Analysis Of Learning Objects Using Hadoop

Published Date: 02 Nov 2017

INTRODUCTION

With recent advances in Cloud Computing , it can be used for many data- intensive applications . Data intensive applications are that data is more voluminous and more sophisticated mechanisms are required for the analysis purpose. These data in Cloud cannot be stored or analysed using Relational Database. Many real world data intensive applications are Weather Forecast , Forensic data , Health care records and many more. An interesting fact is that the above said data sets are all unstructured .

For the analysis purpose of the large data sets some technique is needed . Hadoop is thus a frame work used for the analysis and storage of large data set. The large data set available in internet is termed as Big Data. Hadoop employs Map Reduce Technique for parallelizing propose. The data is stored in HDFS (Hadoop Distributed File System).

In this paper , hadoop framework is considered in the e-learning domain. The data set taken here is Learning Objects (Los) . Learning Objects are nothing but are study materials used in e-learning. The materials includes documents , power point presentations , audios ,videos , quizzes , assessments , questions , answers , grading etc. Learning Objects are referenced using metadata which contains all search words or tags towards the learning objects. Here in this paper we propose how learning objects can be implemented using map reduce method in hadoop .

RELATED WORKS

In [1] , Hongyong et al , massive data management in health care was presented using Map Reduce and Hive methods in Hadoop and considered the performance factors. The paper [2] presents how the map reduce and hadoop works and also develops a simple GUI for ordinary users for using hadoop without real programming. Gao Xin et al in [3] suggests an efficient method of data storage in a campus using hadoop framework. The paper by Ashok Chandrasekar et al , they propose a simple technique of classifying metadata using importance factor. Shamil in [4] generalises the data intensive applications using map reduce and also describes about Google File Systems(GFS) and Hadoop Distributed File Systems(HDFS) .

EXISTING SYSTEM

Learning Objects are small units that can be fitted in any number of ways to provide customized learning experience(Hodgins and Conner 2000). Either these small chunks can be separately used or can be combined together for learning purpose. These learning objects are reusable and self contained . The learning objects are maintained by Learning Management System(LMS) . For using Learning Objects , simple username and password verification procedure is required . Then particular keyword to be searched will be given , the metadata finds the corresponding locations of the learning objects related to the keyword will be returned. Day by day informations included in the learning objects are increased , thus data storage and performance factors are affected .

PROPOSED SYSTEM

The proposed system has been designed to overcome the data storage and performance bottlenecks in the learning management system. Here a new technology Hadoop , is used for effective retrieval of learning objects by searching the keywords parallelly and producing desired output. The processing time is also reduced by using hadoop . The system involves user authentication then selection of the learning style to be used and after that the map reduce technique for learning objects .

DESIGN AND IMPLEMENTATION

The proposed system involves the following stages :

1) User Authentication.

2) Selection of Learning Style.

3) Keyword searching using Metadata.

4) Analysis of Learning Objects using Hadoop.

5) Result to User.

Each stage is explained in detail in the below.

1) User Authentication.

To use a particular learning system , user need to register using his/her login and password . Only registered users can use the learning objects provided by the Learning Management System. Registration is just a simpe process which takes just few minutes.

2) Selection of Learning Style.

Learning Styles are the way by which a user prefers to learn. Each person has his/her way of learning . Some may prefers texts , some videos , some audios , some ppts etc. Some learners may be fast in learning some may not be . Thus a selection of learning style is needed for effective usage of learning objects. Several Learning Style models are available today based on the user’s personality , attitude and behaviour . In this paper , Felder Silverman Model of learning style is used. The model consists of a questionnaire . This questionnaire consists of 44 questions based on which user’s style of learning is selected . The model consists of 4 groups :

1) Perception : Sensing vs Intuitive.

Sensory learning is based on sights , sounds ,physical sensations etc. These kind of learners believe on learning by facts.

Intuitive learning uses the possibilities and insights available. Learners are interested in learning by understanding meaning.

2) Input : Visual vs Verbal.

Visual learning is by using pictures , diagrams , films , timelines etc.

Verbal learning is based on what they hear or on what they read.

3) Processing : Active vs Reflective.

Active group people like to study in groups and Reflective learners like to have self study.

4) Understanding : Sequential vs Global .

In sequential style , learner follows a linear order of learning of the materials provided thus finally a big picture of the topic is obtained.

Global understanding first makes a general undeerstanding of the materials provided to get a general approach of the topic covered and then starts learning based on his interest.

Out of the 44 questions , each of the above 4 categories have 11 questions each . The learner will be asked to fill the questionnaire which determines the category to which he/she belongs and corresponding materials will be provided.

3) Keyword Search using Metadata.

After learning style is determined , necessary learning objects have to be found for the learner . For this purpose the concept of metadata is used. Metadata is data about the data . Metadata contains tags or index to the location where the data is originally stored. Thus for searching purpose we usually give keywords and corresponding locations will be displayed.

4) Analysis of Learning Objects using Hadoop.

In this we are explaining how map reduce and hadoop framework is used in the e-learning. Hadoop is a framework used for parallel execution of a large data set thus providing better and effective data storage mechanism , performance , scalability and fault tolerance .The hadoop uses a technology called Map Reduce and HDFS for data storage. The data set we considering are learning objects . These may be available as different data blocks . The original data set available in blocks are then stored in HDFS and the database used in hadoop is referred as HBase. HDFS follows master-slave architecture. There is a master node called Name Node /Job Tracker and several slave nodes called data nodes / task trackers. These nodes actually performs map reduce function. Map Reduce has 2 functions : Map function and Reduce function . The data set is first split into different blocks . These blocks are fed as input into map function. The name node allocates the splitted data blocks into different slave nodes. Each slave node separately performs mapping function on the data block using key/value pairs . From each data node intermediate results will be obtained with key/value pairs. These intermediate results are given to Reduce function. The intermediate results are combined using reduce functions and the output is given to user.

5) Result to User .

Result obtained using map-reduce functions will be given to user.