CN108920596A - A kind of personalized recommendation algorithm and terminal - Google Patents
A kind of personalized recommendation algorithm and terminal Download PDFInfo
- Publication number
- CN108920596A CN108920596A CN201810678309.4A CN201810678309A CN108920596A CN 108920596 A CN108920596 A CN 108920596A CN 201810678309 A CN201810678309 A CN 201810678309A CN 108920596 A CN108920596 A CN 108920596A
- Authority
- CN
- China
- Prior art keywords
- product
- data
- user
- recommended products
- evaluation data
- 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
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 embodiment of the invention discloses a kind of personalized recommendation algorithm and terminal, wherein method includes：Obtain user tag data；The initial value of product is determined according to the user tag data, and picks out pre- recommended products from the product that initial value is higher than the first preset fraction；Quantified to determine pre- recommended products recommender score respectively according to the first evaluation dimension and the second evaluation dimension；Pre- recommended products using recommender score higher than the second preset fraction is as consequently recommended Products Show to user.The embodiment of the present invention carries out comprehensive normalization to pre- recommended products from different evaluation dimensions and scores, and by the highest Products Show of recommender score to user, guarantees that recommended product meets user demand, to improve business profession degree and user satisfaction.
Description
Technical field
The present invention relates to computer application technology more particularly to a kind of personalized recommendation algorithm and terminals.
Background technique
In the information age, targetedly recommend that its is interested, meets the product of demand to user, for enterprise and use
The value at family is self-evident.More than 60% Netflix user be found by the recommendation of system oneself interested video and
Film, and music radio station Pandora is then by providing the user with feedback system --- like, do not like and skip to obtain user
Interest model, recommend the single-row table of relevant song to user in conjunction with the history interbehavior of user, make every effort to meet user
Preference.
System recommended to realize mainly by recommended method to user, and recommended method is most core in entire recommender system
The part of the heart, most critical has been largely fixed the superiority and inferiority of recommender system performance.Currently, recommended method mainly includes：It is based on
Commending contents, collaborative filtering recommending are recommended based on correlation rule, based on effectiveness recommendation, knowledge based recommendation and combined recommendation.Though
So above-mentioned recommended method has been widely applied, but is still faced with many problems, such as personalized recommendation degree is low, recommendation spirit
Poor activity and how new user is recommended and how to recommend new product to user's (cold start-up problem) etc..
Summary of the invention
The embodiment of the present invention provides a kind of personalized recommendation algorithm and terminal, with help user in the information of magnanimity quickly
It was found that really necessary product, improves user's stickiness.
In a first aspect, the embodiment of the invention provides a kind of personalized recommendation algorithm, this method includes：
Obtain user tag data；
The initial value of product is determined according to the user tag data, and is higher than the first preset fraction from initial value
Pre- recommended products is picked out in product；
The first fractional value of the pre- recommended products is determined according to the first evaluation dimension, wherein first evaluation dimension
Urgent significance level, enterprise value and three aspects of user experience including product；
The second fractional value of the pre- recommended products is determined according to the second evaluation dimension, wherein second evaluation dimension
For user internet behavioral data；
Determine that the 4th fractional value of the pre- recommended products, the 4th evaluation dimension are product according to the 4th evaluation dimension
Inventory；
The recommended hour of the pre- recommended products is determined according to the first fractional value of the pre- recommended products, the second fractional value
Number；
The pre- recommended products is ranked up from high to low by recommender score, obtains sorted lists；
The pre- recommended products that preset quantity is filtered out from sorted lists recommends user as recommended products.
Second aspect, the embodiment of the invention provides a kind of terminal, which includes:For executing as described in relation to the first aspect
Method unit.
The third aspect, the embodiment of the invention provides another terminal, which includes processor, input equipment, output
Equipment and memory, the processor, input equipment, output equipment and memory are connected with each other, wherein the memory is used for
Storage supports terminal to execute the application code of the above method, and the processor is configured for executing above-mentioned first aspect
Method.
Fourth aspect, the embodiment of the invention provides a kind of computer readable storage medium, the computer storage medium
It is stored with computer program, the computer program includes program instruction, and described program instruction makes institute when being executed by a processor
State the method that processor executes above-mentioned first aspect.
Compared with prior art, beneficial effects of the present invention include：
Three urgency level, enterprise value and user experience aspects of algorithm fusion product of the invention, recommend pre-
Product, which is done, to be considered in all directions and quantifies to give a mark, and the product for being most suitable for recommending user is chosen；
The behavioral data of algorithm fusion user online of the invention, tracks the behavioral data and current row of user's history
For data, depth analysis user is to the preference of product, so as to according to the preference real-time update recommendation results of user；
Algorithm of the invention considers the factor that consumer products order history and product inventory, and recommendation results is made to enable user full
While meaning, it is ensured that user can order in time；
The embodiment of the present invention goes out pre- recommended products according to the note data preliminary screening of user, and from different evaluation dimensions
Comprehensive normalization scoring is carried out to pre- recommended products, by the highest Products Show of recommender score to user, guarantee is recommended
Product meet user demand, to improve business profession degree and user satisfaction.
Detailed description of the invention
Technical solution in order to illustrate the embodiments of the present invention more clearly, below will be to needed in embodiment description
Attached drawing is briefly described, it should be apparent that, drawings in the following description are some embodiments of the invention, general for this field
For logical technical staff, without creative efforts, it is also possible to obtain other drawings based on these drawings.
