CN111988744B - Position prediction method based on user moving mode - Google Patents

Position prediction method based on user moving mode Download PDF

Info

Publication number
CN111988744B
CN111988744B CN202010898332.1A CN202010898332A CN111988744B CN 111988744 B CN111988744 B CN 111988744B CN 202010898332 A CN202010898332 A CN 202010898332A CN 111988744 B CN111988744 B CN 111988744B
Authority
CN
China
Prior art keywords
user
weather
movement
pattern
mode
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
CN202010898332.1A
Other languages
Chinese (zh)
Other versions
CN111988744A (en
Inventor
苏畅
严杨志
谢显中
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.)
Chongqing University of Post and Telecommunications
Original Assignee
Chongqing University of Post and Telecommunications
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 Chongqing University of Post and Telecommunications filed Critical Chongqing University of Post and Telecommunications
Priority to CN202010898332.1A priority Critical patent/CN111988744B/en
Publication of CN111988744A publication Critical patent/CN111988744A/en
Application granted granted Critical
Publication of CN111988744B publication Critical patent/CN111988744B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
    • H04W4/02—Services making use of location information
    • H04W4/029—Location-based management or tracking services
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/953—Querying, e.g. by the use of web search engines
    • G06F16/9537—Spatial or temporal dependent retrieval, e.g. spatiotemporal queries
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • 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"
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
    • H04W4/02—Services making use of location information
    • H04W4/021—Services related to particular areas, e.g. point of interest [POI] services, venue services or geofences

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • Databases & Information Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Human Resources & Organizations (AREA)
  • Evolutionary Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • Signal Processing (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Strategic Management (AREA)
  • Economics (AREA)
  • Evolutionary Computation (AREA)
  • Game Theory and Decision Science (AREA)
  • Development Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Marketing (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention relates to a position prediction method based on a user movement mode, and belongs to the field of machine learning. The method comprises the following steps: adopting an Apriori algorithm to dig out an individual moving mode of each user and finding out internal factors influencing the sign-in of the user; calculating the similarity between the individual movement modes of the user by using a dynamic time warping algorithm DTW; grouping individual moving modes of users through clustering to obtain a central mode of each group, and finding out external factors influencing check-in; training a Markov model with the individual movement patterns and the overall movement patterns, respectively; training a Markov chain model based on IMP and AMP to predict a next location of the user; considering the influence of external weather, creating a weather general characteristic; calculating the similarity between the weather of the current place and the weather of other places by using a Gaussian kernel function, and correcting a predicted result; setting an evaluation standard and a reference method. The invention enables the predicted result to be more suitable for actual life.

