CN106228264A - A kind of personalized place route recommendation method with time correlation - Google Patents
A kind of personalized place route recommendation method with time correlation Download PDFInfo
- Publication number
- CN106228264A CN106228264A CN201610570769.6A CN201610570769A CN106228264A CN 106228264 A CN106228264 A CN 106228264A CN 201610570769 A CN201610570769 A CN 201610570769A CN 106228264 A CN106228264 A CN 106228264A
- Authority
- CN
- China
- Prior art keywords
- place
- user
- time
- registering
- recommended
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 16
- 230000002123 temporal effect Effects 0.000 claims abstract description 25
- 238000000354 decomposition reaction Methods 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 description 5
- PEDCQBHIVMGVHV-UHFFFAOYSA-N Glycerine Chemical group OCC(O)CO PEDCQBHIVMGVHV-UHFFFAOYSA-N 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
- G06Q10/047—Optimisation of routes or paths, e.g. travelling salesman problem
-
- 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
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/01—Social networking
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention discloses a kind of and time correlation personalized place route recommendation method, it is achieved step is as follows, and (1) builds three-dimensional tensor of registering;(2) tensor of registering that reconstruct is three-dimensional;(3) candidate locations list is obtained;(4) the coarseness Annual distribution in place and fine-grained Annual distribution are obtained;(5) the coarseness temporal information divergence in place is calculated;(6) the fine granularity temporal information divergence in place is calculated;(7) probability of user's access locations to be recommended is calculated;(8) recommended route is obtained.The method of the invention is according to the time that the temporal information digging user in user interest and place is interested effective place, making user be met the route recommendation of personal interest, the present invention may be used in the place route Push Service in social networks based on place.
Description
Technical field
The invention belongs to technical field of data processing, further relate to the one in social networks technical field and time
Relevant personalized place route recommendation method.Present invention can apply to push clothes based on the place route in the social networks of position
Business.
Background technology
Along with information technology and the development of mobile terminal technology, and wireless network and the appearance of location aware technology and enter
Step, location Based service develops rapidly, such as Foursquare.Location Based service can be by record of registering
Follow the tracks of the space motion track of user.Meanwhile, along with the expansion of urbanization, the place quantity comprised in city gets more and more, and uses
Certain region, city may be simply understood and is familiar with at family.Other regions in user arrives city, need service provider to enter
Row route recommendation.Have the personal information relevant with user and the historical record of magnanimity, utilize these data actively to push away to user
On the one hand recommend related service can help user to make a choice, and finds their interested, valuable place, and then by these
Place composition route recommendation, to user, improves Consumer's Experience.
Master's thesis " personalized Path mining based on user preference the research " (Master's thesis that Jiang Xiaoling delivers at it
Northeastern University in 2013) in propose a kind of interest preference based on pictorial information digging user go forward side by side walking along the street footpath recommend side
Method.The interest-degree of user is carried out extracting quantization by the method by introducing TF-IDF algorithm.With class belonging to the point of interest in route
Route Dai Ti not described point of interest, and using point of interest category for the contribution degree of user as user's interest-degree to the category,
Propose evaluation function based on user interest degree, it is achieved that personalized route recommendation.The weak point that the method exists is,
Recommendation results is carried out more single, lack of diversity by TF-IDF algorithm;Do not account for the temporal information that route accesses, it is recommended that
It is invalid that the route that system produces is probably.
The patent " a kind of route recommendation method and system " that Baidu In Line Network Technology Co Ltd (Beojing) applies at it
(application number 201410449757.9, publication number CN104197947A) discloses one and is applicable to user in two states switching
Route recommendation method and system under situation, under the first state of mobile device, the preference information according to user obtains at least one
Bar the first recommendation paths;When determining that described mobile device is become the second state by described first State Transferring by needs, according to institute
State preference information and obtain at least one the second recommendation paths in said second condition.The weak point that the method exists is:
During preference information according to user carries out path recommendation, do not account for the current temporal information of user and the time of route
Information, it is impossible to provide the user the route meeting time requirement well.
Summary of the invention
It is an object of the invention to overcome the deficiency of above-mentioned prior art, propose the personalized place road of a kind of time correlation
Method recommended by line, has considered user interest and time context, has improve accuracy and the reliability of recommendation results, can answer
For based on the place route Push Service in the social networks of position.
The concrete thought of the present invention is: to existing data set, sets up the three-dimensional tensor model of user-place-time, meter
Calculate the interest that user puts over the ground, the individual preference of digging user in certain time.In conjunction with the Annual distribution in place, by calculating place
Coarse grain information divergence and fine granularity information divergence, make use of temporal information.The individual preference of synthetic user and temporal information,
Obtain position candidate list.By using longest path planning algorithm, it is achieved that personalized route recommendation.
For achieving the above object, the present invention comprises the following steps:
(1) build three-dimensional to register tensor:
(1a) from data set, obtain the data of registering of all users in city to be recommended;
(1b) from data of registering, each user scoring to any one place section at any time is extracted respectively;
(1c) all of scoring is built into the three-dimensional of user-place-time register tensor;
(2) reconstruct three-dimensional tensor:
(2a) three-dimensional is registered tensor, use Higher-order Singular value decomposition algorithm, calculate all users at any time
Scoring to all places during section;
(2b) all of scoring is constituted the three-dimensional tensor of user-place-time;
(3) candidate locations list is obtained:
(3a) by the locating module that user mobile phone is built-in, the current location of user to be recommended is obtained;
(3b) according to the current location of user to be recommended, city, user place to be recommended is determined, and using this city as treating
Recommend city;
(3c) from the three-dimensional tensor of reconstructing user-place-time, the scoring collection in all places, city to be recommended is extracted
Close;
(3d) to scoring set, being ranked up from high to low according to scoring, K scoring before selecting from sequence, by selected
Take place corresponding to scoring as candidate locations list;
(4) Annual distribution is obtained:
(4a) by the data of registering in city to be recommended, classify according to place name, in statistics candidate locations list
The number of times of registering in each place;
(4b) data of registering in the city to be recommended in month each in the 1 year time of registering are added up, obtain candidate ground
Each place in point list is at the number of times of registering in different months;
(4c) data of registering in the city to be recommended of each hour in the one day time of registering are added up, obtain candidate ground
Each place in point list was at the number of times of registering of each hour;
(4d) with each place the registering time of each place in the number of times of registering in different months is divided by candidate locations list
Number, obtains the coarseness Annual distribution probability in place;
(4e) with each place the registering time of each place in the number of times of registering of different hours is divided by candidate locations list
Number, obtains the fine granularity Annual distribution probability in place;
(5) the coarseness temporal information divergence in calculating place:
Use the information divergence computing formula of coarseness time, the coarseness time in the place in calculating candidate locations list
Information divergence value;
(6) the fine granularity temporal information divergence in calculating place:
Use the information divergence computing formula of fine granularity time, the fine granularity time in the place in calculating candidate locations list
Information divergence;
(7) probit of user's access locations to be recommended according to the following formula, is calculated:
sj=ecj-b·(ckj·fkj)
Wherein, sjRepresent the probit of user's access locations j to be recommended, ecjRepresent that place j is commented by user c to be recommended
Point, cjRepresent the coarseness temporal information divergence of place j, fjRepresent place j fine granularity temporal information divergence, b represent control because of
Son, its value is in the range of b ∈ [0,1];
(8) place route recommendation is obtained:
Use longest path algorithm, the place in the probability and candidate locations list of user's access locations as input,
It is met the place route recommendation of user interest.
The present invention has the advantage that compared with prior art
First, present invention utilizes coarseness time and the fine granularity temporal information in place, be used for calculating the coarse grain in place
Degree temporal information divergence and fine granularity temporal information divergence, has filtered the place that divergence value is zero, does not eliminates the time the most not
The place joined, overcomes prior art and can not meet the ineffectually some produced problem of time requirement so that the present invention improves
The accuracy of recommendation results.
Second, the present invention combines user's scoring to place and the temporal information divergence in place, calculates user to be recommended
The probability of access locations, employs interest information rather than the location information of user, overcomes prior art recommendation results list
The problem of one so that the present invention improves the multiformity of recommendation results.
Accompanying drawing explanation
Fig. 1 is the flowchart of the present invention.
Detailed description of the invention
Below in conjunction with the accompanying drawings 1, the step of the present invention is described in further detail.
Step 1, builds three-dimensional tensor of registering.
From disclosed Foursquare data set, obtain the data of registering of all users in city to be recommended.From number of registering
According to, extract each user scoring to any one place section at any time respectively.All of scoring is built into user-
The three-dimensional of place-time is registered tensor.
Step 2, reconstructs three-dimensional tensor.
Three-dimensional is registered tensor, use Higher-order Singular value decomposition algorithm, when calculating all users to section at any time
Scoring to all places;All of scoring is constituted the three-dimensional tensor of user-place-time.
Step 3, obtains candidate locations list.
By the locating module that user mobile phone is built-in, obtain the current location of user to be recommended;According to user's to be recommended
Current location, determines city, user place to be recommended, and using this city as city to be recommended;From build user-place-time
Between three-dimensional tensor in, extract the scoring set in all places, city to be recommended;To scoring set, enter from high to low according to scoring
Row sequence, K scoring before selecting from sequence, using place corresponding for selected scoring as candidate locations list, the value model of K
Enclosing is 30-50.
Step 4, obtains Annual distribution.
The data of registering in city to be recommended are classified according to place name, statistics candidate locations list in eachly
The number of times of registering of point;The data of registering in the city to be recommended in month each in the 1 year time of registering are added up, obtains candidate
Each place in list of localities is at the number of times of registering in different months;To the city to be recommended of each hour in the one day time of registering
Data of registering add up, obtain the number of times of registering that each place in candidate locations list was at each hour;With each
Point is the number of times of registering in each place in the number of times of registering in different months is divided by candidate locations list, when obtaining the coarseness in place
Between distribution probability;With each place the registering time of each place in the number of times of registering of different hours is divided by candidate locations list
Number, obtains the fine granularity Annual distribution probability in place.
Step 5, calculates the coarseness temporal information divergence in place.
Use the information divergence computing formula of coarseness time, the coarseness time in the place in calculating candidate locations list
Information divergence value;The information divergence computing formula of coarseness time is as follows:
Wherein, cjRepresenting the information divergence of the coarseness time of place j, ∑ represents that sum operation, x represent the coarseness time
Corresponding month value, G () represents Gauss distribution, and log represents operation of taking the logarithm with 2 end of for, djX () represents that place j is in the moon
Part value is coarseness Annual distribution probit during x.
Step 6, calculates the fine granularity temporal information divergence in place.
Use the information divergence computing formula of fine granularity time, the fine granularity time in the place in calculating candidate locations list
Information divergence;The information divergence computing formula of fine granularity time is as follows:
Wherein, fjRepresenting the information divergence of the fine granularity time of place j, ∑ represents that sum operation, y represent the fine granularity time
Corresponding one hour value, G () represents Gauss distribution, and log represents operation of taking the logarithm with 2 end of for, kjY () represents that place j is little
Time value be fine granularity Annual distribution probit during y.
Step 7, according to the following formula, calculates the probit of user's access locations to be recommended:
sj=ecj-b·(ckj·fkj)
Wherein, sjRepresent the probit of user's access locations j to be recommended, ecjRepresent that place j is commented by user c to be recommended
Point, cjRepresent the coarseness temporal information divergence of place j, fjRepresent place j fine granularity temporal information divergence, b represent control because of
Son, its value is in the range of b ∈ [0,1].
Step 8, obtains recommended route.
With longest path algorithm, the place in the probability and candidate locations list of user's access locations, as input, obtains
To the place route recommendation meeting user interest.
Claims (4)
1., with the personalized place route recommendation method of time correlation, comprise the following steps:
(1) build three-dimensional to register tensor:
(1a) from data set, obtain the data of registering of all users in city to be recommended;
(1b) from data of registering, each user scoring to any one place section at any time is extracted respectively;
(1c) all of scoring is built into the three-dimensional of user-place-time register tensor;
(2) reconstruct three-dimensional tensor:
(2a) three-dimensional is registered tensor, use Higher-order Singular value decomposition algorithm, when calculating all users to section at any time
Scoring to all places;
(2b) all of scoring is constituted the three-dimensional tensor of user-place-time;
(3) candidate locations list is obtained:
(3a) by the locating module that user mobile phone is built-in, the current location of user to be recommended is obtained;
(3b) according to the current location of user to be recommended, city, user place to be recommended is determined, and using this city as to be recommended
City;
(3c) from the three-dimensional tensor of reconstructing user-place-time, the scoring set in all places, city to be recommended is extracted;
(3d) to scoring set, it is ranked up from high to low according to scoring, K scoring before selecting from sequence, comments selected
Divide corresponding place as candidate locations list;
(4) Annual distribution is obtained:
(4a) by the data of registering in city to be recommended, classify according to place name, statistics candidate locations list in each
The number of times of registering in place;
(4b) data of registering in the city to be recommended in month each in the 1 year time of registering are added up, obtain candidate locations row
Each place in table is at the number of times of registering in different months;
(4c) data of registering in the city to be recommended of each hour in the one day time of registering are added up, obtain candidate locations row
Each place in table was at the number of times of registering of each hour;
(4d) with each place number of times of registering in each place in the number of times of registering in different months is divided by candidate locations list,
Coarseness Annual distribution probability to place;
(4e) with each place number of times of registering in each place in the number of times of registering of different hours is divided by candidate locations list,
Fine granularity Annual distribution probability to place;
(5) the coarseness temporal information divergence in calculating place:
Use the information divergence computing formula of coarseness time, the coarseness temporal information in the place in calculating candidate locations list
Divergence value;
(6) the fine granularity temporal information divergence in calculating place:
Use the information divergence computing formula of fine granularity time, the fine granularity temporal information in the place in calculating candidate locations list
Divergence;
(7) probit of user's access locations to be recommended according to the following formula, is calculated:
sj=ecj-b·(ckj·fkj)
Wherein, sjRepresent the probit of user's access locations j to be recommended, ecjRepresent the user c to be recommended scoring to place j, cjTable
Show the coarseness temporal information divergence of place j, fjRepresenting the fine granularity temporal information divergence of place j, b represents controlling elements, its
Value is in the range of b ∈ [0,1];
(8) place route recommendation is obtained:
Using longest path algorithm, the place in the probability and candidate locations list of user's access locations, as input, obtains
Meet the place route recommendation of user interest.
A kind of personalized place route recommendation method with time correlation the most according to claim 1, it is characterised in that: step
Suddenly the span of the K described in (3d) is 30-50.
A kind of personalized place route recommendation method with time correlation the most according to claim 1, it is characterised in that: step
Suddenly the information divergence computing formula of the coarseness time described in (5) is as follows:
Wherein, cjRepresenting the information divergence of the coarseness time of place j, ∑ represents that sum operation, x represent that the coarseness time is right
The month value answered, G () represents Gauss distribution, and log represents operation of taking the logarithm with 2 end of for, djX () represents that place j took in month
Value is coarseness Annual distribution probit during x.
A kind of personalized place route recommendation method with time correlation the most according to claim 1, it is characterised in that: step
Suddenly the information divergence computing formula of the fine granularity time described in (6) is as follows:
Wherein, fjRepresenting the information divergence of the fine granularity time of place j, ∑ represents that sum operation, y represent that the fine granularity time is right
The one hour value answered, G () represents Gauss distribution, and log represents operation of taking the logarithm with 2 end of for, kjY () represents that place j hour is taking
Value is fine granularity Annual distribution probit during y.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610570769.6A CN106228264A (en) | 2016-07-19 | 2016-07-19 | A kind of personalized place route recommendation method with time correlation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610570769.6A CN106228264A (en) | 2016-07-19 | 2016-07-19 | A kind of personalized place route recommendation method with time correlation |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106228264A true CN106228264A (en) | 2016-12-14 |
Family
ID=57531693
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610570769.6A Pending CN106228264A (en) | 2016-07-19 | 2016-07-19 | A kind of personalized place route recommendation method with time correlation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106228264A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109425355A (en) * | 2017-09-05 | 2019-03-05 | 上海博泰悦臻网络技术服务有限公司 | Recommended method of navigating and system, car-mounted terminal and vehicle |
CN112183863A (en) * | 2020-09-29 | 2021-01-05 | 上海交通大学 | Fine-grained taxi route recommendation method, system and medium based on gravity model |
-
2016
- 2016-07-19 CN CN201610570769.6A patent/CN106228264A/en active Pending
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109425355A (en) * | 2017-09-05 | 2019-03-05 | 上海博泰悦臻网络技术服务有限公司 | Recommended method of navigating and system, car-mounted terminal and vehicle |
CN112183863A (en) * | 2020-09-29 | 2021-01-05 | 上海交通大学 | Fine-grained taxi route recommendation method, system and medium based on gravity model |
CN112183863B (en) * | 2020-09-29 | 2022-03-25 | 上海交通大学 | Fine-grained taxi route recommendation method, system and medium based on gravity model |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Sinclair et al. | Recreation and environmental quality of tropical wetlands: A social media based spatial analysis | |
CN105532030B (en) | For analyzing the devices, systems, and methods of the movement of target entity | |
Šćepanović et al. | Mobile phone call data as a regional socio-economic proxy indicator | |
Montini et al. | Trip purpose identification from GPS tracks | |
Zhong et al. | Inferring building functions from a probabilistic model using public transportation data | |
Nuzzolo et al. | Advanced public transport and intelligent transport systems: new modelling challenges | |
Chandra et al. | Predictions of freeway traffic speeds and volumes using vector autoregressive models | |
Xia et al. | Modelling spatio-temporal movement of tourists using finite Markov chains | |
Gontier et al. | Comparing GIS-based habitat models for applications in EIA and SEA | |
Yu et al. | Modeling user activity patterns for next-place prediction | |
Hardy et al. | Using innovations to understand tourist mobility in national parks | |
Landis et al. | A general risk‐based adaptive management scheme incorporating the Bayesian Network Relative Risk Model with the South River, Virginia, as case study | |
JP7175783B2 (en) | Program, device and method for estimating stay area population in predetermined area | |
Randin et al. | Introduction of snow and geomorphic disturbance variables into predictive models of alpine plant distribution in the Western Swiss Alps | |
Noyman et al. | Reversed urbanism: Inferring urban performance through behavioral patterns in temporal telecom data | |
CN108875761A (en) | A kind of method and device for expanding potential user | |
Bwambale et al. | Modelling long-distance route choice using mobile phone call detail record data: a case study of Senegal | |
Hu et al. | An adaptive hawkes process formulation for estimating time-of-day zonal trip arrivals with location-based social networking check-in data | |
Bwambale et al. | Getting the best of both worlds: a framework for combining disaggregate travel survey data and aggregate mobile phone data for trip generation modelling | |
Amador-Jiménez et al. | Reliability-based initial pavement performance deterioration modelling | |
Altieri et al. | Give public transit a chance: A comparative analysis of competitive travel time in public transit modal share | |
CN110322067A (en) | Location of mobile users prediction technique based on factor graph model | |
Radinger et al. | Spatial scaling of environmental variables improves species-habitat models of fishes in a small, sand-bed lowland river | |
EP3014491A1 (en) | Displaying demographic data | |
Dalumpines et al. | Making mode detection transferable: extracting activity and travel episodes from GPS data using the multinomial logit model and Python |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20161214 |
|
RJ01 | Rejection of invention patent application after publication |