Fig. 1 is a kind of schematic flow diagram of personalized recommendation algorithm provided in an embodiment of the present invention；
Fig. 2 is the concrete methods of realizing schematic flow diagram for the step S102 that one embodiment of the invention provides；
Fig. 3 is the concrete methods of realizing schematic flow diagram for the step S103 that one embodiment of the invention provides；
Fig. 4 is the concrete methods of realizing schematic flow diagram for the step S104 that one embodiment of the invention provides；
Fig. 5 be another embodiment of the present invention provides a kind of terminal schematic block diagram.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are some of the embodiments of the present invention, instead of all the embodiments.Based on this hair
Embodiment in bright, every other implementation obtained by those of ordinary skill in the art without making creative efforts
Example, shall fall within the protection scope of the present invention.
It should be appreciated that ought use in this specification and in the appended claims, term " includes " and "comprising" instruction
Described feature, entirety, step, operation, the presence of element and/or component, but one or more of the other feature, whole is not precluded
Body, step, operation, the presence or addition of element, component and/or its set.
It is also understood that mesh of the term used in this description of the invention merely for the sake of description specific embodiment
And be not intended to limit the present invention.As description of the invention and it is used in the attached claims, unless on
Other situations are hereafter clearly indicated, otherwise " one " of singular, "one" and "the" are intended to include plural form.
It will be further appreciated that the term "and/or" used in description of the invention and the appended claims is
Refer to any combination and all possible combinations of one or more of associated item listed, and including these combinations.
Personalized recommendation is the difference according to user interest and behavioral characteristic, recommends it respectively required to different users
Information or commodity, to promote Information or merchandise sales.
It is a kind of schematic flow diagram of personalized recommendation algorithm provided in an embodiment of the present invention, as shown in the figure referring to Fig. 1
This method may include step S101-S107：
S101 obtains user tag data.
Third party's data relevant to business are determined outside the data and joint retained in business system according to user
Property analysis go forward side by side row label management, to understand the essential attribute of user, Behavior preference, consumption habit, business service condition
Etc. information.
The label data of user mainly includes user information and socialization relationship.User information can be by user in enterprise
The data retained in system obtain, and mainly include the data such as gender, age, occupation, income and place city.Socialization relationship
Data can then be obtained by introducing relevant to business third party's data, the third party's data that can be introduced mainly include consumption partially
Good, social data, Unionpay's data, credit data etc..
S102 determines the initial value of product according to user tag data, and is higher than the first preset fraction from initial value
Product in pick out pre- recommended products.
In specific implementation, by the label data of user can generally understand the Behavior preference of user, consumption habit,
The essential informations such as business service condition do quantitative analysis scoring to the product of magnanimity according to these essential informations, pre- with first
If score does screening conditions, the product of limited quantity is filtered out from the product of magnanimity as pre- recommended products, so that it is guaranteed that in advance
Recommended products substantially conforms to user interest and preference.
First preset fraction can be by technical staff according to the concrete condition sets itself in specific embodiment, the present invention couple
This is without limitation.
S103 determines the first fractional value of pre- recommended products according to the first evaluation dimension.
The first evaluation dimension includes the urgent significance level of product, enterprise's valence in some embodiments, such as the present embodiment
Three aspects of value and user experience.
In specific implementation, after filtering out pre- recommended products, by quantify the urgent significance level of product, enterprise value with
And user experience determines the first fractional value of pre- recommended products, to further choose the product for being most suitable for recommending user, really
The accuracy recommended is protected, usage rate of the user is increased.
The urgent significance level of product mainly includes four degree：It is important and urgent, important it is not urgent, it is urgent it is inessential with
And it is not urgent inessential.The urgent significance level of product shows user to the market demand of product, for important and urgent
Product, the market demand is big, can promote the raising of pre- the first fractional value of recommended products, to promote the recommendation of product；For not tight
Anxious unessential product, shows that the market demand is small, has detrimental effect to the raising of the first fractional value, influence the recommendation of product.
Enterprise value refers to product to enterprise's bring value and profit.The big product of enterprise value shows profit height, can
The raising for promoting pre- the first fractional value of recommended products, to promote the recommendation of product；The small product of enterprise value shows that profit is low,
There is detrimental effect to the raising of the first fractional value, influences the recommendation of product.
User experience is user to the whole cognition impression and reaction for using or it is expected to use product.The good production of user experience
Product can promote the raising of pre- the first fractional value of recommended products, to promote the recommendation of product；The product of poor user experience is to first
The raising of fractional value has detrimental effect, influences the recommendation of product.
The determination of pre- the first fractional value of recommended products should be from the angle of the market demand, it is also desirable to consider that product is brought
Value and user experience situation；It is obtained at three the urgent significance level of product, enterprise value and user experience aspects
Balance, it is ensured that the accuracy of recommendation promotes business profession degree and increases usage rate of the user.
S104 determines the second fractional value of pre- recommended products according to the second evaluation dimension.
In specific implementation, the second evaluation dimension is specially the internet behavioral data of user.To the internet row of user
It is converted into quantifiable normalization operation for data, to obtain the second fractional value of pre- recommended products.
It should be noted that the behavior of the internet each time event of user all can be considered an internet behavioral data, nothing
Essential requirement of user's heart, including page browsing, click, collection, shopping, search, marking, comment etc. are not reflected.Therefore,
The internet behavioral data of tracking user can react user to the preference of product, it is easier to meet the demand of user instantly, from
And more accurately to user's recommended products.