Description

Position prediction method based on user moving mode
Technical Field
The invention belongs to the field of machine learning, and relates to a position prediction method based on a user movement mode.
Background
With the popularization of mobile terminals, human mobile data are more easily obtained, a social network platform based on the position collects a large amount of user check-in data, the research on the human mobile rule becomes a hotspot, the research on the mobile mode of people also becomes possible, and the position prediction is more common. Through position prediction, the mobile preference of the user can be known in advance, the mobile tendency of people flow can also be known, and not only can targeted service be provided for the user, but also benefits are brought to merchants. The existing research mainly analyzes the behavior of a user through the sign-in historical record of the user, finds the movement rule of the user and then predicts the place. Most of the considered factors have time, space, social contact and the like, and mainly aim at the preference of the user and ignore the connection between the positions. In addition, most studies are predicated on the individual movement patterns of the user, and if the user goes to a place that has never been visited, no data is available to train the model; however, the prediction based on the overall data is too coarse-grained, and if the users are currently located at the same place, the final prediction results are all the same place and are not in accordance with the actual situation.
Aiming at the problem that the traditional position prediction model based on the discrete state sequence cannot well predict the position, the invention respectively excavates the individual movement mode and the whole movement mode of the user from the historical sign-in data of the user by considering the association between different positions in the sign-in track of the user, namely, the internal cause and the external cause which influence the sign-in of the user. Wherein, the overall movement mode is mainly embodied by the group behaviors; the individual movement pattern takes into account the dynamic changes of the movement pattern within the individual user, taking into account the personalized location prediction. The time factors are also taken into consideration in the process of mining the movement patterns of the users, and the movement patterns of the users in different time periods in the week are mined aiming at the changes of the user movement patterns in two different time periods, namely a working day and a weekend, so that the actual transfer condition between the spatial positions of the users can be reflected, and the potential time law of the movement of the users is also included, and the predicted result is more suitable for the actual life. In addition, a weather general characteristic is created, and the result is corrected by calculating weather similarity by adopting a Gaussian kernel function.
Disclosure of Invention
In view of the above, the present invention provides a method for predicting a location based on a user movement pattern.
In order to achieve the purpose, the invention provides the following technical scheme:
a method for location prediction based on user movement patterns, the method comprising the steps of:
define the moving pattern MP as: a set of locations sequentially visited by a user in a continuous time; a pattern in which the user frequently moves is referred to as a user's movement pattern, and the movement pattern is denoted as MP ═ l1,l2,l3,…,lnN is the number of positions contained in the mobile mode of the user;
define the individual mobility pattern IMP as: a sequence of locations frequently occurring in a user's personal historical visit location, for a given user, whose movement pattern is a set of all movement patterns in the historical check-in record;
defining the support degree as: the frequency with which the user's movement pattern appears in its movement trajectory; in the historical check-in record of the user, including multiple movement tracks, the support degree of the movement mode may be calculated as:
Figure BDA0002659076960000021
define the global mobile mode AMP as: frequently occurring movement patterns in historical check-ins for all users in different groups; calculating the similarity of the user moving modes by adopting a DTW algorithm, then dividing the similarity into a plurality of groups through clustering, and finding a central mode according to the moving modes in the historical access positions of all users in each group to obtain the integral moving mode of all users in the group;
defining the general weather characteristics: combining the rainfall, temperature and wind speed into a new characteristic according to a weighting fusion mode;
s1: adopting an Apriori algorithm to dig out an individual moving mode of each user and finding out internal factors influencing the sign-in of the user;
s2: calculating the similarity between the individual movement modes of the user by using a dynamic time warping algorithm DTW;
s3: grouping individual moving modes of users through clustering to obtain a central mode of each group, namely an integral moving mode AMP, and finding out external factors influencing check-in;
s4: training a Markov model with the individual movement patterns and the overall movement patterns, respectively;
s5: training a Markov chain model based on IMP and AMP, combining probability vectors of the IMP and AMP, and predicting the next position of the user;
s6: considering the influence of external weather, creating a weather general characteristic;
s7: calculating the similarity between the weather of the current place and the weather of other places by using a Gaussian kernel function, and correcting a predicted result;
s8: setting an evaluation standard and a reference method.
Optionally, step S1 specifically includes:
