass
Return neural network model.
5. the method as described in claim 1, which is characterized in that the commodity sequence according in items list recommends pre- push away
User is recommended, including:
Several commercial product recommendings are to pre- before being come according to the sequence of the commercial productainterests degree predicted in items list from high to low
Recommended user.
6. a kind of individual commodity recommendation system, which is characterized in that including:
Acquisition module, the historical behavior data for obtaining multiple users in preset time period, after being arranged according to pre-defined rule
To the first training sample;
Computing module, for using the historical behavior data after arranging, calculating each user's based on cosine similarity method
The corresponding impact factor of each commodity is as the second training sample in historical behavior data;
Training module, for instructing the first training sample and the second training sample as the training sample of deep learning model
Practice, the deep learning model trained;
Recommending module, for the first training sample of pre- recommended user and the second training sample to be input to the depth trained
It practises in model, the interested items list of user that output model predicts, is recommended according to the commodity sequence in items list
Pre- recommended user.
7. system as claimed in claim 6, which is characterized in that the acquisition module, including:
Screening unit filters out particular category for the historical behavior data of multiple users in the preset time period based on acquisition
Interbehavior data;
Conversion unit is converted into the storage of unique number form for the information in the interbehavior data by particular category;
Sequencing unit is used for the interbehavior data of each user in the interbehavior data of particular category in chronological order
Sequence.
8. system as claimed in claim 6, which is characterized in that the computing module, including:
Computing unit carries out most the vector for being calculated after the N-dimensional vector of particular commodity based on cosine similarity method
Big value normalization, makes in vector that for each numerical value between [0,1], N-dimensional vector is the corresponding impact factor of commodity, and N is pre-
The total quantity of commodity or from pretreated historical behavior data pass through certain rule in advance in historical behavior data that treated
Candidate commodity amount after then screening, the value Value of i-th of number in the impact factor that the corresponding N-dimensional vector of commodity indicatesi's
Circular includes:
Wherein, Count (i-th of commodity of particular commodity &) represent particular commodity and i-th of commodity appear in jointly it is pretreated
Number in historical behavior data, i >=1 and i≤N, Count (particular commodity) represent particular commodity and are arranged according to pre-defined rule
The interaction total degree in historical behavior data afterwards, Count (i-th of commodity) represent i-th of commodity whole according to pre-defined rule
The interaction total degree in historical behavior data after reason, Max (Values), which is represented in the historical behavior data of each user, to be owned
Maximum value in the corresponding impact factor of commodity.
9. system as claimed in claim 6, which is characterized in that the deep learning model includes but not limited to recurrent neural net
Network model.
10. system as claimed in claim 6, which is characterized in that the commodity sequence according in items list is recommended pre-
Recommended user, including:
Several commercial product recommendings are to pre- before being come according to the sequence of the commercial productainterests degree predicted in items list from high to low
Recommended user.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810433175.XA CN108629665B (en) | 2018-05-08 | 2018-05-08 | Personalized commodity recommendation method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810433175.XA CN108629665B (en) | 2018-05-08 | 2018-05-08 | Personalized commodity recommendation method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108629665A true CN108629665A (en) | 2018-10-09 |
CN108629665B CN108629665B (en) | 2021-07-16 |
Family
ID=63695947
