CN109325186B - Behavior motivation inference algorithm integrating user preference and geographic features - Google Patents

Behavior motivation inference algorithm integrating user preference and geographic features Download PDF

Info

Publication number
CN109325186B
CN109325186B CN201810910570.2A CN201810910570A CN109325186B CN 109325186 B CN109325186 B CN 109325186B CN 201810910570 A CN201810910570 A CN 201810910570A CN 109325186 B CN109325186 B CN 109325186B
Authority
CN
China
Prior art keywords
feature
user
geo
weight
features
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
Application number
CN201810910570.2A
Other languages
Chinese (zh)
Other versions
CN109325186A (en
Inventor
姜建武
李景文
陆妍玲
殷敏
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guilin University of Technology
Original Assignee
Guilin University of Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Guilin University of Technology filed Critical Guilin University of Technology
Priority to CN201810910570.2A priority Critical patent/CN109325186B/en
Publication of CN109325186A publication Critical patent/CN109325186A/en
Application granted granted Critical
Publication of CN109325186B publication Critical patent/CN109325186B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Abstract

The invention discloses a behavior motivation inference method fusing user preference characteristics and geographic characteristics. Extracting geographic features and user preference feature source data from national POI data of a high-grade map and internet behavior data of a user, constructing the extracted geographic features and user preference features as a feature set, intersection processing is carried out on the constructed geographic features and the user preference feature set, the specific strategy is to take the feature tag as a judgment basis, the product of the geographic features and the user preference feature set is obtained, the obtained set product is sorted according to the weight in a descending order of the set elements, the features with smaller weights are removed, namely feature screening, the retained features, namely the remarkable behavior motivation of the user are judged according to the feature weight, and the larger the weight of the characteristic is, the more the behavior motivation of the user is biased to the characteristic, otherwise, the more the characteristic of the user deviates from the characteristic, thereby determining the obvious behavior motivation of the user and finally realizing the inference of the behavior motivation.