s11: in a given time range, through analyzing Gowalla, finding out a moving mode with the length of 1 in the check-in record of the user;
s12: then finding out the mobile mode with the length of 2 in sequence, then calculating whether the support degree sigma meets the requirement, and circulating until the length of the mobile mode cannot be increased to obtain a candidate mobile mode;
s13: and finding out the movement mode with the support degree meeting the condition from the obtained candidate movement modes to obtain the individual movement mode of the user.
Optionally, in the step, the euclidean distance between the two points is not simply calculated for the similarity between the two movement modes, but the Haversine distance is calculated, and the coordinates of the two points are introduced to obtain the geospatial distance between the actual two points, which is specifically as follows:
Figure BDA0002659076960000031
wherein:
Figure BDA0002659076960000032
|Mpl represents the length of the movement pattern, i.e. the number of positions in the pattern; rest (M)p) Indicating a movement pattern to remove the first position, d (l, l)i) Representing the true distance between two locations.
Optionally, step S3 specifically includes:
s31: initializing a plurality of classes according to the personal movement mode of a user, and setting a distance threshold tau;
s32: for each movement pattern of each user, calculating the distance between each movement pattern and each class, and selecting the class with the minimum distance;
s33: then, calculating the distance between the moving mode and the class by adopting a DTW algorithm, and if the distance is smaller than a threshold value tau, adding the distance into the class and updating; otherwise, a new class is created for the mobile mode;
s34: and obtaining a clustering result, namely the overall movement pattern of each person.
Optionally, step S4 specifically includes:
s41: after clustering the individual movement patterns of the user, obtaining the overall movement pattern of the user, and combining the obtained overall movement pattern, wherein the next position to be reached is as follows:
Figure BDA0002659076960000033
s42: based on the personal movement pattern, the next location to go is:
Figure BDA0002659076960000034
wherein the content of the first and second substances,
Figure BDA0002659076960000041
is represented by containingMoving pattern of N positions, MPcA set of classes of movement patterns is represented,
Figure BDA0002659076960000042
the expression sequence is
Figure BDA0002659076960000043
The moving mode of (A) appears in the MPcThe number of times of (1) to (d),
Figure BDA0002659076960000044
is shown in MPcIn position liThe number of subsequent occurrences.
Optionally, in step S5, for each person, there is a personal movement pattern and a global movement pattern, which are used for training the markov model respectively; finally, a predicted probability vector is obtained; the vector based on the whole is PAMP＝(l1,l2,l3,…,ln) Vector based on individual movement pattern is PIMP＝(l1,l2,l3,…,ln) Wherein n represents the number of predicted positions; then combining the two obtained results to obtain a final prediction result; the final combination is as follows:
P＝α·PIMP+(1-α)PAMP。
optionally, the step S6 includes:
s61: creating a weather Total feature Xweather＝[Temperature,Rain,Windspeed]；
S62: the three kinds of weather of the user check-in place are subjected to weighted summation, the influence of the three kinds of weather on the user check-in place is comprehensively considered, and the total weather characteristic of each user check-in place is obtained, and the total weather characteristic is specifically represented as follows:
Xweather＝ω1Temperature+ω2Windspeed+ω3Rain
wherein, the weight of rainfall is calculated as follows:
Figure BDA0002659076960000045
Figure BDA0002659076960000046
refers to one of the places l where the user checks iniThe total number of user check-ins in a given rainfall interval,
Figure BDA0002659076960000047
the total number of days of the rainfall interval in the corresponding time period; the weighting calculations for wind speed and temperature are also consistent.
Optionally, in the step S7, after calculating the weather preference of the user through the created weather total feature, calculating the current location X of the user by using a gaussian kernel functionlTo other locations
Figure BDA0002659076960000049
The similarity of weather is obtained to obtain a final prediction result; the specific calculation is as follows:
Figure BDA0002659076960000048
wherein, XlIndicating the weather conditions of the location where the user is currently located,
Figure BDA00026590769600000410
the weather conditions of other locations.
Optionally, the step S8 includes:
s81: taking the Accuracy and the APR of the site prediction as evaluation standards of the experiment;
s82: accuracy: the index defines the proportion of the correct prediction place to the total prediction place in the prediction result list of the user; when the prediction result is consistent with the actual result, p (l) is 1;
Figure BDA0002659076960000051
s83: average percent ranking APR: the prediction problem also has a certain relation with the sequence, user uiSign-in location ljIn the prediction list PR is defined as:
Figure BDA0002659076960000052
obtaining APR values of all users by taking the average value of the sum of the PR values, wherein the larger the value is, the better the prediction effect is; the formula is as follows:
Figure BDA0002659076960000053
s84: in order to verify the effectiveness of the proposed location prediction method based on the user movement pattern, the following models are selected for comparison with the proposed model:
NextPlace: the method is a classical position prediction method, and is used for predicting user behaviors based on nonlinear time series analysis of arrival time and predicting by using similarity of time series;
SimPreT: associating the historical pattern with the current user trajectory, and determining the next position of the user by using pattern similarity;
