CN105354720B - A method of mixed recommendation is carried out to consumption place based on visual cluster - Google Patents
A method of mixed recommendation is carried out to consumption place based on visual cluster Download PDFInfo
- Publication number
- CN105354720B CN105354720B CN201410478389.0A CN201410478389A CN105354720B CN 105354720 B CN105354720 B CN 105354720B CN 201410478389 A CN201410478389 A CN 201410478389A CN 105354720 B CN105354720 B CN 105354720B
- Authority
- CN
- China
- Prior art keywords
- consumption
- place
- cluster
- consumer
- recommendation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention provides a kind of methods for carrying out mixed recommendation to consumption place based on visual cluster, the steps include: 1) to record credit card purchase the pretreatment for carrying out data;2) it carries out RadViz to consumption place visually to cluster, the similar consumption place cluster of discovery consumption feature;3) RadViz is carried out to consumer visually to cluster, find the cluster of consumer similar in consumption mode;4) the consumption place based on content recommendation method and collaborative filtering recommending method discovery is added in the recommendation list of corresponding consumer spending place by consumer;5) it is further processed consumer spending place recommendation list, consumption place recommendation tables is generated, completes the consumption place personalized recommendation carried out to consumer.The present invention is completed from multi-angle to carry out the personalized recommendation in most possible interested consumption place to consumer, solves the problem of content-based recommendation and collaborative filtering recommending relevant with post-consumer place can not carry out personalized recommendation.
Description
Technical field
The present invention relates to consumption places to recommend, more particularly to there are in the case where a large amount of credit card purchases record, sufficiently digs
Consumer is dug in the preference tendency in certain consumption places, is that consumer recommends most probable from multi-angle using mixed recommendation method
Interested other consumption places.
Background technique
Recommender system has application to including search engine, e-commerce, information retrieval, social network service, news media
Many fields Deng including.Current recommended technology specifically includes that based on commending contents, collaborative filtering recommending and is based on being associated with rule
Then recommend.But these methods all have many disadvantages: content-based recommendation has consumer record according to consumer to be pushed away for it
The consumption place similar with selecting before is recommended, this method is only consumer and provides interested place before it, Bu Nengfa
It can interested place after existing consumer;The recommendation of collaborative filtering is that consumer provides other people similar with its consumption feature
Selected consumption place, this lacking individuality of recommended method;Recommendation based on correlation rule does not account in correlation rule
Each precedence, and the consumption of consumer has stringent precedence.
Traditional data mining technology and algorithm allows user's indigestion and use, and cannot be that driving is carried out with user
Interaction.Visualization is easier to understand data and Result, allows to be compared result and examine, is also used for guide data
Mining algorithm, during making user participate in Analysis of Policy Making.Visual Data Mining Techniques foundation is being visualized and was being analyzed
On the basis of journey, it is to portray structure and show the functionality and human perception mode, exception, tendency and relationship of data
Based on ability, reinforce data mining treatment process with visualization.
Currently, the consumption that credit card purchase becomes many people is first because of the characteristics of its convenience is general, the credit line that can overdraw
Choosing.With the rapid development of computer level, credit card purchase record is acquired very perfect with the method for preservation.
Reasonable analysis credit card purchase records consumption feature and mode it can be found that user, catches the tendentiousness consumer psychology of user,
The accuracy for improving class of subscriber identification, counts the place of similar consumption feature, provides for user more personalized
Recommendation service, this become the consumption place such as bank, credit card company and market, supermarket further promoted enterprises service it is horizontal must
So require.
Summary of the invention
The main object of the present invention is all there is the case where respective advantage and disadvantage for all kinds of recommended methods, proposes that one kind includes
The mixed recommendation method of content-based recommendation and collaborative filtering recommending can with user based on the method that radar map visually clusters
The mode of interaction is participated in, for the personalized recommendation for carrying out consumption place to consumer.
Mentality of designing of the invention are as follows: based on the method that radar map visually clusters, consumption place, consumer are mapped to thunder
Up in figure, in the case where selecting suitable dimension, similar consumption place, consumer can assemble because of consumption mode similar in it
Together, user can choose label and participate in interaction, to recommend to consume place consumption feature with selected before it for consumer
Similar other places, other consumers selected place similar with this consumer spending mode.
