MobiContext: A Context-aware Cloud-Based Venue Recommendation Framework

          

 MobiContext: A Context-aware Cloud-Based Venue Recommendation Framework

ABSTRACT

 In recent years, recommendation systems have seen significant evolution in the field of knowledge engineering. Most of the existing recommendation systems based their models on collaborative filtering approaches that make them simple to implement. However, performance of most of the existing collaborative filtering-based recommendation system suffers due to the challenges, such as: (a) cold start, (b) data sparseness, and (c) scalability. Moreover, recommendation problem is often characterized by the presence of many conflicting objectives or decision variables, such as users’ preferences and venue closeness. In this paper, we proposed MobiContext, a hybrid cloud-based Bi-Objective Recommendation Framework (BORF) for mobile social networks. The MobiContext utilizes multi-objective optimization techniques to generate personalized recommendations. To address the issues pertaining to cold start and data sparseness, the BORF performs data pre-processing by using the Hub-Average (HA) inference model. Moreover, the Weighted Sum Approach (WSA) is implemented for scalar optimization and an evolutionary algorithm (NSGA-II) is applied for vector optimization to provide optimal suggestions to the users about a venue. The results of comprehensive experiments on a large-scale real dataset confirm the accuracy of the proposed recommendation framework.

EXISTING SYSTEM:

 In recent years, recommendation systems have seen significant evolution in the field of knowledge engineering. Most of the existing recommendation systems based their models on collaborative filtering approaches that make them simple to implement. However, performance of most of the existing collaborative filtering-based recommendation system suffers due to the challenges, such as: (a) cold start, (b) data sparseness, and (c) scalability. Moreover, recommendation problem is often characterized by the presence of many conflicting objectives or decision variables, such as users’ preferences and venue closeness.

Disadvantage:

1 .Cold start

 The cold start problem occurs when a recommendation system has to suggest venues to the user that is newer to the system. Insufficient check-ins for the new userresultsin zero similarity value that degrades the performance of the recommendation system. The only way for the system to provide recommendation in such scenario is to wait for sufficient check-ins by the user at different venues.

2. Data sparseness

Many existing recommendation systems suffer from data sparseness problem that occurs when users have visited only a limited number of venues. This results into as parsley filled user-to-venue check-in matrix. The sparseness of such matrix creates difficulty in finding sufficient reliable similar users to generate good quality recommendation.

PROPOSED SYSTEM:

We propose a cloud-based framework consisting of bi-objective optimization methods named as CF-BORF and greedy-BORF. The Genetic Algorithm based BORF (GA-BORF) utilizes Non-dominated Sorting Genetic Algorithm (NSGA-II) to optimize the venue recommendation problem. We introduce a pre-processing phase that performs data refinement using HA. We perform extensive experiments on our internal Open Nebula cloud setup running on 96 core Super micro Super Server SYS-7047GR-TRF systems. The experiments were conducted on real-world “Gowalla” dataset.

Advantage:

Ø  Most of the existing recommendation systems utilize centralized architectures that are not scalable enough to process large volume of geographically distributed data. The centralized architecture for venue recommendations must simultaneously consider users’ preferences, check-in history, and social context to generate optimal venue recommendations. Therefore, to address the scalability issue, we introduce the decentralized cloud-based MobiContext BORF approach.

Ø Memory Efficiency.

FEATURES:

In the future, we would like to extend our work by incorporating more contextual information in the form of objective functions, such as the check-in time, users’ profiles, and interests, in our proposed framework. Moreover, we intend to integrate other approaches, such as machine learning, text mining, and artificial neural networks to refine our existing framework.

PROCESS:

MODULE DESCRIPTION:

Number of Modules

After careful analysis the system has been identified to have the following modules:

1.     User Profiles

2.      Ranking Module

3.     Mapping Module

4.     Recommendation Module

1.      User Profiles

The MobiContextframework maintains records of users’ profiles for each geographical region. A user’s profile consists of the user’s identification, venues visited by the user, and check-in time at a venue.

2.      Ranking Module

On top of users’ profiles, the ranking module performs functionality during the pre-processing phase of data refinement. The pre-processing can be performed in the form of periodic batch jobs running at monthly or weekly basis as configured by system administrator. The ranking module applies model-based HA inference method on users’ profiles to assign ranking to the set of users and venues based on mutual reinforcement relationship. The idea is to extract a set of popular venues and expert users. We call a venue as popular, if it is visited by many expert users and a user as expert if she has visited many popular venues. The users and venues that have very low scores are pruned from the dataset during offline pre-processing phase to reduce the online computation time.

3.     Mapping Module

The mapping module computes similarity graphs among expert users for a given region during pre-processing phase. The purpose of similarity graph computation is to generate a network of like-minded people who share the similar preferences for various venues they visit in a geographical region. The mapping module also computes venue closeness based on geographical distance between the current user and popular venues.

4.      Recommendation Module

 The online recommendation module that runs a service to receive recommendation queries from users. A user’s request consists of: (a) current context (such as, GPS location of user, time, and region), and (b) a bounded region surrounding the user from where the top N venues will be selected for the current user (N is number of venues).The recommendation service passes the user’s query to optimization module that utilizes scalar and vector optimization techniques to generate an optimal set of venues. In our proposed framework, the scalar optimization technique utilizes the CF-based approach and greedy heuristics to generate user preferred recommendations. The vector optimization technique, namely GA-BORF, utilizes evolutionary algorithms, such as NSGA-II to produce optimized recommendations.

SOFTWARE REQUIREMENTS:

          Operating System            : Windows

          Technology                    : Java and J2EE

          Web Technologies                   : Html, JavaScript, CSS

           IDE                               : My Eclipse

           Web Server                            : Tomcat

           Network                       : LAN

           Database                      : My SQL

           Java Version                : J2SDK1.5                 

                

HARDWARE REQUIREMENTS:

         Hardware                            :     Pentium

         Speed                                  :     1.1 GHz

         RAM                                   :    1GB

         Hard Disk                            :    20 GB

         Floppy Drive                        :    1.44 MB

         Key Board                            :    Standard Windows Keyboard

         Mouse                                  :    Two or Three Button Mouse

         Monitor                                :    SVGA

CONCLUSION

We proposed a cloud-based framework MobiContextt hat produces optimized recommendations by simultaneously considering the trade-offs among real-world physical factors, such as person’s geographical location and location closeness. The significance and novelty of the proposed framework is the adaptation of collaborative filtering and bi-objective optimization approaches, such as scalar and vector. In our proposed approach, data sparseness issue is addressed by integrating the user-to-user similarity computation with confidence measure that quantifies the amount of similar interest indicated by the two users in the venues commonly visited by both of them. Moreover, a solution to cold start issue is discussed by introducing the HA inference model that assigns ranking to the users and has a precompiled set of popular unvisited venues that can be recommended to the new user.