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810433175.XA Active CN108629665B (en) | 2018-05-08 | 2018-05-08 | Personalized commodity recommendation method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108629665B (en) |
Cited By (25)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109242654A (en) * | 2018-11-15 | 2019-01-18 | 中国联合网络通信集团有限公司 | A kind of item recommendation method and system |
CN109446430A (en) * | 2018-11-29 | 2019-03-08 | 西安电子科技大学 | Method, apparatus, computer equipment and the readable storage medium storing program for executing of Products Show |
CN109614544A (en) * | 2018-10-30 | 2019-04-12 | 北京奇虎科技有限公司 | A kind of prediction technique and device of the personal information of user |
CN109829761A (en) * | 2019-01-31 | 2019-05-31 | 广州视源电子科技股份有限公司 | A kind of commodity selection method, apparatus, equipment and storage medium |
CN109919737A (en) * | 2019-03-20 | 2019-06-21 | 中电科大数据研究院有限公司 | A kind of recommended method and system of production and sales commodity |
CN110070148A (en) * | 2019-03-15 | 2019-07-30 | 北京木业邦科技有限公司 | A kind of forestry products characteristic analysis method, device and computer-readable medium |
CN110232153A (en) * | 2019-05-29 | 2019-09-13 | 华南理工大学 | A kind of cross-cutting recommended method based on content |
CN110442810A (en) * | 2019-08-08 | 2019-11-12 | 广州华建工智慧科技有限公司 | A kind of mobile terminal BIM model intelligent buffer method based on DeepFM proposed algorithm |
CN110555753A (en) * | 2019-08-14 | 2019-12-10 | 中国平安人寿保险股份有限公司 | recommendation-based ranking control method and device, computer equipment and storage medium |
CN110851718A (en) * | 2019-11-11 | 2020-02-28 | 重庆邮电大学 | Movie recommendation method based on long-time memory network and user comments |
CN111126110A (en) * | 2018-10-31 | 2020-05-08 | 杭州海康威视数字技术股份有限公司 | Commodity information identification method, settlement method and device and unmanned retail system |
CN111325572A (en) * | 2020-01-21 | 2020-06-23 | 深圳前海微众银行股份有限公司 | Data processing method and device |
CN111340115A (en) * | 2020-02-27 | 2020-06-26 | 深圳创新奇智科技有限公司 | Recommendation model training method and system based on characteristic dynamic structure |
CN111612588A (en) * | 2020-05-27 | 2020-09-01 | 腾讯科技(深圳)有限公司 | Commodity presentation method and device, computing equipment and computer-readable storage medium |
CN111639989A (en) * | 2020-04-28 | 2020-09-08 | 上海风秩科技有限公司 | Commodity recommendation method and readable storage medium |
CN111695023A (en) * | 2019-03-11 | 2020-09-22 | 北京京东尚科信息技术有限公司 | Information recommendation method and device, storage medium and equipment |
CN111899061A (en) * | 2020-03-10 | 2020-11-06 | 北京畅行信息技术有限公司 | Order recommendation method, device, equipment and storage medium |
CN111951051A (en) * | 2020-08-14 | 2020-11-17 | 中国工商银行股份有限公司 | Method, device and system for recommending products to customers |
CN112329172A (en) * | 2020-11-05 | 2021-02-05 | 上海交通大学 | Shield tunneling machine cutter head torque prediction method and system based on parallel neural network |
CN112330362A (en) * | 2020-11-04 | 2021-02-05 | 江苏瑞祥科技集团有限公司 | Rapid data intelligent analysis method for internet mall user behavior habits |
CN113360773A (en) * | 2021-07-07 | 2021-09-07 | 脸萌有限公司 | Recommendation method and device, storage medium and electronic equipment |
CN113495966A (en) * | 2020-03-18 | 2021-10-12 | 北京达佳互联信息技术有限公司 | Determination method and device of interactive operation information and recommendation system of video |
CN113538079A (en) * | 2020-04-17 | 2021-10-22 | 北京金山数字娱乐科技有限公司 | Recommendation model training method and device, and recommendation method and device |
CN113705792A (en) * | 2021-08-31 | 2021-11-26 | 平安银行股份有限公司 | Personalized recommendation method, device, equipment and medium based on deep learning model |
CN114820142A (en) * | 2022-06-29 | 2022-07-29 | 国能(北京)商务网络有限公司 | Commodity information recommendation method facing to B-end purchasing user |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103700005A (en) * | 2013-12-17 | 2014-04-02 | 南京信息工程大学 | Association-rule recommending method based on self-adaptive multiple minimum supports |
US20150262282A1 (en) * | 2012-10-05 | 2015-09-17 | Tastebud Technologies, Inc. | Computer-implemented method and system for recommendation system input management |
CN106294758A (en) * | 2016-09-23 | 2017-01-04 | 华南师范大学 | Collaborative recommendation method based on the change of user cognition degree |
