International Journal of Computer

Science and Engineering (IJCSE)

ISSN 2278-9960
Vol. 2, Issue 4, Sep 2013, 83-90
© IASET

A HYBRID RECOMMENDER SYSTEM TO PREDICT THE LOCATION RANKING

AKSHITA1 & SMITA2

1
Research Scholar, Department of Computer Science and Engineering, PDMCEW, Bahadurgarh, Haryana, India
2
Associate Professor, Department of Computer Science and Engineering, PDMCEW, Bahadurgarh, Haryana, India

ABSTRACT

Data Mining is about to perform the intelligent analysis on available statistics to derive the new results as well as
the future aspects. This future aspect analysis is formed as a prediction system or the recommender system. The
recommender system is about to identify the knowledge about the similar user or the event and derive the favorable aspect
based on it.

In this present work, a hybrid recommender system is defined to identify the most favorable location for a user to
spend his vacations. The presented hybrid model is based on the content based similarity as well as event based similarity.
This collaborative system is implemented with authenticated dataset. The obtained results from the system gives acceptable
results in terms of accuracy.

KEYWORDS: Collaborative, Hybrid, Location Based, Recommender System

INTRODUCTION

Data mining can be viewed as a result of the natural evolution of information technology. An evolutionary path
has been witnessed in the database industry in the development of the following functionalities data collection and database
creation, data management (including data storage and retrieval, and database transaction processing), and data analysis
and understanding (involving data warehousing and data mining).

With numerous database systems offering query and transaction processing as common practice, data analysis and
understanding has naturally become the next target. Efficient methods for on-line transaction processing (OLTP), where a
query is viewed as a read-only transaction, have contributed substantially to the evolution and wide acceptance of
relational technology as a major tool for efficient storage, retrieval, and management of large amounts of data. Several
data mining tasks exist [5]. Here figure 1 is showing the basic data mining process flow.

The process begins with a valid data search or the collection. Once the database is generated or retrieved, the next
work is to perform the selection of most required attribute and row set from this dataset. After this the cleaning and the pre
processing stage is seen as the filtration stage. Finally the data evaluation and interpretation will be performed on this
dataset to obtain the exact results from the system.
84 Akshita & Smita

Figure 1: Data Mining Process

Recommendation techniques are information agents that attempt to predict which items out of large pool a user
may b interested in and recommend the best one to the target user. Recommendation techniques have a number of possible
classifications.

The whole classification is broadly categorized into the personalized and non personalized recommendation, and
discusses all the personalized recommendation techniques shown in Figure 2.

Figure 2: Classification of Recommendation Techniques

Personalized recommendation is an enabling mechanism to overcome information overload occurred when
shopping in an internet marketplace, use personalized information for better recommendations to the user.

Non personalized recommendations are the simplest form of recommendations in which without any
consideration of user’s specifications some items are recommended. The most popular method is the recommendation
based on ranking of items. However, since they don’t take user’s preferences into account, the quality of their results are
low. For example in an electronic shop most sold items are recommended to all users [4].
A Hybrid Recommender System to Predict the Location Ranking 85

Content based recommendation systems analyze item descriptions to identify items that are of particular interest
to the user [8]. For instance, if a Netflix user has watched many cowboy movies, then recommend a movie classified in the
database as having the “cowboy” genre. Collaborative based recommendation systems recommend items based on
similarity measures between users and/or items. The items recommended to a user are those preferred by similar users.

Knowledge-based recommendation attempts to suggest objects based on inferences about a user’s needs and
preferences. Knowledge-based approaches are distinguished in that they have functional knowledge: they have knowledge
about how a particular item meets a particular user need, and can therefore reason about the relationship between a need
and a possible recommendation.

EXISTING WORK

In year 2004, Chien-Chih Yu presented a Framework on customer oriented intelligent decision support system. It
was a web based frame work to personalize the B2C web e-services. The work includes the personal management, auction,
negotiation, evaluation, planning, collaboration, transactions, payments, feedback and quality control. In this framework,
almost all the business oriented functionality is implemented and described for investment and the tourism applications.
The work is about to improve the efficiency and the effectiveness of the decision support system in the same area [1].
In 2010, an e-learning based decision support system was presented by Marta Zorrilla for E-Learning. The distance course
learning and teaching process is processed in this application. The work includes the course description, course assignment,
and other course based data mining operations are been handled by the system. The system includes the pattern based
model along with probability analysis for the decision making. The work will help the instructor to answer all the student
query and to find the outcome of standard process performed. The work also includes the reporting tool to present the work
in an effective way [6].

In year 2007, Mohammed N. A. Abdelhakim presented a work on intelligent decision making for the evaluation
and selection of educational multimedia. The work is a web based group decision system that will perform a statistical
analysis on education provider with continuous evaluation to investigate the requirement for developing the educational
application. The work also include the knowledge management along with design and implementation of performance
evaluation to present the collect and process the data from instructors, producers and propose a solution for the educational
consumers for the evaluation of the system[7].

In year 2008, Suresh Kalathur presented a work on the data mining operation on student driven content analysis
while online teaching. The work is presented in the form of a web model integrated with data mining operations to handle
the classroom discussions to predict the student faring. The model also provides the feedback on student discussion
regarding the topic discussion in class, and a comparative analysis with other topics. The analysis includes the text mining
operations relative to the answers submitted by the students [9].

Another mining based analysis on academic data is performed by J.M.Lauria in year 2012 for the analysis of
college student retention. Author has presented an analytical research on academic risk using data mining approaches. The
work is presented in the form of methodological framework to develop the query based model to analyze the course
management respective to academic records and the classification process is performed to work on selected dataset [2].

In Year 2008, Kleanthi Lakiotaki performed a work," UTA-Rec: A Recommender System based on Multiple
Criteria Analysis". UTARec, a Recommender System that incorporates Multiple Criteria Analysis methodologies is
presented. The system’s performance and capability of addressing certain shortfalls of existing Recommender Systems is
demonstrated in the case of movie recommendations. UTARec’s accuracy is measured in terms of Kendall’s tau and ROC
86 Akshita & Smita

curve analysis and is also compared to a Multiple Rating Collaborative Filtering (MRCF) approach [3]. In Year 2008, Juan
A. Recio-García performed a work," Prototyping Recommender Systems in jCOLIBRI". Presented goal is to support
system developers in rapid prototyping recommender systems using Case-Based Reasoning (CBR) techniques. In this
paper Author describe how jcolibri can serve to that goal. Jcolibri is an object-oriented framework in Java for building
CBR systems that greatly benefits from the reuse of previously developed CBR systems [10].

PROPOSED WORK

In this present work a hybrid architecture is defined to perform the recommendation about the location selection to
spend the vacations. In this system, we have defined a dataset with three main tables. First table contains the details related
to the user such as age, gender, occupation etc. The another table is defined to represent the locations. The locations are
defined in terms of distance, type of location, cost factor etc. The third table is the rank table that defines the ranking
allotted by different users to a particular user. The ranking is here assigned between 1 and 5. The presented system is
divided in three main layers.

First, the content based matching is performed to identify the most similar users from the dataset. Then matching
is performed under different attributes such as age, gender and occupation. The age and gender are taken with higher
priority factors whereas the occupation is having the least priority contribution.

Once the similar users are identified, Second the event based match is performed. Here the event is described as
the ranking assigned to a particular location. Each user assigned some ranking to each location. The analysis is here
performed on these assigned ranking. For this some ratio analysis mechanism is used in which the ranking assigned to
particular location by the similar users is identified and a ratio is generated. The another factor while performing this
analysis is the temporal factor. The temporal factor is here defined as the time based analysis. It means instead of analyzing
the whole dataset, a selective dataset is taken for the recommendation process. This dataset is the most recently
recommended locations by the users. Based on the temporal factor based model, the ratio analysis is obtained.

Now in last, these two vectors are combined to obtain the final result. Equal weightage is assigned to both kind of
analysis and finally the rank for a particular location by a particular user is identified. Here these aspects are described in
detail

Content Based Analysis

User Attribute Similarity is calculated by using Demographic information of users. The main idea behind making
predictions using demographic data is the assumption that people with similar characteristics enjoy similar Sites. It is
believed that age, gender, occupation and hometown play an important role on Site preferences and a set of users who have
a high level of demographic similarity with the target user is found. Then the similarity is used as initial value for user-
based similarity calculation.

The similarity between two users can be calculated as:

(1)

Where f represents a feature of the user from the set of all demographic features F,

w represents the relative weight of feature f,

u1f and u2f represent the values of f for u1 and u2

A Hybrid Recommender System to Predict the Location Ranking 87

Rank Similarity

Rating similarity is calculated by using Pearson Correlation Coefficient. Pearson's correlation coefficient is a
measure of the strength of the association between the two variables.

Let U be the set of users, P be the set of items and data as set of triplet (i, x, r), where i ϵ U is a user, x ϵ P is an
item and r is a rating of item x by user i. Moreover, r ix denotes the rating of item x by user i. P i is subset of P that denotes
the set of item rated by i. Ux denotes set of users that have rated item x. User Rating similarity is given by UB-PCC as

(2)

Where is the average rating user i give to all items and is the average rating user j give to all items, rix is the
rating given by user i to item x and rjx is the rating of j to item x.

Hybrid Approach

On the basis of calculated similarities, predictions are generated. This can be done using class formation. As after
determining the similarities, various classes can be formed on the basis of similarity values. The rating for an unseen item
or a new user can be determined.

The basic flow of the presented work is shown in figure 3.

Figure 3: Flow Chart of Proposed Work

RESULTS

The presented work is implemented on an authenticated dataset in matlab environment. In this work, we have
taken three tables called user table, location table and ranking table. At first the similarity based match is performed on
user table to identify the similar users to process. In second layer, the rank similarity is performed on all three tables
collectively and derives the ratio analysis. Finally these two are collected to obtain the rank for the particular location. To
present the results, the dataset is divided in two sets called training set and testing set under 10 fold methods. According to
this, 90% record dataset is taken as the training dataset and 10% is taken as the testing dataset. Now the analysis is
performed on this testing dataset and the ranking is predicted. These predicted values are compared with existing ranked
88 Akshita & Smita

values. The difference between these two values is taken as the error. The analysis of this work is done based on this error
analysis. The Results obtained from this error analysis is given as under.

Table 1: Error Analysis

Method Mean Absolute Error
10 elements dataset 0.131
20 elements dataset 0.139
30 elements dataset 0.127

As we can see in table 1, as the numbers of testing records are increased the accuracy of obtained results is also
increased. The system is providing about 87% accuracy level.

Below graphs shows MAE calculation fir different size testing datasets.

Figure 4: MAE Analysis Using Proposed Collaborative (Hybrid) Filtration

CONCLUSIONS

In this present work a hybrid recommendation system is presented to predict the user interest location based on
similarity match. In this work, content based and rank based similarity measures are obtained and merged to obtain the
collective results from the system. The obtained results show the effectiveness of the system in terms of higher degree of
accuracy.

REFERENCES

1. Chien-Chih Yu," A Web-Based Consumer-Oriented Intelligent Decision Support System for Personalized E-
Services".2004 Proceeding of the 6th ICEC 1-58113-930 (PP 229-237).

2. Eitel J.M. Lauría, Joshua D. Baron" Mining academic data to improve college student retention: An open source
perspective", LAK’12,Proceeding of 978-1-4503-1111-3/12/04

3. Kleanthi Lakiotaki," UTA-Rec: A Recommender System based on Multiple Criteria Analysis", RecSys’08,
October 23–25, 2008, Lausanne, Switzerland. ACM 978-1-60558-093-7/08/10 (pp 219-225)

4. Kyumars Sheykh Esmaili, Mahmood Neshati, Mohsen Jamali, Hassan Abolhassani and Jafar Habibi, “Comparing
Performance of Recommendation Techniques in the Blogsphere” ,in ECAI 2006 workshop on recommender
system.
A Hybrid Recommender System to Predict the Location Ranking 89

5. Lukasz Kurgan and Petr Musilek (2006); A survey of Knowledge Discovery and Data Mining process models.
The Knowledge Engineering Review. Volume 21 Issue 1, March 2006, pp 1–24, Cambridge University Press,
New York, NY, USA doi: 10.1017/S0269888906000737.

6. Marta Zorrilla," A Decision Support System to improve e-Learning Environments", 978-1-60558-990-9/10/03

7. Mohammed N. A. Abdelhakim," A Web-Based Group Decision Support System for the Selection and Evaluation
of Educational Multimedia", Emme ’07 Proceedings of the international workshop on Educational multimedia and
multimedia education, 978-1-59593-783-4/07/0009.

8. Michael J. Pazzani and Daniel Billsus, “Content-Based Recommendation Systems”, the Adaptive Web, LNCS
4321, pp. 325 – 341, 2007.

9. Suresh Kalathur," Enriching Student Experience with Student Driven Content while Teaching an Online Data
Mining Class", SIGITE’08 Proceedings of the 9th ACM SIGITE conference on Information technology
education, 978-1-60558-329-7/08/10, pp 125-130.

10. Juan A. Recio-García, Belen Diaz Agudo "Prototyping Recommender Systems in jCOLIBRI", RecSys’08,
Proceedings of the 2008 ACM conference on recommender system, October 23–25, 2008, Lausanne, Switzerland.
ACM 978-1-60558-093-7/08/10 (pp243-250).