Description

Behavior motivation inference algorithm integrating user preference and geographic features
Technical Field
The invention relates to an accurate inference algorithm for user behavior motivation in intelligent information service, which realizes accurate inference of user behavior motivation trend by organically combining geographic characteristics with user preference characteristics. Belonging to the field of intelligent service research and application.
Background
The intelligent service is embodied by information technology development and service mode change, and the inferred user behavior motivation is added in the intelligent service process, so that the real-time intelligent information service with higher accuracy and stronger individuation can be realized. The current intelligent information service mainly considers user preference factors, generally combines two aspects of user preference and commodity characteristics to screen service contents for users and pushes the service contents to the users, but the mode ignores the influence of behavior motivation related to geographic characteristics on the service contents, and causes the loss of the information service contents in the aspects of individuation and accuracy. According to statistics, 80% of human activity information is related to the geographic position, and if the geographic characteristics and the user characteristics are fused to infer a user behavior motivation and apply the user behavior motivation to the intelligent service process, the personalization and the accuracy of service content can be effectively improved.
Disclosure of Invention
The invention provides a user behavior motivation inference algorithm based on fusion of geographic characteristics and user preference characteristics, aiming at the problem that intelligent service is lost in personalization and precision due to neglect of user behavior motivation, and by carrying out correlation analysis on the geographic characteristics and the user preference, the potential behavior motivation of a user is inferred and added into the recommendation process of the intelligent information service, so that the purpose of improving the personalization and precision of the intelligent service is realized.
In order to achieve the purpose, the invention constructs two large sets of geographic characteristics and user preference characteristics, and obtains a comprehensive set of the geographic characteristics and the user preference characteristics by processing and performing correlation analysis on the sets, wherein the comprehensive set is an inference list of user behavior motivation, and the obvious behavior motivation of the user can be inferred according to the list. The specific process of the inference algorithm is as follows:
(1) extracting geographic features and user preference features
The geographic characteristic and user preference characteristic source data are respectively from national POI (Point of Interest) data of a Goodpasture map and Internet behavior data of a user. The geographic features are obtained by processing the position as a key word and the type of the POI as a variable through three steps of statistics, screening and weight calculation. The method comprises the steps of firstly decomposing behavior data of a user into unit behaviors containing contents in six aspects of objects, time, places, contents, behaviors and the number of times of re-return, then calculating behavior weights according to the six elements, finally modeling user interest of the data, and finally constructing a 'user portrait' to obtain user preference characteristics.
(2) Constructing a set of geographic features and user-preferred features
And (2) constructing the geographic features and the user preference features extracted in the step (1) into feature sets, wherein each feature item is composed of a feature label and a feature weight, when the geographic features and the user preference features are constructed, the feature labels in the two sets adopt uniform normalized phrases, and the feature items in the sets are arranged in a descending order according to the weights. And recording the geographic feature set as Geo and the user preference feature set as Pre, then:
Geo＝{Geo1,Geo2,Geo3,...,Geom}
Pre＝{Pre1,Pre2,Pre3,...,Pren}
wherein, Geoi＝{LGeoi,PGeoi},LGeoiRepresenting a geographical feature tag, PGeoiA weight representing the geographic feature; prei＝{Lprei,PPrei},LPreiFeature tag, PPre, representing user preferencesiA weight representing the user preference characteristic.
(3) Intersecting geographic features with a set of user-preferred features
And (3) performing intersection processing on the geographic features and the user preference feature set constructed in the step (2), wherein the specific strategy is to perform normalization processing on feature tags in the Geo set and the Pre set by taking the feature tags as a judgment basis, namely only keeping set elements shared by the Geo set and the Pre set. The geographical feature and the user preference feature set after intersection are respectively set as Geo 'and Pre', and the calculation formula is as follows
Geo'＝Geo-(Geo-(Geo∩Pre))
Pre'＝Pre-(Pre-(Geo∩Pre))
(4) Multiplying a geographical feature by a set of user-preferred features
After the processing of the step (3), the two sets of Geo 'and Pre' reach unification in the quantity of the feature labels and the feature items, when the product operation is carried out on Geo 'and Pre', only the feature weight with the same label in the two sets is calculated, the label is unchanged when the product operation is carried out, the weight after the product operation is the product of the label weights in the two sets, and finally, a product set Comprehensive is obtained as follows:
Comprehensive＝Geo'×Pre'
＝{Com1,Com2,Com3,...,Comk}
wherein:
Comi＝{LComi,PComi}
＝{LGeoi+LPrei,PGeoi×PPrei}
in the above formula, LGeoi+LPreiThe comprehensive label representing the geographic characteristic and the user preference characteristic is normalized before product calculation, so that the two characteristic labels are taken as any one in actual representation, namely PGeoi×PPreiRepresenting a composite weight of the geographic feature and the user-preferred feature.