In addition, it is contemplated that user carried out the behavior recommended or ordered to product, this is than the behavioral datas such as clicking, browsing
The preference of user can more be reacted.When operation being normalized determining the second fractional value of pre- recommended products, to assigning user
Number and the preset weighted value of product subscription history is recommended to want higher.
However, being easy to be influenced by the time, i.e. user internet behavioral data when tracking the behavioral data of user
Valuable extent value is decayed at any time, and behavioral data time of origin is closer away from the current time, and obtained data can more characterize
The behavior in user future.The taste of user can change over time after all, so the time is closer to pre- recommendation production
Second fractional value of product influences bigger.
S105 determines the recommender score of pre- recommended products according to the first fractional value of pre- recommended products, the second fractional value.
In specific implementation, sum the first fractional value and the second fractional value with the recommended hour of the pre- recommended products of determination
Number.The first fractional value, the second fractional value of pre- recommended products indicate pre- recommended products respectively in the score of product and user level.
Recommender score is to carry out comprehensive normalization scoring to pre- recommended products from different evaluation dimensions, around the preference of user,
The factor for further considering product and enterprise, precisely recommends user to realize.
Pre- recommended products is ranked up by S106 from high to low by recommender score, obtains sorted lists.
Recommender score has successively reacted the matching degree with user preference from high to low.Recommender score is higher, more meets use
The preference at family, the more suitable recommendation that product is carried out to user；Recommender score is lower, more deviation user preference, is not suitable for user
Carry out Products Show.
S107, the pre- recommended products that preset quantity is filtered out from sorted lists recommend user as recommended products.
In specific implementation, the Products Show of the highest preceding n quantity of score is filtered out from sorted lists to user.N table
Show preset quantity, for example, n can be 1 or 2 or 3, specific value can be determined by technical staff according to business needs, the present invention
It is not specifically limited in this embodiment.
In some embodiments, such as the present embodiment, step S107 specific implementation be can be：According in sorted lists
Recommender score dynamically preset the second score, judge whether the recommender score of pre- recommended products is higher than the second preset fraction；If
The recommender score of pre- recommended products is higher than the second preset fraction, then recommends user using pre- recommended products as recommended products；If
The recommender score of pre- recommended products is not higher than the second preset fraction, then it represents that pre- recommended products does not meet the preference of user, to protect
Card recommends profession degree and user satisfaction, this pre- recommended products cannot be recommended user.
It should be noted that the second preset fraction of setting can quantitatively filter out the Products Show of preset quantity to user.
The concrete condition sets itself that second preset fraction can be needed by those skilled in the art according to business, the present invention, which does not do this, to be had
Body limits.
Three urgency level of the algorithm fusion product of the present embodiment, enterprise value and user experience aspects, to pushing away in advance
It recommends product and does equilibrium and consider；
The behavioral data of the algorithm fusion user online of the present embodiment, is analyzed according to the characteristics of user internet behavioral data
Preference of the user to product；
The present embodiment goes out pre- recommended products according to the note data preliminary screening of user, and from product and two kinds of user behavior
Different evaluation dimensions carries out comprehensive normalization to pre- recommended products and scores, by the highest Products Show of recommender score to use
Family guarantees that recommended product meets user demand, to improve business profession degree and user satisfaction.
In some embodiments, such as the present embodiment, when calculating the recommender score of pre- recommended products, by the inventory of product
Normalization operation is included in account for.It is also needed by factors such as shelf life of products and promotions when recommending to user
It to consider the inventory of pre- recommended products, its out of date or not in stock product can be replaced by other approximate products and be pushed away
It recommends, it is ensured that user orders in time.
It should be noted that it is true to be referred to existing related data according to the method that operation is normalized in the inventory of product
Fixed, the present invention does not do specific restriction to this.The factor of combination product inventory accounts for the recommendation to product, can effectively keep away
The case where exempting from product supply shortage or expired equal reduction user satisfaction appearance.
Referring to fig. 2, the schematic flow diagram of the concrete methods of realizing of the step S102 provided for one embodiment of the invention, such as
Shown in figure, this approach includes the following steps S201-S208.
S201 determines the first score of product according to TOP-N algorithm.
In specific implementation, descending arrangement is carried out using first score of the TOP-N algorithm to product, from the product of magnanimity
Choose the product of the first score top n quantity.The first score that product is indicated with Q1, the N number of quantity obtained using TOP-N algorithm
The first score of product be：Q11, Q12..., Q1N。
It should be noted that being referred to existing related money according to the method that TOP-N algorithm determines the first score of product
Material determines that the present invention does not do specific restriction to this.
S202 determines the second score of product according to association rule algorithm.
The second score of product of the second score that product is indicated with Q2, the N number of quantity obtained using association rule algorithm is：
Q21, Q22..., Q2N。
In specific implementation, the product of the second score top n quantity is determined according to association rule algorithm, specific algorithm can be with
It is determined referring to existing related data, the present invention does not do specific restriction to this.
S203 determines the third score of product according to collaborative filtering.
The product third score of the third score that product is indicated with Q3, the N number of quantity obtained using collaborative filtering is：
Q31, Q32..., Q3N。
In specific implementation, the product of third score top n quantity is determined according to collaborative filtering, specific algorithm can be with
It is determined referring to existing related data, the present invention does not do specific restriction to this.
S204, according to the first score, the second score and third score and in advance for TOP algorithm, association rule algorithm and
The weighted value of collaborative filtering setting determines the initial value of product.
In specific implementation, the weighted value W1 of TOP algorithm, the weighted value W2 of association rule algorithm are preset and was cooperateed with
The weighted value W3 of algorithm is filtered, and has W1+W2+W3=1.