HMM-based: the model simultaneously considers non-Gaussian and spatiotemporal characteristics in actual human check-in data by constructing a hybrid Markov model.
The invention has the beneficial effects that: according to the method, the relevance among different positions in the user sign-in track is considered, and the individual movement mode and the overall movement mode of the user are respectively mined from the historical sign-in data of the user, namely, the internal cause and the external cause of the user sign-in are influenced. Wherein, the overall movement mode is mainly embodied by the group behaviors; the individual movement pattern takes into account the dynamic changes of the movement pattern within the individual user, taking into account the personalized location prediction. The time factors are also taken into consideration in the process of mining the movement patterns of the users, and the movement patterns of the users in different time periods in the week are mined aiming at the changes of the user movement patterns in two different time periods, namely a working day and a weekend, so that the actual transfer condition between the spatial positions of the users can be reflected, and the potential time law of the movement of the users is also included, and the predicted result is more suitable for the actual life. In addition, a weather general characteristic is created, and the result is corrected by calculating weather similarity by adopting a Gaussian kernel function.
Additional advantages, objects, and features of the invention will be set forth in part in the description which follows and in part will become apparent to those having ordinary skill in the art upon examination of the following or may be learned from practice of the invention. The objectives and other advantages of the invention may be realized and attained by the means of the instrumentalities and combinations particularly pointed out hereinafter.
Drawings
For the purposes of promoting a better understanding of the objects, aspects and advantages of the invention, reference will now be made to the following detailed description taken in conjunction with the accompanying drawings in which:
FIG. 1 is a general flow diagram of the present invention;
FIG. 2 is check-in location information in a data set;
FIG. 3 is a comparison of prediction accuracy for different movement patterns;
FIG. 4 is a comparison of the accuracy of the model in two cities;
FIG. 5 is a comparison of model APR values.
Detailed Description
The embodiments of the present invention are described below with reference to specific embodiments, and other advantages and effects of the present invention will be easily understood by those skilled in the art from the disclosure of the present specification. The invention is capable of other and different embodiments and of being practiced or of being carried out in various ways, and its several details are capable of modification in various respects, all without departing from the spirit and scope of the present invention. It should be noted that the drawings provided in the following embodiments are only for illustrating the basic idea of the present invention in a schematic way, and the features in the following embodiments and examples may be combined with each other without conflict.
Wherein the showings are for the purpose of illustrating the invention only and not for the purpose of limiting the same, and in which there is shown by way of illustration only and not in the drawings in which there is no intention to limit the invention thereto; to better illustrate the embodiments of the present invention, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of an actual product; it will be understood by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted.
The same or similar reference numerals in the drawings of the embodiments of the present invention correspond to the same or similar components; in the description of the present invention, it should be understood that if there is an orientation or positional relationship indicated by terms such as "upper", "lower", "left", "right", "front", "rear", etc., based on the orientation or positional relationship shown in the drawings, it is only for convenience of description and simplification of description, but it is not an indication or suggestion that the referred device or element must have a specific orientation, be constructed in a specific orientation, and be operated, and therefore, the terms describing the positional relationship in the drawings are only used for illustrative purposes, and are not to be construed as limiting the present invention, and the specific meaning of the terms may be understood by those skilled in the art according to specific situations.
The invention relates to a position prediction method based on a user moving mode. Firstly, excavating a moving mode of each user; then clustering is carried out according to the similarity by utilizing a DTW algorithm to obtain an integral moving mode; next, a markov model is trained using the individual movement pattern and the entire movement pattern, respectively, to predict the position. Finally, the total weather characteristics are created, and the result is corrected by calculating weather similarity through a Gaussian kernel function.
In order to enable a more concise and clear description, some noun definitions are explained:
define the Movement Pattern (MP) as: a set of locations that a user visits in a sequential temporal order. A pattern in which the user frequently moves is referred to as a movement pattern of the user, and the movement pattern may be expressed as MP ═ l1,l2,l3,…,lnN is the movement of the userNumber of positions contained in the pattern.
Define an Individual Movement Pattern (IMP) as: the user's individual historical visit locations frequently appear in a sequence of locations whose movement patterns are a set of all movement patterns in the historical check-in record for a given user.