When the product operation is performed on the Geo 'and Pre' sets, in order to prevent feature distortion, when the product is performed on two same features, the difference between the two features is compared with the difference between the average values of the two sets, if the former is greater than the latter, the weight of the comprehensive feature is the product of the value with larger weight of the label in the two sets multiplied by the proportion occupied by the label in the two labels and then multiplied by the weight in the preference set, and if the former is less than or equal to the latter, the weight of the comprehensive feature is the product of the weights of the labels in the two sets, and the judgment process is as follows:
let two characteristic Geo exist in Geo 'and Pre' setsiAnd PreiThe two feature tags are identical in content, i.e., LGeoi＝LPreiThe weights are PGeoiAnd PPrei。
② calculating the average value of the characteristic weight of the set Geo
Figure GDA0003104221860000031
The feature weight average of the set Pre' is
Figure GDA0003104221860000032
When Geo 'and Pre' are multiplied, the feature GeoiAnd PreiThe integrated weight values of (a) are as follows:
Figure GDA0003104221860000033
wherein:
Figure GDA0003104221860000034
Figure GDA0003104221860000035
Figure GDA0003104221860000036
(5) ranking and feature screening
Sorting the Comprehensive set obtained in the step (4) in a descending order according to the weight of the set elements, and removing the features with smaller weight, namely feature screening, wherein the processing flow of the feature screening is as follows:
calculating average value of Comprehensive set
Figure GDA0003104221860000037
Figure GDA0003104221860000038
② feature screening, if
Figure GDA0003104221860000041
The feature is retained if
Figure GDA0003104221860000042
The feature is rejected.
(6) Determination of user's significant behavioral motivation
Judging the possibility of the user behavior motivation according to the feature weight, wherein the feature retained in the Comprehensive set obtained in the step (5) is the significant behavior motivation of the user, and the greater the feature weight is, the more the user behavior motivation is biased to the feature, otherwise, the more the user feature deviates from the feature.
Drawings
FIG. 1 is a flow diagram of user behavioral motivation inference.
Detailed Description
Example (b):
the method is suitable for deducing the potential behavior motivation of the user according to the geographic characteristics and the user preference characteristics and acting on the recommendation process of the intelligent service. For convenience of illustration, the present embodiment is performed in the flow sequence of fig. 1, and the specific implementation steps are as follows:
(1) extracting geographic features and user preference features
The geographic characteristics of the present example are selected from the geographic characteristics of the Sichuan city, the seven-star district, Guilin city, and the geographic characteristics are obtained through statistics and analysis by retrieving a POI database. The user preference characteristics adopt personal experimental data, including 5 characteristic items of food, entertainment, shopping, science and technology and reading.
(2) And (3) constructing geographic feature and user preference feature sets according to the data acquired in the step (1), wherein each set comprises 5 features of food, entertainment, shopping, movies, science and technology and reading.
Geo { { gou: 0.9}, { entertainment: 0.8}, { shopping: 0.7}, { science and technology: 0.6}, { movie: 0.5}}
Pre { { entertainment: 0.8}, { gourmet: 0.7}, { read: 0.6}, { shopping: 0.2}, { science and technology: 0.1}}
(3) Intersecting geographic features with a set of user-preferred features
Geo ≈ Pre { { gourmet: tmp1}, { entertainment: tmp2}, { shopping: tmp3}, { science and technology: tmp4}
In the above formula, Tmp temporary substitution weight is used to find intersection so as to screen out feature items with the same feature labels.
Solving the geographical feature set after intersection:
Geo'＝Geo-(Geo-(Geo∩Pre))
{ { cate: 0.9}, { entertainment: 0.8}, { shopping: 0.7}, { science and technology: 0.6}}
Solving a user preference feature set after intersection:
Pre'＝Pre-(Pre-(Geo∩Pre))
{ { entertainment: 0.8}, { gourmet: 0.7}, { shopping: 0.2}, { science and technology: 0.1}}
(4) Multiplying a geographical feature by a set of user-preferred features
First, the average of the two sets of feature weights is calculated:
Figure GDA0003104221860000051
Figure GDA0003104221860000052
then, the absolute value of the difference between the two sets of weight averages is calculated:
Figure GDA0003104221860000053
finally, the product of the two sets is calculated according to the product rule:
analysis of the elements in Geo 'and Pre' revealed that the difference between the absolute values of the shopping features in the two sets was 0.7-0.2-0.5, and the difference between the absolute values of the scientific features was 0.6-0.1-0.5, both greater than
Figure GDA0003104221860000054
And
Figure GDA0003104221860000055
the absolute value of the difference is 0.3, so the overall weight of the shopping feature in the overall set is
Figure GDA0003104221860000056
The combined weight of the scientific and technological characteristics is
Figure GDA0003104221860000057
The weight of the other feature is the product of the weights of the corresponding features of the two sets. The set of products is calculated as follows:
comprehensive { { entertainment: 0.64}, { gourmet: 0.63}, { shopping: 0.13}, { science and technology: 0.10}}
(5) Ranking and feature screening
The weight of the product set Comrehenive is calculated:
Figure GDA0003104221860000058
keeping the features with the feature weight larger than compressive, and arranging the features in descending order according to the weight, wherein the list of the finally inferred user behavior motivations is as follows:
comprehensive { { entertainment: 0.64}, { gourmet: 0.63}}
(6) Determination of user's significant behavioral motivation
Through the above process analysis and calculations, it is finally determined that the user's set of significant behavioral incentives is Comprehensive { { entertainment: 0.64}, { gourmet: 0.63}, i.e., the motivation of the user for this line may be entertainment and food, and since the weight of the entertainment feature is 0.64, which is greater than the feature weight of food 0.63, the most significant motivation of the user for the line is entertainment, which is inferior to food.

