CN108256914B - Interest point type prediction method based on tensor decomposition model - Google Patents

Interest point type prediction method based on tensor decomposition model Download PDF

Info

Publication number
CN108256914B
CN108256914B CN201810042721.7A CN201810042721A CN108256914B CN 108256914 B CN108256914 B CN 108256914B CN 201810042721 A CN201810042721 A CN 201810042721A CN 108256914 B CN108256914 B CN 108256914B
Authority
CN
China
Prior art keywords
user
tensor
interest
point
interest point
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
CN201810042721.7A
Other languages
Chinese (zh)
Other versions
CN108256914A (en
Inventor
王晓玲
贺韻宇
靳远远
刘坤
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.)
East China Normal University
Original Assignee
East China Normal University
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 East China Normal University filed Critical East China Normal University
Priority to CN201810042721.7A priority Critical patent/CN108256914B/en
Publication of CN108256914A publication Critical patent/CN108256914A/en
Application granted granted Critical
Publication of CN108256914B publication Critical patent/CN108256914B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0201—Market modelling; Market analysis; Collecting market data

Landscapes

  • Business, Economics & Management (AREA)
  • Strategic Management (AREA)
  • Engineering & Computer Science (AREA)
  • Accounting & Taxation (AREA)
  • Development Economics (AREA)
  • Finance (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Game Theory and Decision Science (AREA)
  • Data Mining & Analysis (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The invention discloses a tensor decomposition model-based interest point category prediction method, which comprises the following steps of: a) acquiring the coordinate information of an access point and the stay time of a user in T time periods every day from the new energy automobile track data; b) defining the range of a user staying area as an uncertain area according to the time period of the access point and the staying time length, acquiring the id of the interest point type in the area, and constructing a tensor; c) clustering the track data corresponding to all users; d) predicting the interest point types of the user by using a tensor decomposition model; e) fusing the information of the work, the position of the place of residence and the place of departure, using the similarity of the types of interest points between the place of departure and the destination as a tensor decomposition item constraint, and calculating a low-rank approximate solution of the tensor; f) and (4) iterative optimization of the target function by using an alternating least square method to obtain the probability of the user accessing different interest point types as a prediction result. By adopting the method and the device, the accurate prediction of the types of the interest points accessed by the user can be realized.

Description

Interest point type prediction method based on tensor decomposition model
Technical Field
The invention belongs to the field of mobile behavior analysis, and particularly relates to a point of interest category prediction method based on a tensor decomposition model.
Background
With the development of positioning technology and mobile computing, more and more trajectory data are collected and applied to various researches, which becomes an increasingly important research subject. The trajectory data reflects the movement rules, behavior preferences, even interest preferences, and the like of various moving objects such as people, vehicles, and the like. Meanwhile, research on user interest points is also receiving wide attention from the industrial and academic circles.
How to identify and predict the categories of the points of interest visited by the mobile user from the trajectory data is an important application in mobile behavior analysis. Since the reported position of the mobile device (e.g., the mobile station position, the social network check-in position, the vehicle GPS coordinates, the public transportation coordinates, etc.) is usually very inaccurate and has a large deviation from the user's real position, it is very challenging to determine the point of interest that the user finally visits. Predicting the interest point category accessed by the user by using the trajectory data of the new energy automobile, and describing interest preference and behavior habit of the existing user of the new energy automobile; the method has important practical significance in mining potential customers, popularizing new energy automobiles and the like. However, the above-mentioned problem that the deviation between the parking position and the actual position is large and the parking position is limited (in the parking lot or near the charging pile) also exists.
Tensor decomposition is essentially a method using the idea of collaborative filtering, and takes into account preference information of users of the same type, and mines the relationship between potential features and fills missing values, so that a method for predicting behavior preference of a certain user becomes an important method for prediction and recommendation. Research shows that auxiliary information such as the similarity between two dimensions has a good supplementary effect on a tensor decomposition model. The method adopting auxiliary information as a regularization term is less used in the field of mobile data prediction.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a tensor decomposition model-based interest point type prediction method to realize accurate prediction of interest point types of mobile users.
In order to achieve the above object, the method for predicting interest point categories based on a tensor decomposition model of the present invention includes the following steps:
s1: acquiring the coordinate information of an access point and the stay time of a user in T time periods every day from the new energy automobile track data;
s2: defining the range of a user staying area as an uncertain area according to the time period of the access point and the staying time length, acquiring the id of the interest point type in the area, and constructing a tensor
Figure BDA0001549956150000021
S3: clustering the track data corresponding to all users, wherein the track data comprises residential area clustering and working area clustering which are used as auxiliary information;
s4: predicting the interest point category of the user by utilizing a tensor decomposition model with a negative-unlabeled constraint;
s5: fusing auxiliary information of a place of work, a place of residence and a place of departure, using the similarity of the types of interest points between the place of departure and the destination as a conversion term constraint of tensor decomposition, and calculating tensor
Figure BDA0001549956150000022
A low rank approximate solution of;
s6: and (4) iterative optimization of the target function by using an alternating least square method to obtain the probability of the user accessing different interest point types as a prediction result.
The step S2 constructs the third order tensor as follows
Figure BDA0001549956150000023
The three dimensions of the tensor are user, time and interest point categories respectively, and the corresponding dimensions are N, T, C respectively; each element xijkThe value of (d) represents the probability that user i visits a point of interest of category k during time period j.
The step S3 uses a DBScan density clustering algorithm.
The negative-unlabeled constraint satisfied in step S4 is:
Figure BDA0001549956150000024
wherein, χijk∈[0,1]Representing the probability, Ω, that user i visits a point of interest of category k during time period jijAnd defining an uncertain region for the access point of the user i in the time period j.
The step S5 obtains a low rank approximate solution by the following tensor decomposition method:
Figure BDA0001549956150000031
in the formula (I), the compound is shown in the specification,
Figure BDA0001549956150000032
is F two norm, represents the square root of the sum of the squares of all elements,
Figure BDA0001549956150000033
for the estimated value obtained after iteration, U(n)Is a hidden vector matrix of the user with tensor x at the nth latitude, S is a similarity matrix for describing auxiliary information,
Figure BDA0001549956150000034
is Laplacian matrix, beta is weight value of auxiliary information matrix S, and lambda is weight value of user hidden vector matrix.
The interest point type prediction method based on the tensor decomposition model obtains user access point information from new energy automobile track data, an uncertain area is defined for each access point, and the type of the interest points in the uncertain area is obtained, so that a tensor is constructed. And then decomposing the tensor by combining the auxiliary information of the user and the interest point and the negative-unlabeled constraint, and filling the missing value and simultaneously obtaining a prediction result. By adopting the method and the device, the accurate prediction of the types of the interest points accessed by the user can be realized.
Drawings
FIG. 1 is a flowchart of the general structure of the interest point type prediction method based on tensor decomposition model.
Detailed Description
The following description of the embodiments of the present invention is provided in order to better understand the present invention for those skilled in the art with reference to the accompanying drawings.
S1: obtaining user access point coordinate information and residence time from new energy automobile track data, and obtaining a format as follows: and (4) access point information candidate set of < user ID, time, longitude and latitude and access duration >. Obtaining access point information of each user in a determined time period T according to the change of the vehicle state flag bit in the original track data;
s2: determining the area range which the user can access according to the time period of the access point and the length of the stay time, and then acquiring the interest point belonging to the area range by using a k-nearest neighbor algorithm, thereby obtaining the type ID corresponding to the interest point. The formula for calculating the distance between the two access points A, B is as follows:
Distance＝R×arccosθ；
where θ is cos LatA × cosLatB × cos (LonA-LonB) + sinLatA × sinLatB. Lat represents the latitude of the coordinate, Lon represents the longitude of the coordinate, and R represents the radius of the earth. Subsequently constructing the third order tensor
Figure BDA0001549956150000044
The three dimensions and the corresponding numbers are respectively: n users, T time periods, and C interest point categories. The tensor X is expanded into C matrices according to model-1, and the form is as follows:
Figure BDA0001549956150000041
wherein, the tensor size of the user-time-interest point category is N T C, xijkRepresenting the probability of a user i appearing on a point of interest category k in a time period j, e.g. x, with a value between 0 and 1111The value of (d) represents the magnitude of the probability that user 1 accesses the point of interest category 1 during time period 1.
S3: clustering the access points of all users, defining time periods of possible access to residential places and working places, and respectively carrying out DBScan density clustering on the access points of each user. Finding a direct density reachable point q in the neighborhood of any core point P, and calculating the density based on the following formula:
ρ(x)＝|Nε(x)|
wherein N isε(x) X represents a set of X fields.
S4: predicting the interest point category of the user by utilizing a tensor decomposition model with a negative-unlabeled constraint, wherein the negative-unlabeled constraint is as follows:
Figure BDA0001549956150000042
element χ in tensorijk∈[0,1]Represents the probability, Ω, that user i visits a point of interest of category k during time period jijThe uncertainty region where the access point of user i at time j is located. If the interest point category appears in the uncertain region corresponding to the time period j of the user i, the value is between 0 and 1, and the sum of the category probabilities in all the uncertain regions is 1. If the point of interest class appears outside the corresponding uncertainty region, its value is 0.
S5: fusing the similarity of the user auxiliary information and the types of interest points between the departure place and the destination as a regular term of tensor decomposition, and calculating the tensor
Figure BDA0001549956150000043
The calculation formula is as follows:
Figure BDA0001549956150000051
in the formula (I), the compound is shown in the specification,
Figure BDA0001549956150000052
is F two norm, represents the square root of the sum of the squares of all elements,
Figure BDA0001549956150000053
for the estimated value obtained after iteration, U(n)Is a hidden vector matrix of tensor x at the nth latitude, S is an auxiliary information similarity matrix,
Figure BDA0001549956150000057
is Laplacian matrix, beta is weight value of auxiliary information matrix S, and lambda is weight value of user hidden vector matrix.
Calculating the similarity of the interest point categories between the departure place and the destination of the departure place by adopting the following formula:
Figure BDA0001549956150000054
wherein, Depij、DesijRepresenting the departure place and the destination corresponding to the user i in the time period j; c represents interest point kind set in the access point uncertain area, n and U represent intersection and union of two sets, t represents specific time, and alpha is fission constant.
Calculating the similarity of the residence places of the users a and b by adopting the following formula:
Figure BDA0001549956150000055
wherein, dist (H)a-Hb) Representing the distance between the residence places of the user a and the user b; α is the fission constant.
Figure BDA0001549956150000056
Wherein, dist (W)a-Wb) Representing the distance between the user a and the user b; α is the fission constant.
S6: and (4) iterative optimization of the target function by using an alternating least square method to obtain the probability of the user accessing different interest point types as a prediction result.
Examples
As shown in fig. 1, the specific steps of the interest point category prediction method based on the tensor decomposition model of the present invention include:
s101: acquiring an access point set:
obtaining user access point coordinate information and residence time from new energy automobile track data, and obtaining a format as follows: and (4) access point information candidate set of < user ID, time, longitude and latitude and access duration >. The candidate set obtains access point information of each user in a determined time period T according to changes of vehicle state flag bits in the original trajectory data, and table 1 is a partial access point information candidate set in this embodiment.
TABLE 1
User ID Time Longitude (G) Dimension (d) of Duration/min
LGXC76C3******79 2015-09-15 13:34:07 121.430020 31.345190 20
LGXC76C3******79 2015-09-15 16:08:49 121.485560 31.314280 17
LGXC76C3******79 2015-09-16 07:47:34 121.448590 31.284560 442
LGXC76C3******47 2015-07-03 07:34:12 121.324320 31.289080 280
LGXC76C3******47 2015-07-03 13:50:54 121.378530 31.241330 1090
S102: defining an uncertainty region and constructing a tensor:
defining the range of the user staying area as an uncertain area according to the time period of the access point and the staying time length, and then acquiring the interest points belonging to the uncertain area range by using a k-nearest neighbor algorithm so as to obtain the type ID corresponding to the interest points. The types of interest points selected in this embodiment are 28, including: medical, school, government agency, tourist attraction, shopping center, shop, hotel, gourmet, company, recreational and recreational, transportation facility, real estate, life services, sports fitness, cultural media, finance, and the like. A day is divided into 7 time periods: 0 to 7 hours, 7 to 10 hours, 10 to 13 hours, 13 to 16 hours, 16 to 19 hours, 19 to 22 hours, and 22 to 24 hours.
The formula for calculating the distance between the two access points A, B is as follows:
Distance＝R×arccosθ
where θ is cos LatA × cosLatB × cos (LonA-LonB) + sinLatA × sinLatB. Lat represents the latitude of the coordinate, Lon represents the longitude of the coordinate, and R represents the radius of the earth.
Subsequently constructing the third order tensor
Figure BDA0001549956150000061
The three dimensions and the corresponding numbers are respectively: n users, T time slots and C interest point categories, where N is 50, T is 7 and 180, and C is 28 in this embodiment. Tensor X is expanded into C matrices according to model-1 (user dimension), of the form:
Figure BDA0001549956150000062
wherein the content of the first and second substances,tensor size of user-time-interest point kind is N x T C, xijkRepresenting the probability of a user i appearing on the point of interest category k for a time period j, e.g. x, with a value between 0 and 1111The value of (d) represents the magnitude of the probability that user 1 accesses the point of interest category 1 during time period 1. The probability of initialization in this embodiment is
Figure BDA0001549956150000071
P represents the number of POI categories corresponding to the time period j of the user i, and the tensor is expressed in a sparse form (the missing values in the sparse tensor are all 0) as shown in table 2.
TABLE 2
User' s Time period Species of Probability of
1 1 1 0.5
1 1 3 0.5
2 6 11 0.25
2 6 7 0.25
2 6 8 0.25
2 6 13 0.25
S103: extracting auxiliary information:
clustering the access points of all users, defining time periods of possible access to residential places and working places, and respectively carrying out DBScan density clustering on the access points of each user. Finding a direct density reachable point q in the neighborhood of any core point P, and calculating the density based on the following formula:
ρ(x)＝|Nε(x)|
wherein N isε(x) X ∈ X d (y, X ≦ epsilon } represents the set of poi in the X neighborhood.
The locations of residences and places of work after some users are clustered in this embodiment are shown in table 3.
TABLE 3
User ID Latitude and longitude of residence Working ground latitude
LGXC76C3******56 121.345508,31.257684 121.397267,31.277853
LGXC76C3******91 121.539853,31.330799 121.665357,31.173331
LGXC76C3******25 121.524631,31.258117 121.524617,31.258115
S104: adding negative-unlabeled constraint:
predicting the interest point category of the user by utilizing a tensor decomposition model with a negative-unlabeled constraint, wherein the negative-unlabeled constraint is as follows:
Figure BDA0001549956150000081
element χ in tensorijk∈[0,1]Represents the probability, Ω, that user i appears on the point of interest category k in time period jijAn uncertainty region is defined for user i at the access point of time j. If the interest point category appears in the uncertainty region of the corresponding user i in the time period j, the value is between 0 and 1, and the sum of the probabilities of all the categories in the uncertainty region is 1. If the interest point category appears outside the corresponding uncertain region, the probability of the user accessing the interest point category is 0. In this example
Figure BDA0001549956150000082
For the POI category probability vector of the user i in the uncertain region corresponding to the time period j, the negative-unlabeled constraint aims to obtain(Vector)
Figure BDA0001549956150000083
Projection onto a probability simplex, i.e. satisfying the formula:
Figure BDA0001549956150000084
s105: constructing an assistant feature regularization term:
fusing the similarity of the user auxiliary information and the types of interest points between the departure place and the destination as regularization terms of tensor decomposition, and calculating the tensor
Figure BDA0001549956150000085
The calculation formula is as follows:
Figure BDA0001549956150000086
in the formula (I), the compound is shown in the specification,
Figure BDA0001549956150000087
is F two norm, the square root U representing the sum of the squares of all elements(n)Is a hidden vector matrix of tensor x at the nth latitude, S is an auxiliary information similarity matrix,
Figure BDA00015499561500000810
is Laplacian matrix, beta is weight value of auxiliary information matrix S, and lambda is weight value of user hidden vector matrix.
Calculating the similarity of the interest point categories between the departure place and the destination by adopting the following formula:
Figure BDA0001549956150000088
wherein, Depij、DesijRespectively representing a departure place and a destination corresponding to the time period j of the user i; c represents interest point kind set in access point uncertain area, n and U represent intersection and union of two sets, t represents specific time, alpha isA fission constant.
Calculating the similarity of the residence places of the users a and b by adopting the following formula:
Figure BDA0001549956150000089
wherein, dist (H)a-Hb) Representing the distance between the residence places of the user a and the user b; α is the fission constant.
Figure BDA0001549956150000091
Wherein, dist (W)a-Wb) Representing the distance between the user a and the user b; α is the fission constant.
In the present embodiment, distance calculation dist (W)a-Wb)、dist(Ha-Hb) Satisfies the following formula:
Distance＝R×arccosθ
where θ is cos LatA × cosLatB × cos (LonA-LonB) + sinLatA × sinLatB. Lat represents the latitude of the coordinate, Lon represents the longitude of the coordinate, and R represents the radius of the earth.
S106: and (3) generating a prediction result:
in this embodiment, an alternating least square method is used for fitting, and a model is trained to obtain a final prediction result. The tensor of prediction results in this embodiment is represented in sparse form as in table 4.
TABLE 4
User' s Time period Species of Probability of
1 1 1 0.0109
1 1 3 0.9891
2 6 11 0.5327
2 6 7 0.3018
2 6 8 0.0341
2 6 13 0.1314
Although illustrative embodiments of the present invention have been described above to facilitate the understanding of the present invention by those skilled in the art, it should be understood that the present invention is not limited to the scope of the embodiments, and various changes may be made apparent to those skilled in the art as long as they are within the spirit and scope of the present invention as defined and defined by the appended claims, and all matters of the invention which utilize the inventive concepts are protected.

Claims (1)

1. A point of interest category prediction method based on a tensor decomposition model is characterized by comprising the following steps:
s1: acquiring the coordinate information of an access point and the stay time of a user in T time periods every day from the new energy automobile track data;
s2: defining the range of a user staying area as an uncertain area according to the time period of the access point and the staying time length, acquiring the id of the interest point type in the area, and constructing a tensor
Figure FDA0003251360630000011
S3: clustering the track data corresponding to all users, wherein the track data comprises residential area clustering and working area clustering which are used as auxiliary information;
s4: predicting the interest point category of the user by utilizing a tensor decomposition model with a negative-unlabeled constraint;
s5: fusing auxiliary information of a place of work, a place of residence and a place of departure, using the similarity of the types of interest points between the place of departure and the destination as a conversion term constraint of tensor decomposition, and calculating tensor
Figure FDA0003251360630000012
A low rank approximate solution of;
s6: iterative optimization of a target function is carried out by using an alternating least square method, and the probability that a user accesses different interest point types is obtained and used as a prediction result; wherein:
the step S2 constructs the third order tensor as follows
Figure FDA0003251360630000013
The three dimensions of the tensor are user, time and interest point categories respectively, and the corresponding dimensions are N, T, C respectively; each element
Figure FDA0003251360630000014
The value of (d) represents the probability that user i visits the point of interest of category k during time period j;
said step S3 uses DBScan density clustering algorithm;
the negative-unlabeled constraint satisfied in step S4 is:
Figure FDA0003251360630000015
wherein the content of the first and second substances,
Figure FDA0003251360630000016
representing the probability, Ω, that user i visits a point of interest of category k during time period jijAn uncertain region is defined for an access point of the user i in the time period j;
the step S5 obtains a low rank approximate solution by the following tensor decomposition method:
Figure FDA0003251360630000017
in the formula (I), the compound is shown in the specification,
Figure FDA0003251360630000018
is F two norm, represents the square root of the sum of the squares of all elements,
Figure FDA0003251360630000019
for the estimated value obtained after iteration, U(n)Is tensor
Figure FDA00032513606300000110
A user implicit vector matrix at the nth latitude, S is a similarity matrix for describing auxiliary information,
Figure FDA00032513606300000111
is a Laplacian matrix and is a matrix of Laplacian,beta is the weight value of the auxiliary information matrix S, and lambda is the weight value of the user hidden vector matrix.
CN201810042721.7A 2018-01-17 2018-01-17 Interest point type prediction method based on tensor decomposition model Active CN108256914B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810042721.7A CN108256914B (en) 2018-01-17 2018-01-17 Interest point type prediction method based on tensor decomposition model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810042721.7A CN108256914B (en) 2018-01-17 2018-01-17 Interest point type prediction method based on tensor decomposition model

Publications (2)

Publication Number Publication Date
CN108256914A CN108256914A (en) 2018-07-06
CN108256914B true CN108256914B (en) 2022-02-11

Family

ID=62741361

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810042721.7A Active CN108256914B (en) 2018-01-17 2018-01-17 Interest point type prediction method based on tensor decomposition model

Country Status (1)

Country Link
CN (1) CN108256914B (en)

Families Citing this family (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110309434B (en) * 2018-10-10 2023-10-13 腾讯大地通途（北京）科技有限公司 Track data processing method and device and related equipment
CN110196891B (en) * 2018-11-15 2024-03-15 腾讯大地通途（北京）科技有限公司 Method and device for determining block type, storage medium and electronic device
CN109684604B (en) * 2018-12-06 2020-06-30 北京航空航天大学 City dynamic analysis method based on context-aware nonnegative tensor decomposition
CN109858951B (en) * 2018-12-29 2023-04-07 蔚来(安徽)控股有限公司 Prediction of power-on or power-switching requirements of new energy automobile
CN109918582B (en) * 2019-03-06 2023-06-20 上海评驾科技有限公司 Method for identifying single interest point of user based on space-time data
CN110210604B (en) * 2019-05-21 2021-06-04 北京邮电大学 Method and device for predicting movement track of terminal equipment
CN110223786B (en) * 2019-06-13 2021-08-13 重庆亿创西北工业技术研究院有限公司 Method and system for predicting drug-drug interaction based on nonnegative tensor decomposition
CN111931998B (en) * 2020-07-27 2023-05-02 大连海事大学 Individual travel mode prediction method and system based on mobile positioning data
CN113938817A (en) * 2021-09-10 2022-01-14 哈尔滨工业大学（威海） Vehicle owner travel position prediction method based on vehicle position information
CN113918837B (en) * 2021-10-15 2024-02-06 山东大学 Method and system for generating city interest point category representation

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105513370A (en) * 2015-12-29 2016-04-20 浙江大学 Traffic zone dividing method based on sparse vehicle license identification data
CN105513489A (en) * 2016-01-15 2016-04-20 上海交通大学 Method for building city noise map
CN106909560A (en) * 2015-12-23 2017-06-30 香港中文大学 Point of interest sort method
CN106960044A (en) * 2017-03-30 2017-07-18 浙江鸿程计算机系统有限公司 A kind of Time Perception personalization POI based on tensor resolution and Weighted H ITS recommends method

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101365993B1 (en) * 2011-12-29 2014-02-21 홍익대학교 산학협력단 Data processing method, data processing apparatus, data acquisition method, and information providing method

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106909560A (en) * 2015-12-23 2017-06-30 香港中文大学 Point of interest sort method
CN105513370A (en) * 2015-12-29 2016-04-20 浙江大学 Traffic zone dividing method based on sparse vehicle license identification data
CN105513489A (en) * 2016-01-15 2016-04-20 上海交通大学 Method for building city noise map
CN106960044A (en) * 2017-03-30 2017-07-18 浙江鸿程计算机系统有限公司 A kind of Time Perception personalization POI based on tensor resolution and Weighted H ITS recommends method

Also Published As

Publication number Publication date
CN108256914A (en) 2018-07-06

Similar Documents

Publication Publication Date Title
CN108256914B (en) Interest point type prediction method based on tensor decomposition model
JP6864329B2 (en) Methods and equipment for searching or comparing sites using routes or route distances between sites and locations within the transportation system.
Liu et al. Characterizing mixed-use buildings based on multi-source big data
Langford An evaluation of small area population estimation techniques using open access ancillary data
Bhat et al. Accommodating immigration status and self-selection effects in a joint model of household auto ownership and residential location choice
Zhang et al. City2vec: Urban knowledge discovery based on population mobile network
Seya et al. A comparison of residential apartment rent price predictions using a large data set: Kriging versus deep neural network
Gao et al. Identification of urban regions’ functions in Chengdu, China, based on vehicle trajectory data
Deng et al. Identify urban building functions with multisource data: A case study in Guangzhou, China
Xia et al. Decision Tree‐Based Contextual Location Prediction from Mobile Device Logs
Yang et al. Pedestrian network generation based on crowdsourced tracking data
Zhao et al. Mapping hourly population dynamics using remotely sensed and geospatial data: a case study in Beijing, China
Chen et al. An analysis of movement patterns between zones using taxi GPS data
Yuan et al. Recognition of functional areas based on call detail records and point of interest data
Chen et al. UVLens: Urban village boundary identification and population estimation leveraging open government data
Zhao et al. How does the urban built environment affect online car-hailing ridership intensity among different scales?
Wang et al. A PSO-neural network-based feature matching approach in data integration
Liao et al. Fusing geographic information into latent factor model for pick-up region recommendation
JP7144818B2 (en) Method and Apparatus for Searching or Comparing Sites Using Routes or Route Distances Between Sites and Locations in a Transportation System
Yang et al. Fusing mobile phone and travel survey data to model urban activity dynamics
Rodríguez-Echeverría et al. Density-based spatial clustering and ordering points approach for characterizations of tourist behaviour
US10791418B2 (en) Method and system for modeling user and location
CN112364776A (en) Urban functional area recognition model establishing and classifying method and device
Hummler et al. Web mining to inform locations of charging stations for electric vehicles
Liu et al. Comparing spatial interaction models and flow interpolation techniques for predicting “cold start” bike‐share trip demand

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