The initial value of product is determined by following formula
P=W1*Q1+W2*Q2+W3*Q3
Wherein, P indicates product initial value, Q1 indicates the first score of product, and Q2 indicates the second score of product, Q3
Indicate the third score of product.
S205 picks out product alternately product of the initial value higher than the first preset fraction.
In specific implementation, initial value P is higher than to the product alternately product of the first preset fraction.
Such as in the present embodiment, there is the initial value of f product to be higher than the first preset fraction, then share the standby of f quantity
Select product and corresponding initial value：P1,P2,…,Pf, and P1≥P2≥…≥Pf；Wherein PfFor the initial of f-th alternate product
Score value.
It should be noted that the first preset fraction can by technical staff according to the concrete condition in specific embodiment voluntarily
Setting, which is not limited by the present invention.
S206 judges whether the quantity of alternate product is more than preset amount threshold.
It is being embodied, such as in the present embodiment, preset amount threshold is k, and the quantity of alternate product is f.
S207 is produced from alternative if the quantity of alternate product is more than preset amount threshold by initial value from high to low
It is the product of amount threshold as pre- recommended products that quantity is selected in product.
In specific implementation, if f >=k, there is the pre- recommended products of k number amount：P1,P2,…,Pk, and P1≥P2≥…≥
Pk, wherein PkFor the initial value of k-th of pre- recommended products.
S208, if the quantity of alternate product is less than preset amount threshold, using alternate product as pre- recommended products.
In specific implementation, if f<K then has the pre- recommended products of f quantity：P1,P2,…,Pf, and P1≥P2≥…≥
Pf；Wherein PfFor the initial value of f-th of pre- recommended products.
Referring to Fig. 3, it is the schematic flow diagram of the concrete methods of realizing for the step S103 that one embodiment of the invention provides, such as schemes
It is shown, this approach includes the following steps S301-S302：
S301, receiving pre- recommended products respectively, these three are commented in the urgency level of product, enterprise value and user experience
Score on valence latitude.
In specific implementation, by senior user respectively to pre- recommended products the urgency level of product, enterprise value and
These three evaluation dimensions of user experience are given a mark.
For example, can be by senior user voluntarily to pre- recommended products in the urgency level of product, enterprise value and user's body
It tests three aspects and inputs score.
S302, according to pre- recommended products in the score of the urgency level of product, enterprise value and user experience and in advance
The first fractional value H of pre- recommended products is determined for the weighted value of the urgency level of product, enterprise value and user experience setting；
Wherein, H indicates the first fractional value of pre- recommended products.
In one embodiment, senior user is to pre- recommended products in the urgency level of product, enterprise value and use
The score of family experience input is respectively (full marks 100 divide)：80,95,80；The urgency level weighted value of product is 0.4, enterprise value
Weighted value is 0.3, user experience weighted value is 0.3；Then there is the first fractional value H=80*0.4+95*0.3+ of the pre- recommended products
80*0.3=84.5.
It should be noted that the urgency level weighted value of product, enterprise value weighted value and user experience weighted value can
By technical staff according to the concrete condition sets itself in specific embodiment, which is not limited by the present invention.
Referring to fig. 4, the schematic flow diagram of the concrete methods of realizing of the step S104 provided for one embodiment of the invention, such as
Shown in figure, this approach includes the following steps S401-S405：
S401 picks out the internet row of preset quantity from user in the internet behavioral data in default measurement period
It is data as evaluation data.
In specific implementation, the behavioral data for tracking user can react user to the preference of product, it is easier to meet user
Instantly demand, thus more accurately to user's recommended products.In terms of the excitation to user preference, the internet row of user
Positive feedback behavior data and negative-feedback behavioral data can be divided into for data.Positive feedback behavior data refer to can forward reaction use
Family meets user demand to the preference of product, promotes the behavioral data of product sale, such as browsing recommendation data, repeatedly
Click bubble data, user buys data etc.；Conversely, negative-feedback behavioral data is then preference of the back reaction user to product,
Deviate user demand, be easily reduced the behavioral data of customer satisfaction, such as closes recommending data, prompting next time data, user
Complain data and no-operand according to etc..Since the internet behavioral data of user is more, the present invention is not specifically limited in this embodiment.
In addition, the behavioral data of user, and be easy to be influenced by the time, i.e., user internet behavioral data is valuable
Degree decays at any time, and the time interval that internet behavioral data the occurs current time is closer, can more characterize user in the future
Behavior, value degree it is higher；Conversely, internet behavioral data occur time interval it is current time it is remoter, be more difficult to characterize
The behavior in user future, value degree are lower.
It is being embodied, such as in the present embodiment, the evaluation data of user include closing recommending data, prompting next time number
According to, customer complaint data, browsing recommendation data, repeatedly click number of bubbles accordingly and no-operand according to, user to product
The subscription data etc. of recommending data and user to product.
The true preference of user and demand can be comprehensively reacted in conjunction with the internet behavioral data of user, due to can be used as
The user internet behavioral data for evaluating data is more, and which is not limited by the present invention.
S402 determines history decay factor and evaluation data of the evaluation data in measurement period in the current statistic time
The current attenuation factor.
The valuable extent value of evaluation data is decayed at any time, and the time interval that evaluation data the occur current time gets over
Closely, the behavior in user's future can be more characterized, value degree is higher；Conversely, time interval current time that evaluation data occur gets over
Far, it is more difficult to characterize the behavior in user's future, value degree is lower.
In specific implementation, history decay factor indicates attenuation degree of the evaluation data in measurement period at any time；When
Preceding decay factor indicates attenuation degree of the evaluation data in the current statistic time at any time.
In specific implementation, history decay factor and the current attenuation factor pass through following formula
It determines, wherein t indicates date, fi(t) decay factor, a evaluated data i on the t date is indicatediIndicate evaluation data
Degree that i decays at any time, C indicate that constant, current_date indicate the date of current statistic time, hist_recom_
Date indicates that evaluation data i indicates evaluation data i most by the date of user's execution, last_recom_date in measurement period
The date closely once recommended.
S403 determines review number according to the history decay factor of evaluation data and the preparatory weighted value for evaluation data setting
According to the historical scores in measurement period.
In specific implementation, pass through following formula
Determine historical scores of the evaluation data in default measurement period, wherein S indicates the evaluation data of pre- recommended products
Historical scores, S in default measurement periodiIndicate historical scores, W of the evaluation data i in default measurement periodiIndicate pre-
First indicate that measurement period, n indicate the evaluation data of n quantity for weighted value, the T of evaluation data i setting.
S404 determines review number according to the current attenuation factor of evaluation data and the preparatory weighted value for evaluation data setting
According to the RealTime scores in the current statistic time.
In specific implementation, pass through following formula
ΔSi=Wifi(current_date)
Evaluation data are determined in the RealTime scores of current statistic time, wherein Δ S indicates the evaluation data of pre- recommended products
In the RealTime scores of current statistic time, Δ SiIndicate the RealTime scores in the current statistic time of evaluation data i.
S405 will evaluate the historical scores of data and the summation of RealTime scores as second point of the pre- recommended products
Numerical value.
In specific implementation, pass through following formula
R=S+ Δ S
Determine the second fractional value of the pre- recommended products, wherein R indicates the second fractional value of pre- recommended products.
Specific implementation
For example, in one embodiment, there is pre- recommended products：Product 1, product 2；The recommended hour of pre- recommended products is calculated below
Number：
1, for convenience of calculating, the user of pre- recommended products might as well be recommended number and the behavioral data of product subscription history are arranged
It is 0, then can not considers that the user of product 1 and product 2 recommends number and product subscription history.
2, for convenience of calculating, 0 might as well be set by the commodity stocks weight of pre- recommended products, then can not considers product 1
With the stock factor of product 2.
3, the first fractional value of product 1 and product 2 calculates as follows：
The score that senior user inputs product 1 and product 2 in the urgency level of product, enterprise value and user experience
Respectively (full marks 100 divide)：
Product 1：90,85,85；
Product 2：80,95,80；
Wherein, set the urgency level weighted value of product as 0.4, enterprise value weighted value be 0.3, user experience weighted value
It is 0.3；
Then there is the first fractional value H1=90*0.4+85*0.3+85*0.3=87 of product 1；
First fractional value H2=80*0.4+95*0.3+80*0.3=84.5 of product 2.
4, the second fractional value of product 1 and product 2 calculates as follows：
In the present embodiment, the decay factor and its weight of the evaluation data of product 1 and product 2 are as shown in table 1 below, product 1
It is as shown in table 2 below in the evaluation data of measurement period and current date with product 2.
The decay factor and its weight table of the evaluation data of table 1
The evaluation data statistic of table 2 product 1 and product 2
It should be noted that the number 1 in table 2, which represents, generates evaluation data, number 0 represents and does not generate evaluation data, right
In the date for not generating evaluation data in measurement period, the not no practical significance of decay factor, therefore do not have to calculate.
Historical scores S1 of the evaluation data of product 1 in measurement period：
S1It closes and recommends=-2*e-4-2*e-1.5
S1Next time reminds=-1*2-0.5
S1Browse recommendation=+2*3-0
S1=S1It closes and recommends+S1Next time reminds+S1Browse recommendation=0.83.
Wherein S1It closes and recommendsIndicate product 1 in the historical scores that evaluation data are that closing is recommended；S1Next time remindsIndicate that product 1 is being commented
Valence mumber is according to the historical scores reminded for next time；S1Browse recommendationIndicate that product 1 is obtained in the history that evaluation data are browsing recommendation
Point.
Historical scores S2 of the evaluation data of product 2 in measurement period：
S2It closes and recommends=-2*e-0.5
S2Next time reminds=-1*2-4
S2Browse recommendation=+2*3-1.5+2*3-0
S2=S2It closes and recommends+S2Next time reminds+S2Browse recommendation=1.10.
Wherein S2It closes and recommendsIndicate product 2 in the historical scores that evaluation data are that closing is recommended；S2Next time remindsIndicate that product 2 is being commented
Valence mumber is according to the historical scores reminded for next time；S2Browse recommendationIndicate that product 2 is obtained in the history that evaluation data are browsing recommendation
Point.
Score of the evaluation data of product 1 and product 2 in current date：
Δ S1=Δ S1Browse recommendation=+2*3-0=2
Δ S2=Δ S2It closes and recommends=-2*e-0=-2
Then there is the second fractional value R1=S1+ Δ S1=0.83+2=2.83 of product 1；
Second fractional value R2=S2+ Δ S2=1.10-2=-0.9 of product 2.
5, the recommender score of product 1 and product 2 is determined according to the first fractional value of product 1 and product 2, the second fractional value：
1 recommender score of product：H1+R1=87+2.83=89.93；
2 recommender score of product：H2+R2=84.5-0.9=85.5.
There is 1 recommender score of product to be greater than 2 recommender score of product, and the recommender score of product 1 is more than 85, therefore by product 1
Preferential recommendation is to user.
It should be noted that 85 be the second preset fraction in the present embodiment, in specific implementation can by technical staff according to
Actual conditions sets itself, which is not limited by the present invention.
Referring to Fig. 5, the embodiment of the present invention provides a kind of terminal 50, which includes for executing described in above embodiments
Method unit.As shown, the terminal 50 in the present embodiment includes receiving unit 51, the determining list of module of selection 52, first
First 53, second determination unit 54, third determination unit 55, sequencing unit 56, recommendation unit 57：
Receiving unit 51, for obtaining user tag data.
Module of selection 52 is higher than for determining the initial value of product according to user tag data, and from initial value
Pre- recommended products is picked out in the product of one preset fraction.
First determination unit 53, for determining the first fractional value of pre- recommended products according to the first evaluation dimension.
Second determination unit 54, for determining the second fractional value of pre- recommended products according to the second evaluation dimension, wherein the
Two evaluation dimensions are user internet behavioral data.
Third determination unit 55 determines that pre- recommendation produces for the first fractional value, the second fractional value according to pre- recommended products
The recommender score of product.
Sequencing unit 56 obtains sorted lists for pre- recommended products to be ranked up from high to low by recommender score.
Recommendation unit 57, the pre- recommended products that preset quantity is filtered out from sorted lists recommend use as recommended products
Family.
The above description is merely a specific embodiment, but scope of protection of the present invention is not limited thereto, any
Those familiar with the art in the technical scope disclosed by the present invention, can readily occur in various equivalent modifications or replace
It changes, these modifications or substitutions should be covered by the protection scope of the present invention.Therefore, protection scope of the present invention should be with right
It is required that protection scope subject to.
Claims (10)
1. a kind of personalized recommendation algorithm, which is characterized in that the method includes：
Obtain user tag data；
The initial value of product is determined according to the user tag data, and is higher than the product of the first preset fraction from initial value
In pick out pre- recommended products；
The first fractional value of the pre- recommended products is determined according to the first evaluation dimension, wherein first evaluation dimension includes
Three urgent significance level, enterprise value and user experience aspects of product；
The second fractional value of the pre- recommended products is determined according to the second evaluation dimension, wherein second evaluation dimension is to use
Family internet behavioral data；
The recommender score of the pre- recommended products is determined according to the first fractional value of the pre- recommended products, the second fractional value；
The pre- recommended products is ranked up from high to low by recommender score, obtains sorted lists；
The pre- recommended products that preset quantity is filtered out from sorted lists recommends user as recommended products.
2. personalized recommendation algorithm according to claim 1, which is characterized in that described true according to the user tag data
The initial value of fixed output quota product, including：
The first score of product is determined according to TOP-N algorithm；
The second score of product is determined according to association rule algorithm；
The third score of product is determined according to collaborative filtering；
It TOP-N algorithm, association rule algorithm and is cooperateed with according to first score, the second score and third score in advance
The weighted value of filter algorithm setting determines the initial value of product.
3. the method according to claim 1, wherein the product for being higher than the first preset fraction from initial value
In pick out pre- recommended products, including：
Pick out product alternately product of the initial value higher than the first preset fraction；
Whether the quantity for judging the alternate product is more than preset amount threshold；
If the quantity of alternate product is more than preset amount threshold, number is selected from alternate product from high to low by initial value
Amount is the product of amount threshold as pre- recommended products；
If the quantity of alternate product is less than preset amount threshold, using alternate product as pre- recommended products.
4. personalized recommendation algorithm according to claim 1, which is characterized in that determine institute according to first evaluation dimension
The first fractional value of pre- recommended products is stated, including：
Pre- recommended products obtaining at the urgent significance level of product, enterprise value and three aspects of user experience is received respectively
Point；
According to pre- recommended products the urgent significance level of the product, enterprise value and user experience score be in advance
Urgent significance level, enterprise value and the weighted value of user experience setting of product determine first point of the pre- recommended products
Numerical value.
5. personalized recommendation algorithm according to claim 1, which is characterized in that described to determine institute according to the second evaluation dimension
The second fractional value of pre- recommended products is stated, including：
The internet behavioral data that preset quantity is picked out in the internet behavioral data in default measurement period from user is made
To evaluate data；
Determine history decay factor and evaluation data of the evaluation data in the measurement period in the current statistic time
The current attenuation factor；
Review number is determined according to the history decay factor of the evaluation data and the preparatory weighted value for setting described in evaluation data
According to the historical scores in default measurement period；
Determine that evaluation data exist according to the current attenuation factor of the evaluation data and the preparatory weighted value for evaluation data setting
The RealTime scores of current statistic time；
Using the summation of the historical scores of the evaluation data and RealTime scores as the second fractional value of the pre- recommended products.
6. personalized recommendation algorithm according to claim 5, which is characterized in that the history decay factor and described work as
Preceding decay factor passes through following formula
It determines, wherein t indicates date, fi(t) decay factor, a evaluated data i on the t date is indicatediIndicate evaluation data i with
The degree of time decaying, C indicate that constant, current_date indicate the date of current statistic time, hist_recom_date table
Show date, last_recom_date expression evaluation data i the last time that evaluation data i is executed in measurement period by user
The date of recommendation.
7. personalized recommendation algorithm according to claim 6, which is characterized in that the history according to the evaluation data
Decay factor and the in advance historical scores for the determining evaluation data of weighted value of evaluation data setting in default measurement period, packet
It includes：
Pass through following formula
Determine historical scores of the evaluation data in default measurement period, wherein S indicates the evaluation data of pre- recommended products pre-
If historical scores, S in measurement periodiIndicate historical scores, W of the evaluation data i in default measurement periodiExpression be in advance
Weighted value, the T of evaluation data i setting indicate that measurement period, n indicate the evaluation data of n quantity.
8. personalized recommendation algorithm according to claim 7, which is characterized in that described according to the current of the evaluation data
Decay factor and the in advance RealTime scores for the determining evaluation data of weighted value of evaluation data setting in the current statistic time, packet
It includes：
Pass through following formula
ΔSi=Wifi(current_date)
Evaluation data are determined in the RealTime scores of current statistic time, wherein Δ S indicates that the evaluation data of pre- recommended products are being worked as
The RealTime scores of preceding statistical time, Δ SiIndicate the RealTime scores in the current statistic time of evaluation data i.
9. personalized recommendation algorithm according to claim 1, which is characterized in that user internet behavioral data includes
Close recommending data, next time reminds data, customer complaint data, browsing recommendation data, repeatedly clicks bubble data, without behaviour
Make the subscription data of data, user to the recommending data of product and user to product.
10. a kind of terminal, which is characterized in that including：For executing the unit such as the described in any item methods of claim 1-9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810678309.4A CN108920596B (en) | 2018-06-27 | 2018-06-27 | Personalized recommendation algorithm and terminal |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810678309.4A CN108920596B (en) | 2018-06-27 | 2018-06-27 | Personalized recommendation algorithm and terminal |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108920596A true CN108920596A (en) | 2018-11-30 |
| CN108920596B CN108920596B (en) | 2021-10-29 |
Family
ID=64423897
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810678309.4A Active CN108920596B (en) | 2018-06-27 | 2018-06-27 | Personalized recommendation algorithm and terminal |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108920596B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109753643A (en) * | 2018-12-05 | 2019-05-14 | 平安科技（深圳）有限公司 | Report style creation method, device, equipment and computer readable storage medium |
| CN110196796A (en) * | 2019-05-15 | 2019-09-03 | 无线生活（杭州）信息科技有限公司 | The effect evaluation method and device of proposed algorithm |
| CN111177562A (en) * | 2019-12-31 | 2020-05-19 | 中国银行股份有限公司 | Recommendation and sorting processing method and device for target objects and server |
| CN113468394A (en) * | 2021-07-06 | 2021-10-01 | 北京有竹居网络技术有限公司 | Data processing method and device, electronic equipment and storage medium |
Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101551825A (en) * | 2009-05-15 | 2009-10-07 | 中国科学技术大学 | Personalized film recommendation system and method based on attribute description |
| CN102999589A (en) * | 2012-11-15 | 2013-03-27 | Tcl集团股份有限公司 | Method and system for recommending order |
| CN103164804A (en) * | 2011-12-16 | 2013-06-19 | 阿里巴巴集团控股有限公司 | Personalized method and personalized device of information push |
| CN104021163A (en) * | 2014-05-28 | 2014-09-03 | 深圳市盛讯达科技股份有限公司 | Product recommending system and method |
| CN104090886A (en) * | 2013-12-09 | 2014-10-08 | 深圳市腾讯计算机系统有限公司 | Method and device for constructing real-time portrayal of user |
| CN104102648A (en) * | 2013-04-07 | 2014-10-15 | 腾讯科技（深圳）有限公司 | User behavior data based interest recommending method and device |
| CN104598583A (en) * | 2015-01-14 | 2015-05-06 | 百度在线网络技术（北京）有限公司 | Method and device for generating query sentence recommendation list |
| US20160085859A1 (en) * | 2010-08-15 | 2016-03-24 | John W. Ogilvie | Analytic comparison of libraries and playlists |
| CN105469263A (en) * | 2014-09-24 | 2016-04-06 | 阿里巴巴集团控股有限公司 | Commodity recommendation method and device |
| CN105574216A (en) * | 2016-03-07 | 2016-05-11 | 达而观信息科技（上海）有限公司 | Personalized recommendation method and system based on probability model and user behavior analysis |
| CN106021337A (en) * | 2016-05-09 | 2016-10-12 | 房加科技（北京）有限公司 | A big data analysis-based intelligent recommendation method and system |
| CN106228386A (en) * | 2016-07-12 | 2016-12-14 | 腾讯科技（深圳）有限公司 | A kind of information-pushing method and device |
| CN107016058A (en) * | 2017-03-10 | 2017-08-04 | 浙江工业大学 | A kind of recommendation Forecasting Methodology based on attribute information preference self study |
| CN107729542A (en) * | 2017-10-31 | 2018-02-23 | 咪咕音乐有限公司 | A kind of information methods of marking and device and storage medium |
| CN108197980A (en) * | 2017-12-27 | 2018-06-22 | 中国科学院上海高等研究院 | Illustration generation method/system, storage medium and the terminal of personalized shopper |
-
2018
- 2018-06-27 CN CN201810678309.4A patent/CN108920596B/en active Active
Patent Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101551825A (en) * | 2009-05-15 | 2009-10-07 | 中国科学技术大学 | Personalized film recommendation system and method based on attribute description |