Claims (1)

1. A behavior motivation inference method fusing user preference characteristics and geographic characteristics is characterized by comprising the following specific steps:
(1) extracting geographic features and user preference features
Extracting geographic features and user preference feature source data from national POI data of the Goodpasture map and Internet behavior data of the user; the geographic features are obtained by processing the position as a keyword, the type of the POI as a variable and the three steps of statistics, screening and weight calculation; the method comprises the steps of obtaining user preference characteristics, firstly, decomposing behavior data of a user into unit behaviors containing contents in six aspects of objects, time, places, contents, behaviors and repeated times, then calculating behavior weights according to the six elements, finally, carrying out user interest modeling on the data, and finally constructing a 'user portrait' to obtain the user preference characteristics;
(2) constructing a set of geographic features and user-preferred features
Constructing the geographic features and the user preference features extracted in the step (1) into feature sets, wherein each feature item is composed of a feature label and a feature weight, when the geographic features and the user preference features are constructed, the feature labels in the two sets adopt uniform normalized phrases, and the feature items in the sets are arranged in a descending order according to the weights; and recording the geographic feature set as Geo and the user preference feature set as Pre, then:
Geo＝{Geo1,Geo2,Geo3,...,Geom}；
Pre＝{Pre1,Pre2,Pre3,...,Pren}；
wherein, Geoi＝{LGeoi,PGeoi},LGeoiRepresenting a geographical feature tag, PGeoiA weight representing the geographic feature; prei＝{Lprei,PPrei},LPreiFeature tag, PPre, representing user preferencesiA weight representing the user preference characteristic;
(3) intersecting geographic features with a set of user-preferred features
Performing intersection processing on the geographic features and the user preference feature set constructed in the step (2), wherein the specific strategy is to perform normalization processing on feature tags in the Geo set and the Pre set by taking the feature tags as a judgment basis, namely only keeping set elements shared by the Geo set and the Pre set; the geographical feature and the user preference feature set after intersection are respectively set as Geo 'and Pre', and the calculation formula is as follows
Geo'＝Geo-(Geo-(Geo∩Pre))；
Pre'＝Pre-(Pre-(Geo∩Pre))；
(4) Multiplying a geographical feature by a set of user-preferred features
After the processing of the step (3), the two sets of Geo 'and Pre' reach unification in the quantity of the feature labels and the feature items, when the product operation is carried out on Geo 'and Pre', only the feature weight with the same label in the two sets is calculated, the label is unchanged when the product operation is carried out, the weight after the product operation is the product of the label weights in the two sets, and finally, a product set Comprehensive is obtained as follows:
Comprehensive＝Geo'×Pre'
＝{Com1,Com2,Com3,...,Comk}；
wherein:
Comi＝{LComi,PComi}
＝{LGeoi+LPrei,PGeoi×PPrei}；
in the above formula, LGeoi+LPreiThe comprehensive label representing the geographic characteristic and the user preference characteristic is normalized before product calculation, so that the two characteristic labels are taken as any one in actual representation, namely PGeoi×PPreiA composite weight representing the geographic characteristic and the user preference characteristic;
when the product operation is performed on the Geo 'and Pre' sets, in order to prevent feature distortion, when the product is performed on two same features, the difference between the two features is compared with the difference between the average values of the two sets, if the former is greater than the latter, the weight of the comprehensive feature is the product of the value with larger weight of the label in the two sets multiplied by the proportion occupied by the label in the two labels and then multiplied by the weight in the preference set, and if the former is less than or equal to the latter, the weight of the comprehensive feature is the product of the weights of the labels in the two sets, and the judgment process is as follows:
let two characteristic Geo exist in Geo 'and Pre' setsiAnd PreiThe two feature tags are identical in content, i.e., LGeoi＝LPreiThe weights are PGeoiAnd PPrei；
② calculating the average value of the characteristic weight of the set Geo
Figure FDA0003072597650000021
The feature weight average of the set Pre' is
Figure FDA0003072597650000022
When Geo 'and Pre' are multiplied, the feature GeoiAnd PreiThe integrated weight values of (a) are as follows:
Figure FDA0003072597650000023
wherein:
Figure FDA0003072597650000024
Figure FDA0003072597650000025
Figure FDA0003072597650000026
(5) ranking and feature screening
Sorting the Comprehensive set obtained in the step (4) in a descending order according to the weight of the set elements, and removing the features with smaller weight, namely feature screening, wherein the processing flow of the feature screening is as follows:
calculating average value of Comprehensive set
Figure FDA0003072597650000031
Figure FDA0003072597650000032
② feature screening, if
Figure FDA0003072597650000033
The feature is retained if
Figure FDA0003072597650000034
Then the feature is rejected;
(6) determination of user's significant behavioral motivation
Judging the possibility of the user behavior motivation according to the feature weight, wherein the feature retained in the Comprehensive set obtained in the step (5) is the user's significant behavior motivation, and the greater the feature weight is, the more the user behavior motivation is biased to the feature, otherwise, the more the user feature deviates from the feature, thereby determining the user's significant behavior motivation and finally realizing the inference of the behavior motivation.
CN201810910570.2A 2018-08-11 2018-08-11 Behavior motivation inference algorithm integrating user preference and geographic features Active CN109325186B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810910570.2A CN109325186B (en) 2018-08-11 2018-08-11 Behavior motivation inference algorithm integrating user preference and geographic features

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810910570.2A CN109325186B (en) 2018-08-11 2018-08-11 Behavior motivation inference algorithm integrating user preference and geographic features

Publications (2)

Publication Number Publication Date
CN109325186A CN109325186A (en) 2019-02-12
CN109325186B true CN109325186B (en) 2021-08-17

Family

ID=65263315

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810910570.2A Active CN109325186B (en) 2018-08-11 2018-08-11 Behavior motivation inference algorithm integrating user preference and geographic features

Country Status (1)

Country Link
CN (1) CN109325186B (en)

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8489445B1 (en) * 2008-12-19 2013-07-16 Amazon Technologies, Inc. Determining and displaying item preferences associated with geographic areas
CN103631813A (en) * 2012-08-24 2014-03-12 富士通株式会社 Device and method for site searching and electronic device
CN104166722A (en) * 2014-08-22 2014-11-26 中国联合网络通信集团有限公司 Website recommending method and device
CN104699732A (en) * 2013-12-05 2015-06-10 联想（新加坡）私人有限公司 Method for forming user profile and information processing equipment
CN105046601A (en) * 2015-07-09 2015-11-11 传成文化传媒（上海）有限公司 User data processing method and system
CN105205699A (en) * 2015-09-17 2015-12-30 北京众荟信息技术有限公司 User label and hotel label matching method and device based on hotel comments
CN107291860A (en) * 2017-06-09 2017-10-24 北京邮电大学 Seed user determines method
CN107885784A (en) * 2017-10-17 2018-04-06 北京京东尚科信息技术有限公司 The method and apparatus for extracting user characteristic data

