Development of Big Data App for Classification based on Map Reduce of Naive Bayes with or without Web and Mobile Interface by RESTful API Using Hadoop and Spark

Abstract

Big Data App is a developed framework that we made based on our previous project research and we have uploaded it on github, which is developing lightweight serverless both on Windows and Linux OS with the term of EdUBig as Open Source Hadoop Distribution. In this study, the focus is on solving problems related to difficulties in building a frontend and backend model of a Big Data application which by default only runs scripts through consoles in the terminal. This will be quite a tribulation for the end users when the Big Data application has been released and mass produced to general users (end users) and at the same time how the end users test the performance of the Map Reduce Naive Bayes algorithm used in several datasets. In accordance to these problems, we created the Big Data App framework to make the end users, especially developers, feel easier to build a Big Data application by integrating the frontend using the Web App from Django framework and Mobile App Native, while for the backend, we use Django framework that is able to communicate directly with the script either hadoop batch, streaming processing or spark streaming very easily and also to use the script for pig, hive, web hdfs, sqoop, oozie, etc. the making of which is extremely fast with reliable results. Based on the test results, a very significant result in the ease of data computation processing by the end users and the final results showing the highest classification accuracy of 88.3576% was obtained. Keywords: big data, map reduce of naive bayes, serverless, web and mobile app, restful api, django framework