CN106846082A (en) * | 2016-12-10 | 2017-06-13 | 江苏途致信息科技有限公司 | Tourism cold start-up consumer products commending system and method based on hardware information |
CN107256494A (en) * | 2017-05-18 | 2017-10-17 | 深圳大学 | A kind of item recommendation method and device |
-
2018
- 2018-05-08 CN CN201810433175.XA patent/CN108629665B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20150262282A1 (en) * | 2012-10-05 | 2015-09-17 | Tastebud Technologies, Inc. | Computer-implemented method and system for recommendation system input management |
CN103700005A (en) * | 2013-12-17 | 2014-04-02 | 南京信息工程大学 | Association-rule recommending method based on self-adaptive multiple minimum supports |
CN106294758A (en) * | 2016-09-23 | 2017-01-04 | 华南师范大学 | Collaborative recommendation method based on the change of user cognition degree |
CN106846082A (en) * | 2016-12-10 | 2017-06-13 | 江苏途致信息科技有限公司 | Tourism cold start-up consumer products commending system and method based on hardware information |
CN107256494A (en) * | 2017-05-18 | 2017-10-17 | 深圳大学 | A kind of item recommendation method and device |
Cited By (35)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109614544A (en) * | 2018-10-30 | 2019-04-12 | 北京奇虎科技有限公司 | A kind of prediction technique and device of the personal information of user |
CN109614544B (en) * | 2018-10-30 | 2023-11-03 | 北京奇虎科技有限公司 | Method and device for predicting personal information of user |
CN111126110B (en) * | 2018-10-31 | 2024-01-05 | 杭州海康威视数字技术股份有限公司 | Commodity information identification method, settlement method, device and unmanned retail system |
CN111126110A (en) * | 2018-10-31 | 2020-05-08 | 杭州海康威视数字技术股份有限公司 | Commodity information identification method, settlement method and device and unmanned retail system |
CN109242654A (en) * | 2018-11-15 | 2019-01-18 | 中国联合网络通信集团有限公司 | A kind of item recommendation method and system |
CN109446430A (en) * | 2018-11-29 | 2019-03-08 | 西安电子科技大学 | Method, apparatus, computer equipment and the readable storage medium storing program for executing of Products Show |
CN109829761A (en) * | 2019-01-31 | 2019-05-31 | 广州视源电子科技股份有限公司 | A kind of commodity selection method, apparatus, equipment and storage medium |
CN111695023A (en) * | 2019-03-11 | 2020-09-22 | 北京京东尚科信息技术有限公司 | Information recommendation method and device, storage medium and equipment |
CN110070148A (en) * | 2019-03-15 | 2019-07-30 | 北京木业邦科技有限公司 | A kind of forestry products characteristic analysis method, device and computer-readable medium |
CN109919737A (en) * | 2019-03-20 | 2019-06-21 | 中电科大数据研究院有限公司 | A kind of recommended method and system of production and sales commodity |
CN110232153A (en) * | 2019-05-29 | 2019-09-13 | 华南理工大学 | A kind of cross-cutting recommended method based on content |
CN110442810A (en) * | 2019-08-08 | 2019-11-12 | 广州华建工智慧科技有限公司 | A kind of mobile terminal BIM model intelligent buffer method based on DeepFM proposed algorithm |
CN110555753A (en) * | 2019-08-14 | 2019-12-10 | 中国平安人寿保险股份有限公司 | recommendation-based ranking control method and device, computer equipment and storage medium |
CN110851718B (en) * | 2019-11-11 | 2022-06-28 | 重庆邮电大学 | Movie recommendation method based on long and short term memory network and user comments |
CN110851718A (en) * | 2019-11-11 | 2020-02-28 | 重庆邮电大学 | Movie recommendation method based on long-time memory network and user comments |
CN111325572A (en) * | 2020-01-21 | 2020-06-23 | 深圳前海微众银行股份有限公司 | Data processing method and device |
CN111340115A (en) * | 2020-02-27 | 2020-06-26 | 深圳创新奇智科技有限公司 | Recommendation model training method and system based on characteristic dynamic structure |
CN111340115B (en) * | 2020-02-27 | 2023-05-12 | 青岛创新奇智科技集团股份有限公司 | Recommendation model training method and system based on feature dynamic construction |
CN111899061B (en) * | 2020-03-10 | 2024-04-16 | 北京畅行信息技术有限公司 | Order recommendation method, device, equipment and storage medium |
CN111899061A (en) * | 2020-03-10 | 2020-11-06 | 北京畅行信息技术有限公司 | Order recommendation method, device, equipment and storage medium |
CN113495966A (en) * | 2020-03-18 | 2021-10-12 | 北京达佳互联信息技术有限公司 | Determination method and device of interactive operation information and recommendation system of video |
CN113495966B (en) * | 2020-03-18 | 2023-06-23 | 北京达佳互联信息技术有限公司 | Interactive operation information determining method and device and video recommendation system |
CN113538079A (en) * | 2020-04-17 | 2021-10-22 | 北京金山数字娱乐科技有限公司 | Recommendation model training method and device, and recommendation method and device |
CN111639989B (en) * | 2020-04-28 | 2024-02-02 | 上海秒针网络科技有限公司 | Commodity recommendation method and readable storage medium |
CN111639989A (en) * | 2020-04-28 | 2020-09-08 | 上海风秩科技有限公司 | Commodity recommendation method and readable storage medium |
CN111612588B (en) * | 2020-05-27 | 2023-12-26 | 腾讯科技(深圳)有限公司 | Commodity presenting method and device, computing equipment and computer readable storage medium |
CN111612588A (en) * | 2020-05-27 | 2020-09-01 | 腾讯科技(深圳)有限公司 | Commodity presentation method and device, computing equipment and computer-readable storage medium |
CN111951051B (en) * | 2020-08-14 | 2023-09-26 | 中国工商银行股份有限公司 | Method, device and system for recommending products to clients |
CN111951051A (en) * | 2020-08-14 | 2020-11-17 | 中国工商银行股份有限公司 | Method, device and system for recommending products to customers |
CN112330362A (en) * | 2020-11-04 | 2021-02-05 | 江苏瑞祥科技集团有限公司 | Rapid data intelligent analysis method for internet mall user behavior habits |
CN112329172B (en) * | 2020-11-05 | 2022-12-27 | 上海交通大学 | Shield tunneling machine cutter head torque prediction method and system based on parallel neural network |
CN112329172A (en) * | 2020-11-05 | 2021-02-05 | 上海交通大学 | Shield tunneling machine cutter head torque prediction method and system based on parallel neural network |
CN113360773A (en) * | 2021-07-07 | 2021-09-07 | 脸萌有限公司 | Recommendation method and device, storage medium and electronic equipment |
CN113705792A (en) * | 2021-08-31 | 2021-11-26 | 平安银行股份有限公司 | Personalized recommendation method, device, equipment and medium based on deep learning model |
CN114820142A (en) * | 2022-06-29 | 2022-07-29 | 国能(北京)商务网络有限公司 | Commodity information recommendation method facing to B-end purchasing user |
Also Published As
Publication number | Publication date |
---|---|
CN108629665B (en) | 2021-07-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108629665A (en) | A kind of individual commodity recommendation method and system | |
CN104866474B (en) | Individuation data searching method and device | |
Sivapalan et al. | Recommender systems in e-commerce | |
CN107633430A (en) | A kind of Method of Commodity Recommendation based on community of colony | |
CN103914492B (en) | Query word fusion method, merchandise news dissemination method and searching method and system | |
Wu et al. | Improving performance of tensor-based context-aware recommenders using bias tensor factorization with context feature auto-encoding | |
CN103823908A (en) | Method and server for content recommendation on basis of user preferences | |
CN110175895B (en) | Article recommendation method and device | |
CN102411754A (en) | Personalized recommendation method based on commodity property entropy | |
CN107391582B (en) | The information recommendation method of user preference similarity is calculated based on context ontology tree | |
CN104615779A (en) | Method for personalized recommendation of Web text | |
CN106168980A (en) | Multimedia resource recommends sort method and device | |
Zhang et al. | Multimodal marketing intent analysis for effective targeted advertising | |
CN108805598A (en) | Similarity information determines method, server and computer readable storage medium | |
CN110348930A (en) | Business object data processing method, the recommended method of business object information and device | |
CN107229666A (en) | A kind of interest heuristic approach and device based on commending system | |
US20180139296A1 (en) | Method of producing browsing attributes of users, and non-transitory computer-readable storage medium | |
CN104239335B (en) | User-specific information acquisition methods and device | |
CN110197404A (en) | The personalized long-tail Method of Commodity Recommendation and system of popularity deviation can be reduced | |
CN113689259A (en) | Commodity personalized recommendation method and system based on user behaviors | |
CN116894709A (en) | Advertisement commodity recommendation method and system | |
CN110555719B (en) | Commodity click rate prediction method based on deep learning | |
Konapure et al. | Video content-based advertisement recommendation system using classification technique of machine learning | |
CN109146606A (en) | A kind of brand recommended method, electronic equipment, storage medium and system | |
Kayande et al. | Recommendation systems for community commerce |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |