Data mining means to find out some useful information from a big warehouse of data and the process is aimed at unfolding old records and identifying novel patterns from the data. Data mining is used for classification and prediction. Many techniques and algorithms are available for mining the data. Out of many techniques, the decision tree is the simplest. This paper focuses on comparing the performance accuracy of ID3 and C4.5 techniques of the decision tree for predicting customer churn using WEKA. The data used for this research work has been collected by designing a survey form and getting it filled by around 150 mobile phone users belonging to a different gender, age groups and having different types of connection providers. For the data analysis in WEKA, the cross-validation method is used where a number of folds n (10 as standard as per the software) is used. From the results, it is observed that C4.5 algorithm exhibits better performance than ID3.

ABSTRACT

With the advent of new technologies like software defined networking, cloud computing and Internet of Things, everything needs to be redefined. Software Define Networking (SDN) is the latest approach and an emerging network technology that will bring a major change in the area of networking. Though SDN has been successfully applied to most of the networking area but traffic classification is the area where it is yet to be applied. With the high adoption of cloud services, the traffic on cloud increased rapidly. The technologies need to be clubbed together so that they can survive in the rapidly changing environment. The paper aims at addressing the cloud traffic classification using Differential Services Code Point (DSCP) marking in software defined network environment. This allows us to identify cloud traffic separately from other web services and helps its traffic flows to be provided with special treatment over other internet services. The paper aims to classify cloud traffic along with suggesting some marking schemes to prioritize cloud traffic using DSCP of IP header.

No abstract

No abstract