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8489445B1 (en) * 2008-12-19 2013-07-16 Amazon Technologies, Inc. Determining and displaying item preferences associated with geographic areas
CN103631813A (en) * 2012-08-24 2014-03-12 富士通株式会社 Device and method for site searching and electronic device
CN104699732A (en) * 2013-12-05 2015-06-10 联想（新加坡）私人有限公司 Method for forming user profile and information processing equipment
CN104166722A (en) * 2014-08-22 2014-11-26 中国联合网络通信集团有限公司 Website recommending method and device
CN105046601A (en) * 2015-07-09 2015-11-11 传成文化传媒（上海）有限公司 User data processing method and system
CN105205699A (en) * 2015-09-17 2015-12-30 北京众荟信息技术有限公司 User label and hotel label matching method and device based on hotel comments
CN107291860A (en) * 2017-06-09 2017-10-24 北京邮电大学 Seed user determines method
CN107885784A (en) * 2017-10-17 2018-04-06 北京京东尚科信息技术有限公司 The method and apparatus for extracting user characteristic data

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Understanding Geographical Markets of Online Firms Using Spatial Models of Customer Choice;Wolfgang Jank;《https://doi.org/10.1287/mksc.1050.0145》;20051130;第24卷(第4期);第525-648页 *
地理社交网络数据可视化分析系统的设计、开发与应用;谢妮;《中国优秀硕士学位论文全文库 基础科学辑》;20161215;A008-7 *

Also Published As

Publication number Publication date
CN109325186A (en) 2019-02-12

Similar Documents

Publication Publication Date Title
Sánchez-Núñez et al. Opinion mining, sentiment analysis and emotion understanding in advertising: a bibliometric analysis
Lei et al. Rating prediction based on social sentiment from textual reviews
KR101871747B1 (en) Similarity tendency based user-sightseeing recommendation system and method thereof
CN111898031B (en) Method and device for obtaining user portrait
CN107077486A (en) Affective Evaluation system and method
CN104239399B (en) Potential friend recommendation method in social networks
JP2018169986A (en) Enterprise information providing system and program
Shen et al. A voice of the customer real-time strategy: An integrated quality function deployment approach
Dai et al. BTR: a feature-based Bayesian task recommendation scheme for crowdsourcing system
Sun et al. Leveraging friend and group information to improve social recommender system
CN112765482A (en) Product delivery method, device, equipment and computer readable medium
Dang et al. Adoption of social media search systems: An IS success model perspective
CN111966888A (en) External data fused interpretable recommendation method and system based on aspect categories
CN110851694A (en) Personalized recommendation system based on user memory network and tree structure depth model
Cai et al. An extension of social network group decision-making based on trustrank and personas
Han et al. Developing smart service concepts: morphological analysis using a Novelty-Quality map
Seroussi Utilising user texts to improve recommendations
CN109325186B (en) Behavior motivation inference algorithm integrating user preference and geographic features
JP6696270B2 (en) Information providing server device, program and information providing method
Tuma et al. Online reviews as a source of marketing research data: a literature analysis
Liu et al. User-generated content analysis for customer needs elicitation
Pujahari et al. A new grouping method based on social choice strategies for group recommender system
KR101549188B1 (en) Apparatus and method for measuring brand image
CN114429384A (en) Intelligent product recommendation method and system based on e-commerce platform
CN114168790A (en) Personalized video recommendation method and system based on automatic feature combination

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
EE01 Entry into force of recordation of patent licensing contract
EE01 Entry into force of recordation of patent licensing contract

Application publication date: 20190212

Assignee: Guilin Tianshi Technology Co.,Ltd.

Assignor: GUILIN University OF TECHNOLOGY

Contract record no.: X2022450000162

Denomination of invention: A Behavior Motivation Inference Algorithm Based on the Fusion of User Preferences and Geographic Features

Granted publication date: 20210817

License type: Common License

Record date: 20221124

Application publication date: 20190212

Assignee: Guilin Yidoumi Information Technology Co.,Ltd.

Assignor: GUILIN University OF TECHNOLOGY

Contract record no.: X2022450000128

Denomination of invention: A Behavior Motivation Inference Algorithm Based on the Fusion of User Preferences and Geographic Features

Granted publication date: 20210817

License type: Common License

Record date: 20221123