Defining the support degree as: the frequency with which the user's movement pattern appears in its movement trajectory. In the historical check-in record of the user, including multiple movement tracks, the support degree of the movement mode may be calculated as:
Figure BDA0002659076960000071
define the global movement pattern (AMP) as: all users of different groups have a pattern of movements that frequently occurs in historical check-ins. And calculating the similarity of the movement patterns of the users by adopting a DTW algorithm, then dividing the similarity into a plurality of groups through clustering, and finding a central pattern according to the movement patterns in the historical access positions of all the users in each group to obtain the integral movement pattern of all the users in the group.
Defining the general weather characteristics: the rainfall, temperature and wind speed weather features are combined into a new feature according to a weighting fusion mode.
As shown in fig. 1, the present invention is divided into the following steps:
s1: and excavating the individual movement mode of each user by adopting an Apriori algorithm, and finding out internal factors influencing the sign-in of the user.
S2: the similarity between individual movement patterns of the user is calculated using a Dynamic Time Warping (DTW) algorithm.
S3: and grouping the individual movement patterns of the users through clustering to obtain a central pattern of each group, namely an overall movement pattern (AMP), and finding out external factors influencing check-in.
S4: the markov model is trained with individual and global movement patterns, respectively.
S5: the Markov chain model is trained based on IMP and AMP, and the probability vectors of the two are combined to predict the next position of the user.
S6: the total weather characteristics are created taking into account the influence of the extrinsic weather.
S7: and calculating the similarity between the weather of the current place and the weather of other places by using the Gaussian kernel function, and correcting the predicted result.
S8: setting an evaluation standard and a reference method.
In step S1, an Apriori algorithm is used to mine the movement pattern of the user, where the Apriori algorithm is a data mining algorithm based on association rules and is subsequently applied to mine the movement pattern of the user. A time factor is added in the excavation process of the mobile mode, so that the excavated user mobile mode has a time rule, and the change situation of the user mobile mode along with the time can be known. First, in a given time range (weekday, weekend), Gowalla is analyzed to find out a moving pattern with length of 1 in the check-in record of the user, and the check-in information is shown in FIG. 2. And then finding out the moving mode with the length of 2 in sequence, calculating whether the support degree sigma meets the requirement, and circulating until the length of the moving mode cannot be increased, thus obtaining the candidate moving mode. And finding out the movement mode with the support degree meeting the condition from the obtained candidate movement modes to obtain the individual movement mode of the user.
In step S2, the similarity between two movement patterns is calculated by not simply calculating the euclidean distance between two points, but calculating the Haversine distance, and the geographical distance between two points can be obtained by introducing the coordinates of the two points, so that the calculated distance is more accurate. The method comprises the following specific steps:
Figure BDA0002659076960000081
wherein:
Figure BDA0002659076960000082
|Mpl represents the length of the movement pattern, i.e. the number of positions in the pattern; rest (M)p) Indicating a movement pattern to remove the first position, d (l, l)i) Representing the true distance between two locations.
In step S3, the user' S individual movement pattern is obtained in step S1, and a plurality of classes are initialized and the distance threshold τ is set. For each movement pattern of each user, its distance from each class is calculated and the class with the smallest distance is selected. Then, the distance between the moving pattern and the class is calculated by adopting a DTW algorithm, if the distance is smaller than a threshold value tau, the distance is added into the class and updated, and otherwise, a new class is created for the moving pattern. And finally, obtaining a clustering result, namely the overall movement mode of each person.
In step S4, after clustering the personal movement patterns of the user according to step S3, we obtain the movement patterns of the whole user, and the next position to be reached by combining the obtained whole movement patterns is:
Figure BDA0002659076960000083
based on the personal mobile mode, the next location to go is:
Figure BDA0002659076960000091
wherein the content of the first and second substances,
Figure BDA0002659076960000092
representing a moving pattern, MP, having N positionscA set of classes of movement patterns is represented,
Figure BDA0002659076960000093
the expression sequence is
Figure BDA0002659076960000094
The moving mode of (A) appears in the MPcThe number of times of (1) to (d),
Figure BDA0002659076960000095
is shown in MPcIn position liThe number of subsequent occurrences.
In step S5, for each individual, there is an individual movement pattern and a global movement pattern, and each of these movement patterns is used for training a markov model. A predicted probability vector is finally obtained. The vector based on the whole is PAMP＝(l1,l2,l3,…,ln) Vector based on individual movement pattern is PIMP＝(l1,l2,l3,…,ln) Wherein n represents the number of predicted positions; the two results are then combined to obtain the final predicted result. The final combination is therefore as follows:
P＝α·PIMP+(1-α)PAMP
in step S6, the weather total feature X is first createdweather＝[Temperature,Rain,Windspeed]. Then, carrying out weighted summation on the three kinds of weather of the user check-in place, and comprehensively considering the influence of the three kinds of weather on the user check-in place to obtain the total weather characteristics of each user check-in place, wherein the total weather characteristics are specifically represented as follows:
Xweather＝ω1Temperature+ω2Windspeed+ω3Rain
wherein, the weight of rainfall is calculated as follows:
Figure BDA0002659076960000096
Figure BDA0002659076960000097
refers to one of the places l where the user checks iniThe total number of user check-ins in a given rainfall interval,
Figure BDA0002659076960000098
correspond toThe total number of days that the interval of rainfall occurred within the time period. The weights for wind speed and temperature are similarly calculated.
Wherein, in step S7, after calculating the weather preference of the user according to S6 by the created weather general characteristics, we adopt Gaussian kernel function to calculate the current location of the user (X)l) To other locations
Figure BDA00026590769600000911
And (5) obtaining a final prediction result according to the similarity of the weather. The specific calculation is as follows:
Figure BDA0002659076960000099
wherein, XlIndicating the weather conditions of the location where the user is currently located,
Figure BDA00026590769600000910
the weather conditions of other locations.
In step S8, check-in records of two cities of the Gowalla dataset in one year, which are data of London (LON) and Los Angeles (LA), respectively, are selected as the dataset for the test. Dividing experimental data into a test set, a training set and a verification set, training the training set, then verifying the training set on the verification set, and finally testing the test set. The information of the data set is shown in fig. 2:
accuracy and APR for site prediction were used as evaluation criteria for the experiment. The definition is as follows:
accuracy: the index defines the proportion of the correctly predicted location to the total predicted location in the prediction result list of the user. When the prediction result coincides with reality, p (l) becomes 1.
Figure BDA0002659076960000101
Average Percent Ranking (APR): the prediction problem also has a certain relation with the sequence, user uiSign-in location ljIn advance ofThe measurement list PR is defined as:
Figure BDA0002659076960000102
and taking the average value of the sum of the PR values to obtain the APR values of all users, wherein the larger the value is, the better the prediction effect is. The formula is as follows:
Figure BDA0002659076960000103
secondly, in order to verify the effectiveness of the proposed location prediction method based on the user movement pattern, the following models are selected for comparison with the proposed model:
NextPlace: the method is a classical position prediction method, and predicts the user behavior based on the nonlinear time series analysis of the arrival time, and predicts by using the similarity of the time series.
SimPreT: historical patterns are associated with the current user trajectory, and pattern similarity is used to determine the next location of the user.
HMM-based (Hybrid Markov Model-based): the model simultaneously considers non-Gaussian and spatiotemporal characteristics in actual human check-in data by constructing a hybrid Markov model.
Fig. 3 shows the results of the comparison of the predicted performance of the Individual Movement Pattern (IMP) and the overall movement pattern (AMP). From the figure we can see that the predicted performance gradually decreases with increasing position when only the overall movement pattern is considered; the method based on the individual activity mode obviously encounters the cold start problem in the early stage, the prediction accuracy is low, but the prediction accuracy is greatly improved along with the accumulation of the individual historical activity information. Especially in the later stages, it performs better than methods based on individual movement patterns and overall activity patterns. Despite this, we still consider both individual and overall movement patterns in our approach. It is believed that preserving the user's overall pattern makes our prediction method more robust and can handle some cases that cannot be predicted by relying only on personal movement patterns.
The accuracy ratio of the proposed model to other models is shown in fig. 4. It can be seen from the figure that the accuracy of the model proposed by the invention is much higher than that of NextPlace and SimPreT algorithm, and is slightly higher than that of HMM-based algorithm. The data in LA data set are respectively increased by 15%, 5% and 2.1%; the improvement in the LON data set was 14%, 6.5%, 4.2%, respectively. The illustration of considering both the user's individual and overall movement patterns, as well as weather factors, helps to improve the predicted outcome.
As shown in FIG. 5, in LA, the performance of the APR of the proposed model is better than that of other models, the APR of the model is improved by 19% compared with NextPlace, 7% compared with SimPret model and 4% compared with HMM-based model, and meanwhile, on LON data set, the model of the invention is improved by 18% compared with NextPlace, 9.2% compared with SimPret model and 5% compared with HMM-based model.
Finally, the above embodiments are only intended to illustrate the technical solutions of the present invention and not to limit the present invention, and although the present invention has been described in detail with reference to the preferred embodiments, it will be understood by those skilled in the art that modifications or equivalent substitutions may be made on the technical solutions of the present invention without departing from the spirit and scope of the technical solutions, and all of them should be covered by the claims of the present invention.

Claims (7)

1. A position prediction method based on a user moving mode is characterized in that: the method comprises the following steps:
define the moving pattern MP as: a set of locations sequentially visited by a user in a continuous time; a pattern in which the user frequently moves is referred to as a user's movement pattern, and the movement pattern is denoted as MP ═ l1,l2,l3,…,lnN is the number of positions contained in the mobile mode of the user;
define the individual mobility pattern IMP as: a sequence of locations frequently occurring in a user's personal historical visit location, for a given user, whose movement pattern is a set of all movement patterns in the historical check-in record;
defining the support degree as: the frequency with which the user's movement pattern appears in its movement trajectory; in the historical check-in record of the user, the history check-in record includes a plurality of movement tracks, and then the support degree of the movement mode is calculated as follows:
Figure FDA0003523736780000011
define the global mobile mode AMP as: frequently occurring movement patterns in historical check-ins for all users in different groups; calculating the similarity of the user moving modes by adopting a DTW algorithm, then dividing the user moving modes into a plurality of groups through clustering, and finding a central mode according to the moving modes in the historical access positions of all users in each group to obtain the integral moving mode of all users in the group;
defining the general weather characteristics: combining the rainfall, temperature and wind speed into a new characteristic according to a weighting fusion mode;
s1: adopting an Apriori algorithm to dig out an individual moving mode of each user and finding out internal factors influencing the sign-in of the user;
s2: calculating the similarity between the individual movement modes of the user by using a dynamic time warping algorithm DTW;
s3: grouping individual moving modes of users through clustering to obtain a central mode of each group, namely an integral moving mode AMP, and finding out external factors influencing check-in;
s4: training a Markov model with the individual movement patterns and the overall movement patterns, respectively;
s5: training a Markov chain model based on IMP and AMP, combining probability vectors of the IMP and AMP, and predicting the next position of the user;
s6: considering the influence of external weather, creating a weather general characteristic;
s7: calculating the similarity between the weather of the current place and the weather of other places by using a Gaussian kernel function, and correcting a predicted result;
s8: setting evaluation Standard and reference method'
The step S1 specifically includes:
s11: in a given time range, through analyzing Gowalla, finding out a moving mode with the length of 1 in the check-in record of the user;
s12: then finding out the mobile mode with the length of 2 in sequence, then calculating whether the support degree sigma meets the requirement, and circulating until the length of the mobile mode cannot be increased to obtain a candidate mobile mode;
s13: finding out a mobile mode with the support degree meeting the condition from the obtained candidate mobile modes to obtain an individual mobile mode of the user;
in the step, the Euclidean distance between two points is not simply calculated for the similarity of two moving modes, but the Haversine distance is calculated, and the coordinates of the two points are transmitted to obtain the geographic space distance between the two actual points, which is specifically as follows:
Figure FDA0003523736780000021
wherein:
Figure FDA0003523736780000022
|Mpl represents the length of the movement pattern, i.e. the number of positions in the pattern; rest (M)p) Indicating a movement pattern to remove the first position, d (l, l)i) Representing the true distance between two locations.
2. The method of claim 1, wherein the location prediction based on the user's moving pattern is: the step S3 specifically includes:
s31: initializing a plurality of classes according to the personal movement mode of a user, and setting a distance threshold tau;
s32: for each movement pattern of each user, calculating the distance between each movement pattern and each class, and selecting the class with the minimum distance;
s33: then, calculating the distance between the moving mode and the class by adopting a DTW algorithm, and if the distance is smaller than a threshold value tau, adding the distance into the class and updating; otherwise, a new class is created for the mobile mode;
s34: and obtaining a clustering result, namely the overall movement pattern of each person.
3. The method of claim 2, wherein the location prediction based on the user's moving pattern is: the step S4 specifically includes:
s41: after clustering the individual movement patterns of the user, obtaining the overall movement pattern of the user, and combining the obtained overall movement pattern, wherein the next position to be reached is as follows:
Figure FDA0003523736780000023
s42: based on the personal movement pattern, the next location to go is:
Figure FDA0003523736780000031
wherein the content of the first and second substances,
Figure FDA0003523736780000032
representing a moving pattern, MP, having N positionscA set of classes of movement patterns is represented,
Figure FDA0003523736780000033
the expression sequence is
Figure FDA0003523736780000034
The moving mode of (A) appears in the MPcThe number of times of (1) to (d),
Figure FDA0003523736780000035
is shown in MPcIn position liThe number of subsequent occurrences.
4. A method according to claim 3, wherein the method comprises: the step S5 includes, for each person, a personal movement pattern and a whole movement pattern, and each of the movement patterns is used for training a markov model; finally, a predicted probability vector is obtained; the vector based on the whole is PAMP＝(l1,l2,l3,…,ln) Vector based on individual movement pattern is PIMP＝(l1,l2,l3,…,ln) Wherein n represents the number of predicted positions; then combining the two obtained results to obtain a final prediction result; the final combination is as follows:
P＝α·PIMP+(1-α)PAMP。
5. the method of claim 4, wherein the location prediction based on the user's moving pattern is: the step S6 includes:
s61: creating a weather Total feature Xweather＝[Temperature,Rain,Windspeed]；
S62: the three kinds of weather of the user check-in place are subjected to weighted summation, the influence of the three kinds of weather on the user check-in place is comprehensively considered, and the total weather characteristic of each user check-in place is obtained, and the total weather characteristic is specifically represented as follows:
Xweather＝ω1Rain+ω2Windspeed+ω3Temperature
wherein, the weight of rainfall is calculated as follows:
Figure FDA0003523736780000036
Figure FDA0003523736780000037
refers to one of the places l where the user checks iniThe total number of user check-ins in a given rainfall interval,
Figure FDA0003523736780000038
the total number of days of the rainfall interval in the corresponding time period; the weighting calculations for wind speed and temperature are also consistent.
6. The method of claim 5, wherein the location prediction based on the user's moving pattern is: in the step S7, after calculating the weather preference of the user through the created weather total characteristics, the current location X of the user is calculated by using the gaussian kernel functionlTo other locations
Figure FDA0003523736780000039
The similarity of weather is obtained to obtain a final prediction result; the specific calculation is as follows:
Figure FDA0003523736780000041
wherein, XlIndicating the weather conditions of the location where the user is currently located,
Figure FDA0003523736780000042
the weather conditions of other locations.
7. The method of claim 6, wherein the location prediction based on the user's moving pattern is: the step S8 includes:
s81: taking the Accuracy and the APR of the site prediction as evaluation standards of the experiment;
s82: accuracy: the index defines the proportion of the correct prediction place to the total prediction place in the prediction result list of the user; when the prediction result is consistent with the actual result, p (l) is 1;
Figure FDA0003523736780000043
s83: average hundredRanking by ratio APR: the prediction problem also has a certain relation with the sequence, user uiSign-in location ljIn the prediction list PR is defined as:
Figure FDA0003523736780000044
obtaining APR values of all users by taking the average value of the sum of the PR values, wherein the larger the value is, the better the prediction effect is; the formula is as follows:
Figure FDA0003523736780000045
s84: in order to verify the effectiveness of the proposed location prediction method based on the user movement pattern, the following models are selected for comparison with the proposed model:
NextPlace: the method is a classical position prediction method, and is used for predicting user behaviors based on nonlinear time series analysis of arrival time and predicting by using similarity of time series;
SimPreT: associating the historical pattern with the current user trajectory, and determining the next position of the user by using pattern similarity;
HMM-based: the model simultaneously considers non-Gaussian and spatiotemporal characteristics in actual human check-in data by constructing a hybrid Markov model.
CN202010898332.1A 2020-08-31 2020-08-31 Position prediction method based on user moving mode Active CN111988744B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010898332.1A CN111988744B (en) 2020-08-31 2020-08-31 Position prediction method based on user moving mode

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010898332.1A CN111988744B (en) 2020-08-31 2020-08-31 Position prediction method based on user moving mode

Publications (2)

Publication Number Publication Date
CN111988744A CN111988744A (en) 2020-11-24
CN111988744B true CN111988744B (en) 2022-04-01

Family

ID=73441209

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010898332.1A Active CN111988744B (en) 2020-08-31 2020-08-31 Position prediction method based on user moving mode

Country Status (1)

Country Link
CN (1) CN111988744B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112541621B (en) * 2020-12-02 2021-08-31 香港理工大学深圳研究院 Movement prediction method, intelligent terminal and storage medium
CN112800111B (en) * 2021-01-26 2022-08-02 重庆邮电大学 Position prediction method based on training data mining

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101344970A (en) * 2008-08-20 2009-01-14 浙江大学 Group dynamic tracing specific design method based on kalman filtering
CN105825297A (en) * 2016-03-11 2016-08-03 山东大学 Markov-model-based position prediction method
CN108364098A (en) * 2018-02-08 2018-08-03 重庆邮电大学 A kind of weather characteristics register on user the measure of influence
CN109511090A (en) * 2018-10-17 2019-03-22 陆浩洁 A kind of interactive mode tracing and positioning anticipation system
CN109543285A (en) * 2018-11-20 2019-03-29 山东师范大学 A kind of crowd evacuation emulation method and system of fused data driving and intensified learning
CN110972074A (en) * 2019-11-26 2020-04-07 北京诚志纪元科技有限公司 Track reconstruction method for single agent of multi-mobile agent group after finding lost position

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11538553B2 (en) * 2015-03-23 2022-12-27 New York University Systems and methods of fragment-centric topographical mapping (FCTM) to target protein-protein interactions
CN107016126A (en) * 2017-05-12 2017-08-04 西南交通大学 A kind of multi-user's model movement pattern method based on sequential mode mining
CN108804646B (en) * 2018-06-06 2021-05-11 重庆邮电大学 Point of interest sign-in prediction method integrating deep learning and factorization machine

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101344970A (en) * 2008-08-20 2009-01-14 浙江大学 Group dynamic tracing specific design method based on kalman filtering
CN105825297A (en) * 2016-03-11 2016-08-03 山东大学 Markov-model-based position prediction method
CN108364098A (en) * 2018-02-08 2018-08-03 重庆邮电大学 A kind of weather characteristics register on user the measure of influence
CN109511090A (en) * 2018-10-17 2019-03-22 陆浩洁 A kind of interactive mode tracing and positioning anticipation system
CN109543285A (en) * 2018-11-20 2019-03-29 山东师范大学 A kind of crowd evacuation emulation method and system of fused data driving and intensified learning
CN110972074A (en) * 2019-11-26 2020-04-07 北京诚志纪元科技有限公司 Track reconstruction method for single agent of multi-mobile agent group after finding lost position

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
《A Novel Spatial-Temporal Fusion Framework Based on Object Trajectories》;Canghong Jin;《2019 International Conference on Intelligent Computing, Automation and Systems (ICICAS)》;20200402;全文 *
《Personalized Check-in Prediction Model Based on User’s Dissimilarity and Regression》;Chang Su;《IEEE Access ( Volume: 7)》;20190617;全文 *
基于多层聚类的自适应人群分群检测算法;程祥等;《仪表技术》;20190515(第05期);全文 *
基于并行模式挖掘和路径匹配的用户位置预测;许贤泽等;《东北大学学报(自然科学版)》;20200615(第06期);全文 *
基于数据挖掘的移动用户出行轨迹预测;刘丽娴等;《移动通信》;20190515(第05期);全文 *

Also Published As

Publication number Publication date
CN111988744A (en) 2020-11-24

Similar Documents

Publication Publication Date Title
CN110928993B (en) User position prediction method and system based on deep cyclic neural network
CN108875007B (en) method and device for determining interest point, storage medium and electronic device
CN107247961B (en) Track prediction method applying fuzzy track sequence
EP3241370B1 (en) Analyzing semantic places and related data from a plurality of location data reports
CN107018493B (en) Geographic position prediction method based on continuous time sequence Markov model
CN110414732B (en) Travel future trajectory prediction method and device, storage medium and electronic equipment
US20100111372A1 (en) Determining user similarities based on location histories
CN111988744B (en) Position prediction method based on user moving mode
CN111241419B (en) Method for recommending next interest point based on user relation embedded model
Niu et al. An online-traffic-prediction based route finding mechanism for smart city
Chen et al. Predicting next locations with object clustering and trajectory clustering
CN113139140B (en) Tourist attraction recommendation method based on space-time perception GRU and combined with user relationship preference
CN111444243A (en) User behavior prediction image method and system based on track information
CN111141301A (en) Navigation end point determining method, device, storage medium and computer equipment
CN114461931A (en) User trajectory prediction method and system based on multi-relation fusion analysis
CN115952355A (en) Next interest point recommendation system based on space-time power law attention
CN112818262A (en) Map POI searching method, system, device and medium based on user data
CN115774819A (en) Interest point recommendation method and system based on hierarchical recurrent neural network
CN110944295B (en) Position prediction method, position prediction device, storage medium and terminal
Sun et al. Predicting future locations with semantic trajectories
CN116150511A (en) Next position recommending method based on space-time context and category preference
CN115495661A (en) Self-adaptive interest point recommendation method based on long-term and short-term preference of user
CN115510333A (en) POI prediction method based on space-time perception and combined with local and global preferences
CN113032688B (en) Method for predicting access position of social network user at given future time
JP5422539B2 (en) Action prediction method, apparatus, and program

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant