IRJET- Next Location Prediction

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 07 Issue: 02 | Feb 2020 www.irjet.net p-ISSN: 2395-0072
© 2020, IRJET | Impact Factor value: 7.34 | ISO 9001:2008 Certified Journal | Page 1018
Next Location Prediction
Ashvini M. Gupta1, Pratiksha M. Khedkar2, Nikita R. Choudhari3, Amar G. Waghade4
1,2,3Student, Dept. of Computer Science & Engineering, DES’S COET, Dhamangaon Rly
4Professor, Dept. of Computer Science & Engineering, DES’S COET, Dhamangaon Rly
---------------------------------------------------------------------***----------------------------------------------------------------------
Abstract - Location information of Web pages plays an
important role in location-sensitive tasks such as Web search
ranking for location-sensitive queries. However, such
information is usually ambiguous, incompleteorevenmissing,
which raises the problem of location predictionforWebpages.
Meanwhile, Web pages are massive and often noisy, which
pose challenges to the majority of existing algorithms for
location prediction. While this problem has been explored
earlier in the batch data set-up, we propose in this paper new
solutions in the streaming data set-up. We examine four
incremental learning methods using aDampedwindowmodel
namely, Multivariate multiple regression, Spherical-spherical
regression, Randomized spherical K-NN regression and an
Ensemble of these methods fortheireffectivenessinsolving the
destination prediction problem. This application allows the
user to plan their travel ahead by providing a route for the
destination and facilities near the location and also routes to
these locations with the help of Google Maps.
Key Words: Location Prediction, Data Analysis, k-
means, K-NN.
1. INTRODUCTION
Location-based Service (LBS) is a kind of information
service that provides usersgeographical positionslocatedby
mobile devices and wireless network. There exists wealth
information in the location data, such as user’s interests,
user’s hobbies and user’s behavior pattern. LBS may be
employed in a number of applications, including: location-
based advertising, personalized weather services,
entertainment, personal life and so on. An effective location
prediction or recommendation can make the users have
good experience. Location-based social network services
(LBSNs) have been booming during the past five years.
Nowadays, it is common for a user to attach his location
when he publishes a photo or a status using his online social
network account. Moreover, users may just share their
locations, called check in, to tell their friends where they are
or to engage in social games as in Foursquare.
Since these large amount of location and social relation
data become available, studying human mobility and its
connection with social relationships becomesquantitatively
achievable. Understanding human mobility can lead to
compelling applications including locationrecommendation
urban planning, immigration patterns, etc. Previous works,
including, show that human mobility is influenced by social
factors. However, there is onecommonshortcoming:they all
treat friends of users equally. Similar to other social
behaviors, in most cases mobility is influenced by specific
communities but not all friends. For example, the
aforementioned colleagues can influence the place a user
goes for lunch but probably have nothing to do with his
weekend plans. Meanwhile, where a uservisitsonweekends
largely depends on his friends or family, but not his
colleagues.
Therefore, the impact on a user’s mobility should be
considered from the perspectives of communities instead of
all friends. In a broader view, community is arguably the
most useful resolution to study social networks. Real-time
location prediction can have other interesting applications
for the transport dispatch system such as in vehicle
allocation for the future rides (Powell, 1986, 1987),
diversion in real time (Ichoua et al., 2000), ride-sharing
(Tran, 2015) and reduction of the total idle time (Miao et al.,
2015). Powell (1986) is a key reference on the dynamic
vehicle allocation problem which has seen many extensions
and applications over the years. In recent years, we have
seen that several on-demand transportationservicessuchas
Uber (i.e. UberPool), Ola (i.e. OLA Share), etc. provide ride-
sharing as well. Tran (2015) worked on the real-time ride-
sharing schedule and dispatch problembyusingthelocation
information of the passengers who requested a pickup and
the location of drivers in the near-by region. Ichoua et al.
(2000) worked on diversion issues in real-time vehicle
dispatching system. Miao et al. (2015) worked on the taxi
dispatch problem using real-time sensor information to
reduce the taxi idle driving time.
Overall, real-time location prediction helps the
transport dispatch system in increasing the reliability and
efficiency of their services which finally leads to an increase
in profits. Data generated by GPS can be put to use for tasks
such as prediction of human movement which can lead to
support for more advancedapplicationslikeadaptive mobile
services. Existing location based applications like Google
Maps provide various services like satellite images, real –
time traffic data, street view maps, satellite imagery and
planning of routes for travelling by car, two-wheelers or
public transport or by walking. Most of these location based
systems store the user's location history and their tracking
data. Some of these applications provides other useful
information such as turn by turn data for reaching a
destination, travel times and route information that is
provided by users. Location dependent information is
downloaded over a mobile network.

2. SOFTWARE TOOLS SPECIFICATION
At the same time, mobile user location prediction is also
a very difficult and challenging task. First of all, due to the
occlusion of high buildings in the city and the user’s random
switching device positioning function, it is easy to cause the
loss of some important data and there is a lack of training
data. Second, the activities of mobileusersareuncertain,and
the mobile modes exhibited by the same user may be quite
different in different time periods. Furthermore, the user’s
trajectory data is both discrete andmassive,andresearchers
cannot directly use the raw data for location prediction. The
existing research methods usually dividelocationprediction
into three steps:
(1) Preprocessing of the original trajectory data and
screening out of meaningless and abnormal location points;
(2) In the remaining location points,importantlocations
with special significance to users are constructed according
to point density, visit frequency and other information,such
as the user’s home address, work unit and other areas
inseparable from their daily life;
(3) According to the training model of user’s historical
access information to location points, the next location of
users can be predicted given the current and historical
location points of users. In terms of trajectory preprocessing
and important locationsconstruction,onlyarticlesusingGPS
data and check-in data as experimental data contains these
two steps, articles using Wi-Fi location data directly take
each access point as an important location.Similarly,articles
using data recorded by base stations also take each base
station as an important location, while articles using data
recorded by traffic bayonets take each traffic bayonet as an
important location.
The main idea of Infomapcanbesummarizedasfollows:
information flow in a network can characterize the behavior
of the whole network, which consequently reflects the
structure of the network. A group of nodes among which
information flows relatively fast can be considered as one
community. Therefore, Infomap intends to use information
flow to detect communities in a network. In the beginning,
Infomap simulates information flow in a network with
random walks. Then the algorithm partitions the network
into communities and exploitsHuffmancodingtoencodethe
network at two levels. At the community level, thealgorithm
assigns a unique code for each community based on the
information flow among different communities; at the node
level, the algorithm assigns a code for each node based on
the information flow within the community.
Although the potential location information in non
location terms are ignored. Here this paper considers both
location terms and non location terms and search queries.
We will find how many times that the locationisdisplayed in
that particular meta data content. In twitter data we have
location-tagged tweets but there will be no information
about location of web pages. The stopwords are removed by
using the steps stemming rooting etc.
Researchers often cluster trajectory sampling points
based on their density to extract important locations.
However, commonly used density clustering algorithms
usually have the shortcomings that the clustering result is
too sensitive to the input parameters. How to determine the
appropriate input parameters has become an urgent
problem to be solved. In this paper, weproposeanimproved
density peak clustering algorithm called ADPC to cluster the
feature points and extract the important locations. The
density peak clustering algorithm was a new density
clustering algorithm proposed by Rodriguez et al..
Compared with the classical density clustering
algorithms, the algorithm is simple in principle, requires
only one input parameter, and does not require iteration,
and it can cluster data of various shapes. However, it is
necessary to select the clustering center manually through
the decision graph, which not only increasestheredundancy
of the algorithm, but also causes subjective troubles. The
improved density peak clustering algorithm ADPC
incorporates the self-adaptive improvement of the cut-off
distance and the clustering center, so that it can
automatically determine the best input parameters and
clustering center according to the distribution of sampling
points.
3. IMPLEMENTATION
The Proposed system is a connected app to Moves
that predicts the destination of the user with the location
history available from Moves and provides facilities like
stores, restaurants, local government offices etc. close tothe
destination (in the case that a correct prediction has been
made) or the current location (in the case that either the
prediction is incorrect or cannot be made) within a given
radius. The authorization code obtained from the callback
url is used to get an access token through an authenticated
server-side call. Access token is required to make Moves API
calls. When a user authorizes.
To access their data, an access token is generated.
However an access token has an expiry date. For this reason
a refresh token is sent with the access token in the json
response. The refresh token is used to refresh the access
token when it’s validity expires. Using Moves API daily data
is obtained and saved to the database and with each row of
location data, the refresh token of the user’s Moves account
is also saved. In view of the above problems, the first‐order
Markov model only utilizes the current interestregionof the
user, and the utilization rate of the historical trajectory data
is not high; while the high‐ order Markov model has
problems of difficulty in matching and high sparse rate, the
historical trajectory data cannot be fullyutilizedasthesame.
Therefore, combining low‐order models with high‐order
models is a feasible method. Then, reasonably determining
the influence coefficient of eachmodel withdifferentorderis

an important prerequisite for ensuring good prediction
performance. From a macro perspective, high‐orderMarkov
model usually has higher prediction accuracy, and should
have greater “discourse power” in the fusionmodel.Inorder
to determine the specific influence coefficient of each model
with different order, a mobile user location prediction
method based on Adaboost‐Markov model is proposed in
this paper. As a representative lifting method, Adaboost
algorithm can combine a plurality of weak classifiers by
changing the probability distributionoftrainingdata and the
weight coefficient of weak classifier to generate a strong
classifier [44]. In this paper, the fusion model order k is
adaptively determined, and the first‐order to k th‐order
Markov model are used as k weak predictors. The
probability distribution of user trajectory data and the
weight coefficients of each Markov model are changed by
Adaboost algorithm, and finally a multi‐orderfusionMarkov
is generated, which is used to predict user importan
4. CONCLUSION
The application of location based service becomes moreand
more popular and it is important to provide superiorservice
for the user. The most common method for location
prediction is Markov Model. But it only considers the user’s
location movement sequence and it is impossible to give a
prediction related to the time information. So to overcome
this problem our web-based app ‘Next location prediction’
predicts the intended destination of user using location
history and machine learning algorithms and providing
route for the same. The main goal of the web application is
to allow the user to plan their travel even before they start.
This application is very user friendly that helps to solve user
travelling problem.
REFERENCES
[1] Li Y, Guo A, Liu S, Gao Y, Zheng YT. A location based
reminder system for advertisement. International
Conference on Multimedea. 2010;1501–1502.
[2] Guo B, Fujimura R, Zhang DQ, Imai M. Design-in-
play: improving the variability of indoor pervasive
games. Multimedia Tools & Applications, 2012;
59(1):259–277.
[3] Zheng Y, Xie X, Ma WY. GeoLife: A Collaborative
Social Networking Service among User, Location
and Trajectory. Bulletin of the Technical Committee
on Data Engineering. 2011; 33(2):32–39.
[4] Zheng Y. Trajectory Data Mining:AnOverview.ACM
Transaction on Intellignet SystemsandTechnology.
2015;6(3):41.
[5] Zheng K, Zheng Y, Xie X, Zhou XF. Reducing
Uncertainty of Low-Sampling-Rate Trajectories.

IEEE, International ConferenceonData Engineering.
2012;1144–1155.
[6] Xue AY, Zhang R, Zheng Y, Xie X, Huang J, Xu ZH.
Destination prediction by sub-trajectory synthesis
and privacy protection against such prediction.
IEEE, International ConferenceonData Engineering.
2013;254–265
[7] M.I. López, J.M Luna, C. Romero and S. Ventura,
“Classification via clustering for predicting final
marks based on student participation in forums”.
[8] M. Young, The Technical Writer’s Handbook. Mill
Valley, CA:University Science, 1989
[9] Economou, G., Pothos, V., and Ifantis, A. (2004).
Geodesic distance and mst based image
segmentation. In In XII European Signal Processing
Conference (EUSIPCO 2004), pages 941–944. Ellis,
B. (2014). Real - Time Analytics: Techniques to
analyze and visualize streaming data. Wiley
Publishing, 1st edition. isbn: 1118837916,
9781118837917.
[10] Fisher, N. I., Lewis, T., and Embleton, B. J. J. (1993).
Statistical Analysis of Spherical Data. Cambridge
university press.

IRJET- Next Location Prediction

More Related Content

What's hot (20)

Similar to IRJET- Next Location Prediction (20)

More from IRJET Journal (20)

Recently uploaded (20)

IRJET- Next Location Prediction