| US20160085859A1 (en) * | 2010-08-15 | 2016-03-24 | John W. Ogilvie | Analytic comparison of libraries and playlists |
| CN103164804A (en) * | 2011-12-16 | 2013-06-19 | 阿里巴巴集团控股有限公司 | Personalized method and personalized device of information push |
| CN102999589A (en) * | 2012-11-15 | 2013-03-27 | Tcl集团股份有限公司 | Method and system for recommending order |
| CN104102648A (en) * | 2013-04-07 | 2014-10-15 | 腾讯科技（深圳）有限公司 | User behavior data based interest recommending method and device |
| CN104090886A (en) * | 2013-12-09 | 2014-10-08 | 深圳市腾讯计算机系统有限公司 | Method and device for constructing real-time portrayal of user |
| CN104021163A (en) * | 2014-05-28 | 2014-09-03 | 深圳市盛讯达科技股份有限公司 | Product recommending system and method |
| CN105469263A (en) * | 2014-09-24 | 2016-04-06 | 阿里巴巴集团控股有限公司 | Commodity recommendation method and device |
| CN104598583A (en) * | 2015-01-14 | 2015-05-06 | 百度在线网络技术（北京）有限公司 | Method and device for generating query sentence recommendation list |
| CN105574216A (en) * | 2016-03-07 | 2016-05-11 | 达而观信息科技（上海）有限公司 | Personalized recommendation method and system based on probability model and user behavior analysis |
| CN106021337A (en) * | 2016-05-09 | 2016-10-12 | 房加科技（北京）有限公司 | A big data analysis-based intelligent recommendation method and system |
| CN106228386A (en) * | 2016-07-12 | 2016-12-14 | 腾讯科技（深圳）有限公司 | A kind of information-pushing method and device |
| CN107016058A (en) * | 2017-03-10 | 2017-08-04 | 浙江工业大学 | A kind of recommendation Forecasting Methodology based on attribute information preference self study |
| CN107729542A (en) * | 2017-10-31 | 2018-02-23 | 咪咕音乐有限公司 | A kind of information methods of marking and device and storage medium |
| CN108197980A (en) * | 2017-12-27 | 2018-06-22 | 中国科学院上海高等研究院 | Illustration generation method/system, storage medium and the terminal of personalized shopper |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109753643A (en) * | 2018-12-05 | 2019-05-14 | 平安科技（深圳）有限公司 | Report style creation method, device, equipment and computer readable storage medium |
| CN109753643B (en) * | 2018-12-05 | 2023-10-20 | 平安科技（深圳）有限公司 | Report style creation method, device, equipment and computer readable storage medium |
| CN110196796A (en) * | 2019-05-15 | 2019-09-03 | 无线生活（杭州）信息科技有限公司 | The effect evaluation method and device of proposed algorithm |
| CN110196796B (en) * | 2019-05-15 | 2023-04-28 | 无线生活（杭州）信息科技有限公司 | Effect evaluation method and device for recommendation algorithm |
| CN111177562A (en) * | 2019-12-31 | 2020-05-19 | 中国银行股份有限公司 | Recommendation and sorting processing method and device for target objects and server |
| CN111177562B (en) * | 2019-12-31 | 2024-04-19 | 中国银行股份有限公司 | Recommendation ordering processing method and device for target object and server |
| CN113468394A (en) * | 2021-07-06 | 2021-10-01 | 北京有竹居网络技术有限公司 | Data processing method and device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108920596B (en) | 2021-10-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Sharma et al. | Exchange rate volatility and tourism demand in India: Unraveling the asymmetric relationship | |
| US9230239B2 (en) | Method and system for providing performance metrics | |
| CN108920596A (en) | A kind of personalized recommendation algorithm and terminal | |
| Diehl | When two rights make a wrong: Searching too much in ordered environments | |
| De Oliveira et al. | Brand equity estimation model | |
| Hennig-Thurau et al. | Conceptualizing and measuring the monetary value of brand extensions: The case of motion pictures | |
| Gravelle et al. | Competition, prices and quality in the market for physician consultations | |
| AU2017210638A1 (en) | Social match platform apparatuses, methods and systems | |
| US20090292548A1 (en) | Method, system, and program product for information editorial controls | |
| US7769606B2 (en) | Interactive health insurance system | |
| JP5253519B2 (en) | Method, apparatus and storage medium for generating smart text | |
| US20090192808A1 (en) | Method and Device for Providing a Personal Product Recommendation | |
| Gao et al. | Information acquisition and voluntary disclosure in an export‐processing system | |
| US20150317581A1 (en) | System and method for providing real estate referrals | |
| Schlereth et al. | Measuring consumers' preferences for metered pricing of services | |
| US20150025962A1 (en) | System, method, and computer program for pricing and allocating advertising inventory on digital and web publisher properties | |
| US11309082B2 (en) | System and method for monitoring engagement | |
| Chirra et al. | Barriers to sustainable supply chain flexibility during sales promotions | |
| Ayeh et al. | ‘Estimating tomorrow’s tourist arrivals’: forecasting the demand for China’s tourism using the general-to-specific approach | |
| Oloyede et al. | Measuring the impact of the digital economy in developing countries: A systematic review and meta-analysis | |
| Rajendran et al. | Customer satisfaction index as a performance evaluation metric: A study on indian e-banking industry | |
| Wang et al. | Province-level estimation of waste mobile phones in China and location planning of recycling centers | |
| Essel | Assessing Materials Management Practices Effect on Firm’s Performance in Ghana Using Dominance Analysis: Evidence from a Listed Company | |
| JP2021140525A (en) | Loan management system, method, and program | |
| JP6532065B2 (en) | Operation support device, operation support method, operation support program |
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 |