Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

2020 International Conference on Computer Science and Its Application in Agriculture (ICOSICA)

Recommendation System based on Skyline Query:


Current and Future Research
Ruhul Amin Taufik Djatna Annisa
Computer Science Agroindustrial Engineering Computer Science
IPB University IPB University IPB University
Bogor, Indonesia Bogor, Indonesia Bogor, Indonesia
ruhulamin@apps.ipb.ac.id taufikdjatna@apps.ipb.ac.id annisa@apps.ipb.ac.id

Imas Sukaesih Sitanggang


Computer Science
IPB University
Bogor, Indonesia
imas.sitanggang@apps.ipb.ac.id

Abstract— The availability of information and the increasing fields, including decision support systems that involve many
amount of information resulted in users having difficulty getting criteria, travel, and electronic commerce. In the field of
objects based on their preferences. The main objective of this electronic commerce, for example, the recommendation
paper is to describe the basic skyline query algorithm and its system is used as a decision support system for choosing
application to the recommendation system. The method that is products that have many attributes to consider [6]. Skyline
used in this research is the Systematic Literature Review (SLR) queries have the advantage of filtering unattractive data by
method. The SLR method consisted of procedures to cluster and selecting a set of points that are not dominated by the others
classify specific algorithms issue, to integrate and evaluate focal from a data set that we use [7]. Another example of its
key point from previous research, and to claim previously
application in the analysis of social networks on Twitter by
unknown research gap for future works. The result of this
research shows the emergence of an idea to create a real time
providing friendship recommendations based on the closeness
recommendation system for individual preferences based on between users (accounts) [8]. Some skyline query research [9]
very large dataset both numeric and labeled data by using this research uses the K-dominant skyline query algorithm to
skyline queries, particularly in form of streaming data and the display exciting objects according to user preferences applied
existence of complexity of the algorithm to concern dealing to incomplete data. [8] His study used a skyline sort filter
current and future challenges. (SFS) algorithm to find the closeness between vertices in a
multi-layer graph. [10] This study uses a parallel processing
Keywords— Big Data, recommendation system, skyline query algorithm using the skyline query and MapReduce methods to
process Big Data.
I. INTRODUCTION
A recommendation system based on skyline query can help
The recent development of query optimization relates to users to get very high valuable knowledge and information
very large database operation both in big data and cloud such as individual products preference that match to certain
ecosystem have attracted attention for many real life hidden content in the attributes attached to the product to be
application in the last two decades. Furthermore the increasing selected. It is in the discussion of special topics discussing the
size of the data attributes have led researchers to look for new recommendation system using skyline queries. The author
methods for processing data for efficacy and to obtain sound divides several separate sections including a discussion of the
and useful information [1]. In general, with the query recommendation system, a discussion of skyline queries and a
optimization and application in the electronic commerce critical review of the recommendation system research using
websites offer prospective buyers of the same product in large skyline queries
quantities, and candidates can choose from products that have
similarities. However, new problems arise because potential II. RECOMMENDATION SYSTEM
buyers have difficulty in choosing products. After all, most
product choices are offered to prospective buyers [2]. The A. Definition
recommendation system appears to provide knowledge of A recommendation system is an important part of current
solving such problem that is desired by the user based on artificial intelligence application in providing suggestions for
information derived from user interaction with the product or an item that is beneficial to users [11], and understanding
service provided [3]. customer preferences is an important part of online sales [12].
The results of the recommendation system are suggestions that
Knowledge and information extracted from high
are worth considering by users [13]. In his research [2] states
multidimensional data lately are widely discussed because the
user preferences are influenced by factors other than the
user’s preference potentially appears from the chosen data
attributes of the product to be selected. The recommendation
dimensions [4]. Skyline queries and recommendation systems
system generates recommendations based on personalized data
are applied based on knowledge [5], skyline queries generate
from users, for example, by providing a single product
recommendations based on user preferences, whereas
recommendation [14]. The recommendation system
recommendation systems generally work to produce
recommends products that are in accordance with user
recommendations based on prior knowledge. Skyline query is
preferences so that the recommendation system can increase
present as a recommended method to be applied in several

978-1-7281-6907-1/20/$31.00
Authorized ©2020
licensed use limited to: UNIVERSITAS IEEE
GADJAH MADA. Downloaded on February 15,2021 at 06:37:06 UTC from IEEE Xplore. Restrictions apply.
2020 International Conference on Computer Science and Its Application in Agriculture (ICOSICA)

sales volumes and profits for entrepreneurs [15]. The III. SKYLINE QUERY
recommendation system helps customers in multi-criteria Skyline queries have recently been widely used because of
decision making [14]. The recommendation system is essential their ability to help decision-makers to obtain interesting
for online users to overcome the problem of a large amount of objects from many of the criteria used [4]. Skyline queries are
information available [16]. The recommendation system is used for decision-making systems that have many criteria.
usually given to individuals who have difficulty in making Skyline operator is an extension of the Language Structuring
decisions because they do not have enough knowledge or query language (SQL) that is used to search for interesting
experience [11]. objects from multidimensional data based on user preferences
Three methods that are used in the recommendation system [15]. Some research that has been done using skyline query.
are collaborative filtering (CF), content-based filtering (CB) [9] this research uses the K-dominant skyline query algorithm
and hybrid methods [17]. In the Collaborative filtering to get user preferences on incomplete databases. [8] research
method, the system makes recommendations based on groups using a graph-based skyline database, skyline queries are used
of users who have similar preferences or tastes, usually to count vertices that are not dominated by other vertices. [10]
through ratings. The system can detect any user who has the This study uses a parallel processing algorithm using the
same interest (interest) with the current user, so the system will skyline query and MapReduce methods to process Big Data.
recommend the same item to the user. Skyline has the advantage of filtering unattractive data by
A conventional collaborative filtering (CF) method selecting a set of points that are not dominated by others from
generates recommendations for users based on product ratings a complex database [7]. An object is said to be not dominated
that are given by user groups. Product ratings are used to by other objects if its value is better on all criteria or at least on
determine user preferences for products [2] while the Content- one criterion ([21], but skyline queries have the disadvantage
Based Filtering method is able to recommend similar items to of comparing each point to other points to get a skyline object
users based on items that have been previously purchased, so so that efficiency depends on the number of dimensions on the
this method does not depend on the product rating given by data set [21]. The number of skyline objects cannot be
users to produce recommendations [11]. This collaborative determined by the user, in the worst case, the number of
filtering (CF) method and the Content-Based Filtering method skyline objects can be large, especially for multidimensional
have the disadvantage that it will not function if there is data sets, thereby subtracting from the value of the
insufficient data to calculate the product rankings that are recommendation itself [7].
given by user groups, this is known as the cold-start problem A. Block Nested Loops (BNL)
[2]. The collaborative filtering (CF) method has two main
challenges that must be resolved, namely sparsity data and The BNL algorithm performs a dominance test by
scalability. Sparsity occurs because the rating is small when It comparing one by one of all candidates to produce skyline
is compared to the total number of items available, making it objects [22]. Terminology that must be known in the BNL
difficult to find other users who give similar ratings on the algorithm is a dominant and incomparable object. Dominant
same item [3]. Scalability occurs in large communities and can be interpreted as an object that is not dominated by other
datasets and continues to increase over time, thus requiring objects so that it can be a skyline point, while incomparable,
high complexity [18]. Several studies that are used that is, objects that do not dominate each other. The advantage
collaborative filtering (CF) methods have been conducted [3] of the BNL algorithm is that it is easy to implement and runs
in this study analyzing the differences in algorithms in the well on small data sets. The weakness of the BNL algorithm
recommendation system. is the complexity of the time because this algorithm compares
Content-based filtering systems work to provide product one by one between one point and another point on the dataset
recommendations that are similar to those liked by users in the that is owned, so this algorithm is not good when It is applied
past [19]. Item similarity is calculated based on the features to large data sets [21].
associated with the item being compared. For example, if a B. Sort Filter Skyline (SFS)
user has positively assessed a film in the comedy genre, the
SFS algorithm does presort or sorting data first by adding
system will learn to recommend other films from the comedy
up all the attribute values or finding entropy values, then
genre to that user. Hybrid methods that combine the two
determining the sequence of dominance testing as in the
previous methods to produce more complex recommendations
Block Nested Loop (BNL) algorithm, and then the data is
[13]. The recommendation system with collaborative filtering
compared one by one [22]. The advantages of the SFS
(CF) and content-based filtering (CBF) methods is very
algorithm minimize the number of candidate comparisons by
dependent on the user's transaction to provide a rating, so this
presorting or sorting the data first before conducting
raises the problem shown in Table I.
dominance testing by making comparisons one by one on
TABLE I. PROBLEMS WITH THE RECOMMENDATION each object [22]. Lack of SFS algorithm is to find the point of
METHOD
the skyline by making comparisons one by one on each object
No Problem Description so that this affects the resolution if faced with multiple
1. Cold-start circumstances where the system does not have objects.
historical data from new users, because there have not
been transactions between users and items, so there is C. Divide and Conquer (D and C)
not enough data for the system to produce
recommendations that are relevant to users [20]. The D&C algorithm recursively divides the input data set
2. Sparsity Due to user rating data that are rarely compared to the in the partition m {P1, ..., m} [22]. Next, do a search for
total number of items available, it is difficult to find
other users who are rating similar to the same item skyline points on each partition (Fig. 1) and finally merge
[3]. partitions into one to get a skyline point by comparing the
3. Scalability This problem often occurs in communities and large skyline points of each partition (Fig. 2). In this algorithm,
datasets and continues to increase over time, so it
requires high complexity [18].
points that are dominated by other points can be deleted
immediately. The advantages of this algorithm divide the data

Authorized licensed use limited to: UNIVERSITAS GADJAH MADA. Downloaded on February 15,2021 at 06:37:06 UTC from IEEE Xplore. Restrictions apply.
2020 International Conference on Computer Science and Its Application in Agriculture (ICOSICA)

set into several partitions so that it is faster to get a skyline efficiency of solving problems in the nearest neighbor
point and requires a smaller storage allocation. Weaknesses algorithm. The study was conducted by [8] using the NN
need time to find skyline points in each sub-partition and need algorithm on multi-layer graphs with the aim of determining
time to combine sub-partitions to get skyline points. a skyline object in the form of a collection of vertices that are
not dominated by other vertices based on the shortest distance
between vertices in all graph layers, and optimization is done
to speed up the work of algorithm in a different graph. Other
research conducted by [25] proposes the development of the
NN algorithm by using a tree structure to find nearby objects
that involve spatial and non-spatial data objects, the results of
the study show that the proposed method is more efficient at
the memory requirements which are the main problem
wireless network when sending data to many points.

Fig. 1. Merging process

Fig 3. Examples of skyline queries and nearest neighbor skyline

Each point in Fig. 3(a) represents one hotel. Each user has
a preference when choosing a hotel, and if the preference is a
hotel that is cheap and close to the beach, then it can be
Fig. 2. Bushy merge tree concluded if hotel G dominates or is better than hotel D
because based on the price of $ 50 and distance to the beach
D. Skyline Nearest Neighbor (NN)
2.3 Km, while hotel D has a price of $ 10.5 and distance to the
Algorithms uses the R-Tree (NN-tree) structure, and the beach 2.8 Km. On the other hand, hotel G and K do not
advantage of the R-tree is that it can be used for dominate each other (incomparable), or each one dominates
multidimensional data set operations and supports operations on one dimension only. When viewed from the coordinates of
on point and spatial data [23]. The NN algorithm looks for the hotel G ($ 50;2.8 km) and hotel K ($ 120;0.8 km), in terms of
nearest point or minimum distance between the skyline points price hotel G dominates hotel K, but in terms of distance the
and the closest point. Some applications of the NN algorithm hotel to the beach hotel K dominates hotel G. If you return to
[22], the nearest neighbor algorithm can provide answers to the definition domination, that is, an object is said to not be
the point of interest (POI) problems that are in accordance dominated by another object if the value is better on all criteria
with customer preferences when the user activates GPS on his or at least on one criterion [21]. Based on the definition of
mobile device [24]. dominance, the skyline displays a skyline point consisting of
Skyline nearest neighbor can solve the problem of points A, G, K, N, and T.
searching for nearby data objects that involve spatial and non- Fig. 3(b) shows hotels that are closest to hotels that are
spatial specifications for data objects that come from real-time being skyline points. Chain Hotels 1 and 2 are not skyline
applications [25]. In the multi-layer graph, the nearest objects because the price is a bit higher and the distance is
neighbor skyline is used to find the closest object to the slightly further from the beach, but on other issues that must
skyline object by finding the most dominant node in all layers be facilitated by user requests if one of the hotels that become
in the graph [8]. In his research [26] states the advantage of the skyline point is K the room choice is full or occupied by
the nearest neighbor skyline algorithm is to find the closest other users, so the user thinks to look for another hotel that is
point of the skyline object when the skyline object is nearby or resembles the K hotel that becomes the previous
considered to be less giving recommendations to customers choice, is based on the coordinates in Fig. 3(b) Chain 1 and
when making decisions. Skyline nearest neighbor is a type of Chain 2 hotels can be an alternative used to be another choice.
query that is widely used in geographic information systems
to find the nearest object from a certain point, in this study Research conducted by [27] states that the most efficient
using the R-Tree algorithm. method currently is the Branch-and-Bound Skyline (BBS),
which is the development of the nearest neighbor (NN)
Searching for nearby objects using the NN algorithm is algorithm that works to find the closest object using the tree
widely used in several fields, including databases, computer index structure, the results his research shows that the BBS
vision, multimedia, machine learning and recommendation algorithm is easy to implement and can improve performance.
systems [8]. Several related studies conducted by [26] this Fig. 4 shows the tree structure with node capacity = 3, for an
study developed the nearest neighbor algorithm (NN) with the object with dimension n, the rectangle is the interval in n
aim of finding the closest object from the skyline point to dimensions that includes the objects [28].
provide answers that are easier to use by users in making
decisions, from research showing the effectiveness and

Authorized licensed use limited to: UNIVERSITAS GADJAH MADA. Downloaded on February 15,2021 at 06:37:06 UTC from IEEE Xplore. Restrictions apply.
2020 International Conference on Computer Science and Its Application in Agriculture (ICOSICA)

Refer Problem Aim Analysis and


ence Synthesis
[8] How to find the closest calculates a set In this study, the
object to of vertices that authors
the skyline point in are not performed an
a multi-layer graph dominated by optimization
other vertices using the
in terms of the Dijkstra
shortest algorithm, so
distance in all there was no
layers of the need to search
graph for nearby
objects one by
one.
[2] There are other factors Use clickstrea User ratings for
that affect the m on electronic products and
preferences of users so commerce sites user history on
that it affects the to get user electronic
recommendation history commerce sites
Fig. 4. Example of R-Tree index structure system. records. can be a primary
method for
The nearest neighbor (NN) algorithm searches for the recommendation
nearest object based on the skyline object with indeed the s.
search space using the R-tree structure [29]. Table II [29] How to reduce skyline Consider all By using the K-
objects in the attributes dominate
summarizes the advantages and disadvantages of each of the multidimensional compared to algorithm and
basic skyline query algorithms. datasets get a skyline comparison
object. done in parallel,
TABLE II. SUMMARY OF BASIC SKYLINE QUERY
we get fewer
ALGORITHMS
skyline objects,
Algorithm Type Sorting Advantages Deficiency and comparison
BNL Scan x Small memory Compare all in parallel can
requirements points in the cut time.
data set [30] Volume and Implement skyl The use of the
D&C Scan x Suitable for Allocation of veracity are the main ine queries to right algorithm
multidimensional time problems in managing find objects will affect the
data sets searching for information. that dominate search
skyline points from the for skyline
SFS Scan Entropy Order before Read all data resource points.
scoring comparing sets description
function framework
NN Index The Suitable for I/O accesses (RDF)
closest multidimensional [12] Collaborative Application of Hybrid Case-
distance objects filtering algorithm has hybrid method Based
to the two main challenges, based on item- Reasoning
origin namely sparsity and based CF for (CBR) method
BBS Index The Suitable for Many do scalability more combined with
closest multidimensional checks for personalized average filling is
distance objects dominance. product used to handle
to the recommendati sparsity data
origin ons for users sets, while Self-
Organizing Map
(SOM) is
Table III presents a summary of the analysis and synthesis of optimized with
scientific articles originating from international scientific journals, Genetic
from several articles collected by the author determining 25 articles Algorithm (GA)
to group users in
directly related to the recommendation system using skyline queries, large data sets to
especially in the problem section of the research discussion. I noted reduce the scope
the problems and objectives of the 25 articles in Table III. Next we of item-based
analyzed and synthesized each article in the hope to get new CF
[31] How to process, Provides an Graph databases
knowledge and ideas if applied to this research. analyze and extract understanding have
meaningful of how to advantages
TABLE III. SUMMARY OF ANALYSIS AND SYNTHESIS OF information from the obtain when compared
RELATED ARTICLES graph-based (mining) to some
Refer Problem Aim Analysis and sized databases important relational databa
ence Synthesis information ses, which
[26] How to find the closest The The previous from a graph- support storage
object to application of method based database of unstructured
the skyline point the NN development data, scalable
algorithm uses using R-Tree (Big Data), fast
R-tree was improved searching, extrac
by M + tree, it tion of patterns
was obtained or hidden
that results were information,
more efficient in building
getting objects information
closest to from some data.
the skyline point

Authorized licensed use limited to: UNIVERSITAS GADJAH MADA. Downloaded on February 15,2021 at 06:37:06 UTC from IEEE Xplore. Restrictions apply.
2020 International Conference on Computer Science and Its Application in Agriculture (ICOSICA)

Refer Problem Aim Analysis and Refer Problem Aim Analysis and
ence Synthesis ence Synthesis
[32] How to determine Application of The use of the from facilities
the skyline semantics right algorithm that are not cold,
object in the uncertain and refining m is useful for so this helps the
database ethods to prove cutting time user in making a
in theory (efficiently), and decision to find
not a lot of a business
memory uses in location
processing. [37] How to analyze large The In this study, K-
[33] How to get information Application This study data application of dominant
patterns or extract data of graph minin uses a graph- reverse skyline algorithm has
from multi- g for multi- based database t the advantage of
relational databases relational datab o process and controlling the
ases analyze generated
operational data, skyline object.
the problem of [32] Conventional Skyline The spatial Spatial skyline
ever-increasing queries cannot answer expectation of queries can
data volumes the user's need to move skyline optimize the
and the the community to the queries for efficiency of
interrelation area they want smart cities. community
between data redirection to
from one locations
another to a deemed suitable
graph- or in accordance
based database with user
[22] How to find skyline Application Temporal Datab preferences.
objects of skyline ase is a solution [38] How to find skyline The Skyline join can
in a temporal database queries to filter in handling data points on application of solve the
a set of points in the past Skyline join problem of
that meet the and in the query on finding skyline
requirements future, using the multiple points without
based on a set skyline query relations having to
of evaluation can get essential compare records
criteria from a objects in the one by one to
potentially past data as a get skyline
large points basis for making points
dataset. decisions for [39] How to find skyline Implementatio The application
now. points in n of the RSSA of the RSSA
[34] How does a skyline Application of Skyline queries, multidimensional data algorithm for algorithm with
query look for areas the TFSS in addition to the the R-tree
that are away from algorithm with generating development structure can
unpleasant facilities the skyline points, of the RSQ make searching
and far from development can also be used algorithm more efficient,
competitors to be a of the BBFS to search for advantages over
business location algorithm skyline areas so the R-tree can
that they are be used for
suitable for multidimensiona
finding l dataset
alternative operations and
locations desired supports
by users based operations on
on their point and spatial
preferences. data
[35] How to determine the Application of The purpose of a [7] How to find skyline The Computation is
suitable area for real skyline area to Reverse Skyline points in implementatio done in parallel
estate built. get suitable query is to multidimensional data n uses the K- to find the
areas identify the dominant skyline point,
according to effect of the algorithm. parallel
user query object on computing using
preferences the MapReduce will
multidimensiona produce a faster
l data set with time
respect to [24] How to find the closest Application The use of a
distance point to the object to be of the nearest mobile device is
vectors. An used as a neighbor very helpful in
example of a recommendation algorithm to determining the
Reverse Skyline get the point of user's position
query is Branch interest (POI) based on GPS,
and Bound from this can
Reverse Skyline provide
(BBRS). recommendation
[36] How skyline queries Application of skyline area can s to users based
choose a good location skyline area to be used to find on their
and according to get suitable the best preferences.
the preferences of the areas accordin location, for [25] How to find the closest Applying This study
user g to user example, to find point to the object to be the nearest develops the
preferences a location that is made based on spatial neighbor nearest neighbor
close to the and non-spatial data algorithm to (NN) algorithm
station and also get the closest with the aim of

Authorized licensed use limited to: UNIVERSITAS GADJAH MADA. Downloaded on February 15,2021 at 06:37:06 UTC from IEEE Xplore. Restrictions apply.
2020 International Conference on Computer Science and Its Application in Agriculture (ICOSICA)

Refer Problem Aim Analysis and Refer Problem Aim Analysis and
ence Synthesis ence Synthesis
object to the finding the to serial
skyline object closest object calculation
from the skyline methods.
point to provide
answers that are
easier to use by
Based on the articles that have been reviewed in Table III by
users in making the author, the advantages and disadvantages of the skyline
decisions, from query algorithm are found to be used to solve the problem.
research The Top-K algorithm has the advantage of not depending on
showing the
effectiveness
the value function of the data and is effective in determining
and efficiency of the best object and this algorithm can be run in the
solving MapReduce framework, while the drawback is that this
problems. algorithm requires an objective function from the user to
[40] An increase in data size Application of In this study
obtain maximum or minimum results for the skyline object.
causes the skyline to skyline join comparing
join to not be applied. queries points with one The K-dominant algorithm has the advantage of returning
on distributed another in fewer skyline points to the user by changing k values that are
databases. parallel smaller than n. The value of K is that many objects dominate
in a distributed
other objects and n is the amount of data to be compared.
database, so that
they can solve Meanwhile, the deficiency in K-dominant algorithm is that it
problems more performs multiple comparisons on the dataset to remove the
quickly. dominant points, so this takes time. The R-tree algorithm has
[41] Look for skyline points Application of The top K the advantage of being used for data objects that have many
on uncertain data. Top k algorithm works
probabilistic to get the best K attributes or is multidimensional, while the disadvantage is
skyline queries object that that this algorithm compares the distance from one point to
on uncertain minimizes another to compare the dominance between two points so it
data. certain takes a lot of time.
preference
functions so that
it can minimize
IV. CONCLUSION
the work of The recommendation system using the skyline query
memory.
method is a solution to provide appropriate advice based on
[42] How to determine the Application of This research
region of interest (ROI) skyline uses skyline to preferences. Skyline query is an algorithm to get objects that
using a skyline query with graph- search for are not dominated by other objects, is able to extract
based data to skyline objects interesting objects from multidimensional data based on user
get ROI based on user preferences, a recommendation method to get a small number
preferences and
uses graphs to of interesting objects based on user preferences called skyline
trim the route objects. Algorithms in the skyline are grouped by type and
used to get to type. Each algorithm in a skyline query has a different way of
the area of working on getting a skyline object according to user
interest.
[43] How to find interesting Application of Data mining and preferences. Each algorithm in the skyline query method has
data patterns from a data mining machine a different way of working. Each algorithm must be adjusted
graph-based database (data mining) learning to the problem or challenge to be solved. Some studies suggest
to obtain become, data the most efficient method currently is the Branch-and-Bound
information mining is used
from a graph- to find patterns Skyline (BBS), which is the development of the nearest
based database of data stored neighbor (NN) algorithm that works to find the nearest object
using the in a graph- using the tree index structure, the results of the study show
Framework based database a that BBS is easy to implement and can solve the problem of
nd make it
possible to ever-increasing data volumes that causes high complexity.
process, to
analyze, and to
extract REFERENCES
meaningful
[1] C. Kalyvas and T. Tzouramanis, “A Survey of Skyline Query
information
Processing,” Arxiv, vol. 1704.01788, pp. 1–127, 2017.
from large
[2] J. Iwanaga, N. Nishimura, N. Sukegawa, and Y. Takano, “Electronic
amounts of
Commerce Research and Applications Improving collaborative
graph data.
filtering recommendations by estimating user preferences from
[44] How to improve the Implement Parallel and clickstream data,” Electron. Commer. Res. Appl., vol. 37, no. July, p.
performance of skyline MapReduce to distributed 100877, 2019.
query processing for improve the computing have [3] B. Sarwar, G. Karypis, J. Konstan, and J. Riedl, “Item-Based
large database performance the advantage of Collaborative Filtering Recommendation,” ACM, 2001.
of skyline showing [4] K. Alami and S. Maabout, “Data & Knowledge Engineering A
query process efficient framework for multidimensional skyline queries over streaming data,”
sing solutions by Data Knowl. Eng., no. January, p. 101792, 2020.
shortening data [5] S. Shah, A. Thakkar, and S. Rami, “A Survey Paper on Skyline Query
processing time using Recommendation System,” Int. J. Data Min. Emerg. Technol.,
and the results vol. 6, no. 1, p. 1, 2016.
of research show [6] G. Bordogna and G. Pasi, “A flexible multi criteria information
acceleration filtering model,” Soft Comput., vol. 14, no. 8, pp. 799–809, 2010.
when compared

Authorized licensed use limited to: UNIVERSITAS GADJAH MADA. Downloaded on February 15,2021 at 06:37:06 UTC from IEEE Xplore. Restrictions apply.
2020 International Conference on Computer Science and Its Application in Agriculture (ICOSICA)

[7] H. Tian, A. Siddique, and Y. Morimoto, “An Efficient Processing of 6, pp. 946–955, 2016.
k-Dominant Skyline Query in MapReduce,” ACM, 2014. [37] M. S. S. Basha and A. M. Abbas, “Efficient Reverse Skyline
[8] W. Liu, D. Wen, H. Wang, F. Zhang, and X. Wang, “Skyline nearest Algorithm For Discovering Top K-Dominant Products,” J. Theor.
neighbor search on multi-layer graphs,” Proc. - 2019 IEEE 35th Int. Appl. Inf. Technol., vol. 80, no. 2, pp. 227–246, 2015.
Conf. Data Eng. Work. ICDEW 2019, pp. 259–265, 2019. [38] J. Zhang, Z. Lin, B. Li, W. Wang, and D. Meng, “Efficient skyline
[9] J. Lee, H. Im, and G. W. You, “Optimizing skyline queries over query over multiple relations,” Procedia Comput. Sci., vol. 80, pp.
incomplete data,” Inf. Sci. (Ny)., vol. 361–362, pp. 14–28, 2016. 2211–2215, 2016.
[10] J. Kim and M. H. Kim, An efficient parallel processing method for [39] Y. Gao, Q. Liu, B. Zheng, and G. Chen, “On efficient reverse skyline
skyline queries in MapReduce, vol. 74, no. 2. Springer US, 2018. query processing,” Expert Syst. Appl., vol. 41, no. 7, pp. 3237–3249,
[11] F. Ricci, L. Rokach, and B. Shapira, Recommender Systems 2014.
Handbook, Second. New York: Springer, 2015. [40] M. Bai, J. Xin, G. Wang, R. Zimmermann, and X. Wang, “Skyline-
[12] N. P. Kumar and Z. Fan, “Hybrid User-Item Based Collaborative join query processing in distributed databases,” Front. Comput. SCi,
Filtering,” Procedia - Procedia Comput. Sci., vol. 60, pp. 1453–1461, pp. 1–23, 2015.
2015. [41] Z. Yang, K. Li, X. Zhou, J. Mei, and Y. Gao, “Top k probabilistic
[13] R. Burke, “Hybrid Web Recommender Systems,” Springer, pp. 377– skyline queries on uncertain data,” Neurocomputing, vol. 317, pp. 1–
378, 2007. 14, 2018.
[14] M. Nilashi, O. Ibrahim, E. Yadegaridehkordi, S. Samad, E. Akbari, [42] S. Pande, S. Ranu, and A. Bhattacharya, “SkyGraph: Retrieving
and A. Alizadeh, “Travelers Decision Making Using Online Review regions of interest using skyline subgraph queries,” Proc. VLDB
in Social Network Sites: A Case on TripAdvisor,” J. Comput. Sci., Endow., vol. 10, no. 11, pp. 1382–1393, 2017.
2018. [43] B. Boden, S. Günnemann, H. Hoffmann, and T. Seidl, “Mining
[15] A. Agarwal, A. Aggarwal, and A. Agarwal, “An Approach for Coherent Subgraphs in Multi-Layer Graphs with Edge Labels,” in
Augmenting Selection Operators of SQL Queries using Skyline and international conference on Knowledge discovery and data mining,
Fuzzy-Logic Operators,” Procedia Comput. Sci., vol. 115, pp. 14–21, 2012, pp. 1258–1266.
2017. [44] Y. Li, Z. Li, M. Dong, W. Qu, C. Ji, and J. Wu, “Efficient subspace
[16] D. Gavalas, C. Konstantopoulos, K. Mastakas, and G. Pantziou, skyline query based on user preference using MapReduce,” Ad Hoc
“Journal of Network and Computer Applications Mobile recommender Networks, vol. 35, pp. 105–115, 2015.
systems in tourism,” J. Netw. Comput. Appl., pp. 1–15, 2013.
[17] P. Symeonidis, D. Ntempos, and Y. Manolopoulos, Recommender
Systems for Location-based Social Networks. New York: Springer
New York LLC, 2014.
[18] T. George, “A Scalable Collaborative Filtering Framework based on
Co-clustering,” in International Conference on Data Mining, 2005, pp.
3–6.
[19] M. J. Pazzani and D. Billsus, “Content-Based Recommendation
Systems,” Springer, pp. 325–326, 2007.
[20] A. I. Schein, A. Popescul, L. H. Ungar, and D. M. Pennock, “Methods
and Metrics for Cold-Start Recommendations,” ACM, pp. 253–260,
2002.
[21] S. Borzsonyil and K. Stocker, “SkylineOperator,” IEEE, pp. 421–430,
2001.
[22] C. Kalyvas, T. Tzouramanis, and Y. Manolopoulos, “Processing
Skyline Queries in Temporal Databases,” ACM, pp. 893–899, 2017.
[23] H. P. Kriegel, B. Seeger, R. Schneider, and N. Beckmann, “The R-tree:
an efficient and robust access method for points and rectangles,” ACM,
pp. 448–455, 1990.
[24] S. Choi, G. Ghinita, H. S. Lim, and E. Bertino, “Secure kNN query
processing in untrusted cloud environments,” IEEE Trans. Knowl.
Data Eng., vol. 26, no. 11, pp. 2818–2831, 2014.
[25] H. Jung, Y. D. Chung, and L. Liu, “Processing generalized k-nearest
neighbor queries on a wireless broadcast stream,” Inf. Sci. (Ny)., vol.
188, pp. 64–79, 2012.
[26] Y. C. Chen and C. Lee, “The σ-neighborhood skyline queries,” Inf.
Sci. (Ny)., vol. 322, pp. 92–114, 2015.
[27] D. Papadias, Y. Tao, G. Fu, and B. Seeger, “An Optimal and
Progressive Algorithm for Skyline Queries,” Proc. ACM SIGMOD Int.
Conf. Manag. Data, pp. 467–478, 2003.
[28] D. Papadias, G. Fu, and B. Seeger, “Progressive Skyline Computation
in Database Systems Hong Kong University of Science and
Technology City University of Hong Kong,” ACM, vol. 30, no. 1, pp.
41–82, 2005.
[29] Y. W. Peng and W. M. Chen, “Parallel k-dominant skyline queries in
high-dimensional datasets,” Inf. Sci. (Ny)., vol. 496, pp. 538–552,
2019.
[30] A. Abidi, S. Elmi, M. A. Bach Tobji, A. HadjAli, and B. Ben
Yaghlane, “Skyline queries over possibilistic RDF data,” Int. J.
Approx. Reason., vol. 93, pp. 277–289, 2018.
[31] S. Aridhi and E. Mephu Nguifo, “Big Graph Mining: Frameworks and
Techniques,” Big Data Res., vol. 6, pp. 1–10, 2016.
[32] S. Elmi and J. K. Min, “Spatial skyline queries over incomplete data
for smart cities,” J. Syst. Archit., vol. 90, no. April, pp. 1–14, 2018.
[33] C. Fan, F. Xiao, M. Song, and J. Wang, “A graph mining-based
methodology for discovering and visualizing high-level knowledge for
building energy management,” Appl. Energy, vol. 251, no. January, p.
113395, 2019.
[34] G. W. You, M. W. Lee, H. Im, and S. W. Hwang, “The farthest spatial
skyline queries,” Inf. Syst., vol. 38, no. 3, pp. 286–301, 2013.
[35] Annisa, A. Zaman, and Y. Morimoto, “Reverse Area Skyline in a
Map,” Int. J. Adv. Comput. Sci. Appl., vol. 8, no. 2, 2017.
[36] Annisa, A. Zaman, and Y. Morimoto, “Area Skyline Query for
Selecting Good Locations in a Map,” Inf. Process. Vol.24, vol. 24, no.

Authorized licensed use limited to: UNIVERSITAS GADJAH MADA. Downloaded on February 15,2021 at 06:37:06 UTC from IEEE Xplore. Restrictions apply.

You might also like