A method of mixed recommendation is carried out to consumption place based on visual cluster comprising the steps of:
Step 1): recording the pretreatment for carrying out data to credit card purchase, obtains including consumer, consumption place, consumption
The consumption Basic Information Table Basic_Infor_Tab and consumer of time and spending amount field area of consumption point set P;
Step 2): the consumption Basic Information Table Basic_Infor_Tab obtained according to step 1) carries out consumption place
RadViz is visually clustered, and obtains the similar consumption place cluster of consumption feature;
Step 3): clustering according to the consumption place that step 2) obtains, and carries out RadViz to consumer and visually clusters, is disappeared
Take the cluster of consumer similar in mode;
Step 4): the consumption place cluster obtained according to step 2) carries out content-based recommendation, is obtained according to step 3)
Consumer cluster and carry out collaborative filtering recommending, the consumption place that will be obtained is added in corresponding consumption place recommendation list;
Step 5): the consumer spending place recommendation list Recommend_List established according to step 4), to area of consumption
It presses frequency of occurrence to be ranked up, consumer then is recommended into the most N number of consumption place of frequency of occurrence, that is, is completed to consumption
The recommendation in person's progress personalized consumption place.
The step 1) the following steps are included:
Step 1.1): it is recorded according to credit card trade and establishes consumption Basic Information Table Basic_Infor_Tab, for describing
When and where certain consumer is carrying out the consumption of how many spending amount, i.e., the table include consumer, consumption place, consumption time,
Spending amount field;
Step 1.2): area of consumption point set P corresponding to each consumer creation, from the basic letter of step 1.1) gained consumption
Chosen in breath table the consumer be expert in consumption place, and be added into set P;
Step 1.3): the consumption Basic Information Table established according to step 1.1) calculates associated statistical information, including consumption time
Number Count, spending amount average value Avg, spending amount maximum value Max and spending amount minimum M in;
The step 2) visually clusters that specific step is as follows to the RadViz in consumption place:
Step 2.1): consumption time or statistical information are chosen as dimension anchor point, is evenly distributed on RadViz circumference;
Step 2.2): visualizing consumption place, and consumption place can be because of spring tension that dimension anchor point generates it
And its final position is determined in radar map;
Step 2.3): similar consumption place will flock together in RadViz due to similar consumption feature, it is described certain
The consumption feature in place refers to the time of consumption generation in the consumption place, the amount of money of consumption and consumer, according to adaptive K-
Means algorithm carries out initial clustering to place in RadViz annulus, firstly, setting initial clustering numberWherein X is
It consumes and consumes place number in Basic Information Table Basic_Infor_Tab, meanwhile, the number of iterations I is setn=1, cluster interior record
Number limits Nummin=3, maximum iterations Imax=100;Secondly, selecting K record point at random in RadViz as poly-
Class center Centeri, K initial clustering C is generated using K-Means algorithmi, i=1,2, K;Then judgement is each poly-
Number Num is recorded in classiWhether it is less than record number in class and limits NumminIf Numi< Nummin, then cancel such center
Centeri, K=K-1, the number of iterations In=In+ 1, it reuses K-Means method and RadViz is clustered, if each cluster
Middle record number NumiBoth greater than NumminOr the number of iterations In> maximum iterations Imax, then initial clustering is completed;
Step 2.4): in step 2.3) initial clustering number K, record number limits Num in clustermin, most iteration
Number ImaxIt is adjusted, until obtaining cluster radius less than centroid distance between cluster, cluster radius is cluster CiMass center to side
Until the final cluster result of the maximum value of edge record, wherein the adjusting range of initial clustering number K is greater than 0 no more than X's
Integer clusters interior record number and limits NumminAdjusting range be integer greater than 0, maximum iterations ImaxAdjusting range
For the integer greater than 0.
Different clusters in final cluster result are selected by subscriber frame, and the place in different clusters is distinguished with different colours
Visually to cluster since RadViz is visually clustered in order to which user's naked eyes identify, by select frame and different colours both
Means, so that it may allow user to understand easily and see and identify different clusters.
The step 3) visually clusters that specific step is as follows to the RadViz of consumer:
Step 3.1): according to step 2) to consumption place cluster as a result, the place in one or two cluster is selected to make
For radar map RadViz dimension anchor point, it is evenly distributed on RadViz circumference;
Step 3.2): visualizing consumer, and consumer can be due to the spring force that dimension anchor point generates it in thunder
Up to its final position determining in figure;
Step 3.3): similar consumer will flock together in RadViz due to similar consumption mode, the consumption
The consumption mode of person refers to consumption time, spending amount and consumption place in the consumer when consuming, according to adaptive
K-Means algorithm carries out initial clustering to place in RadViz annulus, firstly, setting initial clustering numberWherein X
To consume place number in consumption Basic Information Table Basic_Infor_Tab, while the number of iterations I is setn=1, cluster interior note
It records number and limits Nummin=3;Secondly, selecting K record point at random in RadViz is used as cluster centre Centeri, use K-
Means algorithm generates K initial clustering Ci, i=1,2, K;Then judge to record number Num in each clusteriWhether
Num is limited less than number is recorded in classminIf Numi< Nummin, then cancel such center Centeri, K=K-1, iteration time
Number In=In+ 1, it reuses K-Means method and RadViz is clustered, if recording number Num in each clusteriBoth greater than
NumminOr the number of iterations In > maximum iterations Imax, then initial clustering is completed;
Step 3.4): in step 3.3) initial clustering number K, record number limits Num in clustermin, most iteration
Number ImaxIt is adjusted, until obtaining cluster radius less than centroid distance between cluster, cluster radius is cluster CiMass center to side
Until the final cluster result of the maximum value of edge record, wherein the adjusting range of initial clustering number K is greater than 0 no more than X's
Integer clusters interior record number and limits NumminAdjusting range be integer greater than 0, maximum iterations ImaxAdjusting range
For the integer greater than 0.
Different clusters in final cluster result are selected by subscriber frame, and the place in different clusters is distinguished with different colours
Visually to cluster since RadViz is visually clustered in order to which user's naked eyes identify, by select frame and different colours both
Means, so that it may allow user to understand easily and see and identify different clusters.
The specific of place is consumed in content-based recommendation method and the discovery of the recommended method of collaborative filtering in the step 4)
Steps are as follows:
Step 4.1): for each place Pi in the place of post-consumer before step 1.2) set P i.e. certain consumer, root
It is visually clustered according to RadViz of the step 2) to consumption place, by other places where Pi in cluster, is added to corresponding consumer
It consumes in place recommendation list Recommend_List, obtains content-based recommendation place;
Step 4.2): visually clustering according to RadViz of the step 3) to consumer, by its where certain consumer in cluster
Place in the corresponding area of consumption point set P of his consumer, is added to corresponding consumer spending place recommendation list Recommend_
In List, the recommendation place of collaborative filtering is obtained.
The step 5) to consumer carry out consumption place personalized recommendation the following steps are included:
Step 5.1): consumption place recommendation list Recommend_List corresponding to each consumer is handled, meter
The weight for calculating each consumption place in list counts what this consumption place occurred in the consumer spending place recommendation list
Number C, and be added in the recommendation tables Recommend_Tab of consumer spending place by sequence arrangement from high to low;
Step 5.2): top n in the consumption place recommendation tables Recommend_Tab after the completion of step 5.1) is consumed into place
Recommend respective objects consumer.
Beneficial effect
The present invention provides a kind of based on visual cluster to the mixed recommendation method in consumption place, from multi-angle complete for
Consumer carries out the personalized recommendation in most possible interested consumption place, solve content-based recommendation can only according to
The problem of recommending similar place and collaborative filtering recommending that can not carry out personalized recommendation toward selection.
The mixed recommendation method proposed by the present invention recorded towards credit card purchase to consumption place, is based on radar map
The visual clustering method of RadViz by the structure of data and is hidden in mode therein with simple and clear mode with graph image
Form show, and allow user's interaction, allow the formation of user's supervision clustering, take full advantage of the very strong perception of the mankind
And judgement, compensate for the wide gap between computerized algorithm and human cognitive.
Detailed description of the invention
Fig. 1 is the method for the invention flow chart;
Fig. 2 is the visual cluster flow chart to consumption place;
Fig. 3 is the visual cluster to consumption place;
Fig. 4 is the visual cluster to consumer;
Fig. 5 is the personalized recommendation flow chart for completing to carry out consumer consumption place.
Specific embodiment
To keep the purpose of the present invention, mentality of designing and advantage clearer, below in conjunction with specific example, and referring to attached drawing,
Invention is further described in detail.
The present invention provides a kind of methods (title) for carrying out mixed recommendation to consumption place based on visual cluster, such as Fig. 1
It is shown, including five key steps: recording the pretreatment for carrying out data to credit card purchase;Carrying out RadViz to consumption place can
Depending on cluster, the similar consumption place cluster of discovery consumption feature;RadViz is carried out to consumer visually to cluster, and finds consumption mode
Similar consumer's cluster;For consumer, according to the consumption based on content recommendation method and collaborative filtering recommending method discovery
Place is added in corresponding consumption place recommendation tables;According to consumer spending place recommendation tables, consumer is consumed in completion
The personalized recommendation in place.
The committed step that method of the invention is related to is described in detail one by one below, it is shown that specific step is as follows:
Step 1 carries out the pretreatment of data to given credit card purchase record, works including three: according to friendship
Easily record establishes consumption Basic Information Table Basic_Infor_Tab, more for being engraved in the consumption of certain place when describing certain consumer
Few amount of money, the i.e. table include consumer, consumption time, consumption place, spending amount field;Each consumer is created accordingly
Area of consumption point set P, from chosen in consumption Basic Information Table the consumer be expert in consumption place, and be added into set
P;Calculate important statistical information, including consumption number of times, spending amount average value, maximum value and minimum value etc..
Step 2 carries out visualization cluster to consumption place, such as Fig. 2 institute according to pretreated consumption Basic Information Table
Show.Consumption time or statistical information are chosen as dimension anchor point, is evenly distributed on RadViz circumference, dimension anchor point is to consumption place
The spring tension of point is recorded, size is proportional to value of this place in the dimension, work of the consumption place record point in spring force
With lower iterative motion, the resultant force size suffered by the point is zero, as shown in figure 3, consumption place record point is in radar map at this time
Determine its final position, the final position calculation formula of consumption place record point is as follows:
Wherein, n is circumference dimension anchor point number;Vn is the value for consuming place in each dimension of circumference;Step 2.1): choosing
Cancel time-consuming or statistical information as dimension anchor point, is evenly distributed on RadViz circumference;
In RadViz figure, similar consumption place will flock together in RadViz due to similar consumption feature, institute
The consumption feature for stating certain place refers to the time of consumption generation in the consumption place, the amount of money of consumption and consumer, according to adaptive
K-Means algorithm is answered to carry out initial clustering to place in RadViz annulus, firstly, setting initial clustering numberWherein
X is consumption place number in consumption Basic Information Table Basic_Infor_Tab, meanwhile, the number of iterations I is setn=1, in cluster
It records number and limits Nummin=3, maximum iterations Imax=100;Make secondly, selecting K record point at random in RadViz
For cluster centre Centeri, K initial clustering C is generated using K-Means algorithmi, i=1,2, K;Then judgement is every
Number Num is recorded in a clusteriWhether it is less than record number in class and limits NumminIf Numi< Nummin, then cancel in such
Heart Centeri, K=K-1, the number of iterations In=In+ 1, it reuses K-Means method and RadViz is clustered, if each poly-
Number Num is recorded in classiBoth greater than NumminOr the number of iterations In> maximum iterations Imax, then initial clustering is completed.
Num is limited to number is recorded in above-mentioned initial clustering number K, clustermin, maximum iterations ImaxIt is adjusted
Whole, until obtaining cluster radius less than centroid distance between cluster, cluster radius is cluster CiThe maximum that is recorded to edge of mass center
Until the final cluster result of value, wherein the adjusting range of initial clustering number K is the integer for being not more than X greater than 0, clusters interior note
It records number and limits NumminAdjusting range be integer greater than 0, maximum iterations ImaxAdjusting range be it is whole greater than 0
Number.
Different clusters in final cluster result are selected by subscriber frame, and the place in different clusters is distinguished with different colours
Visually to cluster since RadViz is visually clustered in order to which user's naked eyes identify, by select frame and different colours both
Means, so that it may allow user to understand easily and see and identify different clusters.
Step 3 carries out RadViz to consumer according to pretreated consumption Basic Information Table Basic_Infor_Tab
Visual cluster, as shown in Figure 4.According to step 2 to consumption place cluster as a result, selecting the place in one or two cluster
As radar map RadViz dimension anchor point, it is evenly distributed on RadViz circumference.Consumer is visualized, consumer can be because of dimension
It spends the spring force that anchor point generates it and determines its final position in RadViz, calculation formula is as follows:
Wherein, n is circumference dimension anchor point number;Vn is value of the consumer in each dimension of circumference;
In RadViz figure, similar consumer will flock together in RadViz due to similar consumption mode, described
The consumption mode of consumer refers to consumption time, spending amount and consumption place in the consumer when consuming, according to certainly
It adapts to K-Means algorithm and initial clustering is carried out to place in RadViz annulus, firstly, setting initial clustering numberIts
Middle X is to consume place number in consumption Basic Information Table Basic_Infor_Tab, while the number of iterations I is arrangedn=1, in cluster
It records number and limits Nummin=3;Secondly, selecting K record point at random in RadViz is used as cluster centre Centeri, use
K-Means algorithm generates K initial clustering Ci, i=1,2, K;Then judge to record number Num in each clusteriIt is
It is no to be less than record number limitation Num in classminIf Numi< Nummin, then cancel such center Centeri, K=K-1, iteration
Number In=In+ 1, it reuses K-Means method and RadViz is clustered, if recording number Num in each clusteriIt is all big
In NumminOr the number of iterations In > maximum iterations Imax, then initial clustering is completed;
Num is limited to number is recorded in above-mentioned initial clustering number K, clustermin, maximum iterations ImaxIt is adjusted,
Until obtaining cluster radius less than centroid distance between cluster, cluster radius is cluster CiThe mass center maximum value that records to edge
Until final cluster result, wherein the adjusting range of initial clustering number K is the integer for being not more than X greater than 0, clusters interior record
Number limitation NumminAdjusting range be integer greater than 0, maximum iterations ImaxAdjusting range be integer greater than 0.
Different clusters in final cluster result are selected by subscriber frame, and the place in different clusters is distinguished with different colours
Visually to cluster since RadViz is visually clustered in order to which user's naked eyes identify, by select frame and different colours both
Means, so that it may allow user to understand easily and see and identify different clusters.
Step 4, for consumer, according to the area of consumption based on content recommendation method and collaborative filtering recommending method discovery
Point is added in corresponding consumption place recommendation list Recommend_List.Specific step is as follows:
1): the place to post-consumer before certain consumer is each place Pi in set P, according to step 2) to area of consumption
The RadViz of point is visually clustered, and by other places where Pi in cluster, is added to corresponding consumer spending place recommendation list
In Recommend_List, content-based recommendation is completed;
2): it is visually clustered according to the RadViz to consumer, other consumers where certain consumer in cluster are corresponding
Place in area of consumption point set P is added in corresponding consumer spending place recommendation list Recommend_List, completes association
With the recommendation of filtering.
At this point, there is the recommendation place repeated in the consumption place recommendation tables of consumer.
Step 5, according to consumer spending place recommendation list Recommend_List, consumer is consumed in completion
The personalized recommendation in place, as shown in Figure 5.It calculates and is consumed in the recommendation list Recommend_List of consumer spending place first
The frequency of occurrence C in place, if there is this place for the first time, is occurred that is, in the consumption place recommendation list of consumer
Number is 1;If this place had already appeared, it is only necessary to which its frequency of occurrence is added 1.Then, by area of consumption press frequency of occurrence from
High to Low sequence arrangement is added in the recommendation tables Recommend_Tab of consumer spending place.Finally, place recommendation tables will be consumed
Respective objects consumer is recommended in top n consumption place in Recommend_Tab, completes to push away the consumption place mixing of consumer
Recommend method.
Claims (7)
1. a kind of method for carrying out mixed recommendation to consumption place based on visual cluster, which is characterized in that comprise the steps of:
Step 1): recording credit card purchase and carry out data processing, and to obtain include consumer, consumes and place, consumption time and disappears
Take the consumption Basic Information Table Basic_Infor_Tab and consumer area of consumption point set P of amount field;
Step 2): the consumption Basic Information Table Basic_Infor_Tab obtained according to step 1) carries out RadViz to consumption place
Visual cluster obtains the similar consumption place cluster of consumption feature;
Step 3): clustering according to the consumption place that step 2) obtains, and carries out RadViz to consumer and visually clusters, and obtains consumption mould
The cluster of consumer similar in formula;
Step 4): the consumption place cluster obtained according to step 2) carries out content-based recommendation, is disappeared according to what step 3) obtained
The person of expense, which clusters, carries out collaborative filtering recommending, and the consumption place that will be obtained is added in corresponding consumption place recommendation list;
Step 5): the consumer spending place recommendation list Recommend_List established according to step 4) presses consumption place
Frequency of occurrence is ranked up, and consumer then is recommended in the most N number of consumption place of frequency of occurrence, that is, complete to consumer into
The recommendation in row personalized consumption place;
The step 2) carries out RadViz to consumption place and visually clusters that specific step is as follows:
Step 2.1): consumption time or statistical information are chosen as dimension anchor point, is evenly distributed on RadViz circumference;
Step 2.2): to consumption place visualize, consumption place can due to the spring tension that dimension anchor point generates it
Its final position is determined in RadViz;
Step 2.3): similar consumption place will flock together in RadViz due to similar consumption feature, consume place
Consumption feature refers to the time of consumption generation in the consumption place, the amount of money of consumption and consumer, is calculated according to adaptive K-Means
Method carries out initial clustering to place in RadViz annulus, firstly, setting initial clustering numberWherein X is that consumption is basic
Place number is consumed in information table Basic_Infor_Tab, meanwhile, the number of iterations I is setn=1, cluster interior record number limitation
Nummin=3, maximum iterations Imax=100;Secondly, selecting K record point at random in RadViz is used as cluster centre
Centeri, K initial clustering C is generated using K-Means algorithmi, i=1,2, K;Then judge to remember in each cluster
Record number NumiWhether it is less than record number in class and limits NumminIf Numi< Nummin, then cancel such center Centeri,
K=K-1, the number of iterations In=In+ 1, it reuses K-Means method and RadViz is clustered, if being recorded in each cluster a
Number NumiBoth greater than NumminOr the number of iterations In> maximum iterations Imax, then initial clustering is completed;
Step 2.4): in step 2.3) initial clustering number K, record number limits Num in clustermin, maximum iterations
ImaxIt is adjusted, until obtaining cluster radius less than centroid distance between cluster, cluster radius is cluster CiMass center remember to edge
Until the final cluster result of the maximum value of record, wherein the adjusting range of initial clustering number K is the integer for being not more than X greater than 0,
Record number limits Num in clusterminAdjusting range be integer greater than 0, maximum iterations ImaxAdjusting range be big
In 0 integer.
2. a kind of method for carrying out mixed recommendation to consumption place based on visual cluster according to claim 1, feature
Be, the step 1) the following steps are included:
Step 1.1): it is recorded according to credit card trade and establishes consumption Basic Information Table Basic_Infor_Tab, disappeared for describing certain
When and where expense person is carrying out the consumption of how many spending amount, i.e. every a line of the table all includes consumer, consumption place, consumption
Time, spending amount field;
Step 1.2): area of consumption point set P corresponding to each consumer creation consumes Basic Information Table from step 1.1) gained
It is middle choose the consumer be expert in consumption place, and be added into set P;
Step 1.3): the consumption Basic Information Table established according to step 1.1) calculates the ASSOCIATE STATISTICS in consumption place and consumer
Information, including consumption number of times Count, spending amount average value Avg, spending amount maximum value Max and spending amount minimum value
Min。
3. a kind of method for carrying out mixed recommendation to consumption place based on visual cluster according to claim 1, feature
It is, in the step 2.4), the different clusters in final cluster result are selected by subscriber frame, and are distinguished not with different colours
With the place in cluster in order to which user's naked eyes identify.
4. a kind of method for carrying out mixed recommendation to consumption place based on visual cluster according to claim 1, feature
It is, the step 3) carries out RadViz to consumer and visually clusters that specific step is as follows:
Step 3.1): according to step 2) to consumption place cluster as a result, selecting the place in one or two cluster as thunder
Up to figure RadViz dimension anchor point, it is evenly distributed on RadViz circumference;
Step 3.2): visualizing consumer, and consumer can be due to the spring force that dimension anchor point generates it in radar map
Its final position of middle determination;
Step 3.3): similar consumer will flock together in RadViz due to similar consumption mode, the consumption of consumer
Mode refers to consumption time, spending amount and consumption place in the consumer when consuming, is calculated according to adaptive K-Means
Method carries out initial clustering to place in RadViz annulus, firstly, setting initial clustering numberWherein X is that consumption is basic
Place number is consumed in information table Basic_Infor_Tab, while the number of iterations I is setn=1, cluster interior record number limitation
Nummin=3;Secondly, selecting K record point at random in RadViz is used as cluster centre Centeri, use K-Means algorithm
Generate K initial clustering Ci, i=1,2, K;Then judge to record number Num in each clusteriWhether class in note is less than
It records number and limits NumminIf Numi< Nummin, then cancel such center Centeri, K=K-1, the number of iterations In=In+ 1,
It reuses K-Means method to cluster RadViz, if recording number Num in each clusteriBoth greater than NumminOr iteration
Number In > maximum iterations Imax, then initial clustering is completed;
Step 3.4): in step 3.3) initial clustering number K, record number limits Num in clustermin, maximum iterations
ImaxIt is adjusted, until obtaining cluster radius less than centroid distance between cluster, cluster radius is cluster CiMass center remember to edge
Until the final cluster result of the maximum value of record, wherein the adjusting range of initial clustering number K is the integer for being not more than X greater than 0,
Record number limits Num in clusterminAdjusting range be integer greater than 0, maximum iterations ImaxAdjusting range be big
In 0 integer.
5. a kind of method for carrying out mixed recommendation to consumption place based on visual cluster according to claim 4, feature
It is, in the step 3.4), the different clusters in final cluster result are selected by subscriber frame, and are distinguished not with different colours
With the place in cluster in order to which user's naked eyes identify.
6. a kind of method for carrying out mixed recommendation to consumption place based on visual cluster according to claim 2, feature
It is, the specific steps of content-based recommendation method and the recommended method of collaborative filtering discovery consumption place in the step 4)
It is as follows:
Step 4.1): for each place Pi in the place of post-consumer before step 1.2) set P i.e. certain consumer, according to step
It is rapid that 2) RadViz in consumption place is visually clustered, by other places where Pi in cluster, it is added to corresponding consumer spending
In the recommendation list Recommend_List of place, content-based recommendation place is obtained;
Step 4.2): visually clustering according to RadViz of the step 3) to consumer, and other where certain consumer in cluster are disappeared
Place in the corresponding area of consumption point set P of the person of expense, is added to corresponding consumer spending place recommendation list Recommend_List
In, obtain the recommendation place of collaborative filtering.
7. a kind of method for carrying out mixed recommendation to consumption place based on visual cluster according to claim 1, feature
Be, the step 5) to consumer carry out consumption place personalized recommendation the following steps are included:
Step 5.1): consumption place recommendation list Recommend_List corresponding to each consumer is handled, and calculates column
The weight in each consumption place in table counts the number that this consumption place occurs in the consumer spending place recommendation list
C, and be added in the recommendation tables Recommend_Tab of consumer spending place by sequence arrangement from high to low;
Step 5.2): top n consumption place in the consumption place recommendation tables Recommend_Tab after the completion of step 5.1) is recommended
Give respective objects consumer.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410478389.0A CN105354720B (en) | 2014-09-18 | 2014-09-18 | A method of mixed recommendation is carried out to consumption place based on visual cluster |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410478389.0A CN105354720B (en) | 2014-09-18 | 2014-09-18 | A method of mixed recommendation is carried out to consumption place based on visual cluster |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105354720A CN105354720A (en) | 2016-02-24 |
CN105354720B true CN105354720B (en) | 2019-04-09 |
Family
ID=55330687
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410478389.0A Active CN105354720B (en) | 2014-09-18 | 2014-09-18 | A method of mixed recommendation is carried out to consumption place based on visual cluster |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105354720B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106610970A (en) * | 2015-10-21 | 2017-05-03 | 上海文广互动电视有限公司 | Collaborative filtering-based content recommendation system and method |
CN106055580B (en) * | 2016-05-23 | 2019-02-05 | 中南大学 | A kind of fuzzy clustering result visualization method based on Radviz |
CN109934614A (en) * | 2019-01-18 | 2019-06-25 | 深圳壹账通智能科技有限公司 | A kind of credit card promotion method and device based on Consumption thermodynamic chart |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110191311A1 (en) * | 2010-02-03 | 2011-08-04 | Gartner, Inc. | Bi-model recommendation engine for recommending items and peers |
CN102780920A (en) * | 2011-07-05 | 2012-11-14 | 上海奂讯通信安装工程有限公司 | Television program recommending method and system |
CN102609523B (en) * | 2012-02-10 | 2018-03-06 | 上海视畅信息科技有限公司 | The collaborative filtering recommending method classified based on taxonomy of goods and user |
-
2014
- 2014-09-18 CN CN201410478389.0A patent/CN105354720B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN105354720A (en) | 2016-02-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Bithas et al. | Unmasking decoupling: redefining the resource intensity of the economy | |
Marcus et al. | Counting with the crowd | |
CN108090800B (en) | Game prop pushing method and device based on player consumption potential | |
CN110490729B (en) | Financial user classification method based on user portrait model | |
CN107851097A (en) | Data analysis system, data analysing method, data analysis program and storage medium | |
CN110335157A (en) | Insurance products recommended method, equipment and storage medium | |
CN107545422A (en) | A kind of arbitrage detection method and device | |
US20130204822A1 (en) | Tools and methods for determining relationship values | |
CN107886241B (en) | Resource analysis method, device, medium, and electronic apparatus | |
CN106776859A (en) | Mobile solution App commending systems based on user preference | |
CN105354720B (en) | A method of mixed recommendation is carried out to consumption place based on visual cluster | |
Chen et al. | Big data analytics on aviation social media: The case of china southern airlines on sina weibo | |
Chiang | Using a data-driven marketing strategy on customer relationship management: an empirical case of urban coffee shops in Taiwan | |
Mladenović et al. | Cross-cultural dimensions influence on business internationalization by soft computing technique | |
CN104866490B (en) | A kind of video intelligent recommended method and its system | |
CN110992111A (en) | Image mining method for applicant user based on big data | |
CN111967970B (en) | Bank product recommendation method and device based on spark platform | |
Zhang et al. | A Precision Marketing Strategy of e‐Commerce Platform Based on Consumer Behavior Analysis in the Era of Big Data | |
CN110008406A (en) | A kind of public sentiment personalized recommendation algorithm based on similar community | |
CN107092617B (en) | Information processing method and device | |
Enza Wella et al. | Service quality dealer identification: the optimization of K-Means clustering | |
Yu et al. | The geographic disparity of agglomeration economies: Evidence from industrial activities in China's emerging greater bay area | |
Bian | Social media marketing optimization method based on deep neural network and evolutionary algorithm | |
Yang et al. | Application of clustering for customer segmentation in private banking | |
Hirschmeier et al. | Information Quality Needs throughout the Purchase Process. |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |