Multi to binary class size based imbalance handling technique in wireless sensor networks
by Neha Singh; Deepali Virmani; Gaurav Dhiman; S. Vimal
International Journal of Nanotechnology (IJNT), Vol. 20, No. 5/6/7/8/9/10, 2023

Abstract: Wireless sensor network (WSN) has entered various disciplines including healthcare, banking, transportation, ocean and wildlife monitoring, earthquake monitoring, and numerous military applications. Now days, there is escalation in size of data which makes it unfeasible to analyse it with accuracy. There are numerous problems that are faced when detecting a pattern between structured and unstructured data which are unworkable by humans, so to make computation fast, easy and accurate, Machine Learning came in existence. Machine Learning is extensively used in WSNs. To make a machine learn, a training dataset is required and output is predicted by testing the dataset. A dataset in WSN has multi-class in its dependent variable. This multiclass classification causes class imbalance problem. The paper proposes multi to binary class size based imbalance handling technique in wireless sensor networks (MBSCIH) technique to solve class imbalance problem in multi-class classification. The technique MBSCIH converts multi-class classification into binary-class classification. The proposed technique MBSCIH is applied on WSNDS, NSL-KDD and KDD-cup 99 datasets and is tested with five major machine learning algorithm: Naive Bayes, Random Forest, decision tree, support vector machine (SVM) and k-nearest neighbour (KNN). The test method used for testing is 10-fold cross validation. Results infer that the proposed method increases the existing efficiency by 15.13%, 0.28%, 0.01%, 0.01%, 0.12% for Naïve Bayes, Decision Tree, SVM, Random Forest and KNN Algorithms respectively for KDD Cup99 dataset. The proposed method increases the efficiency by 43.1%, 2.79%, 0.01%, 0.53% for Naïve Bayes, decision tree, SVM and KNN algorithms respectively for NSL-KDD dataset. Also, the proposed method increases the efficiency by 0.13%, 0.08%, 0.29%, 0.29% for Naïve Bayes, decision tree, SVM and KNN algorithms respectively for WSN-DS dataset.Further the experimental analysis proves that solving multiclass problem through algorithm has enhanced the detection of intrusions in WSNs.

Online publication date: Tue, 10-Oct-2023

The full text of this article is only available to individual subscribers or to users at subscribing institutions.

 
Existing subscribers:
Go to Inderscience Online Journals to access the Full Text of this article.

Pay per view:
If you are not a subscriber and you just want to read the full contents of this article, buy online access here.

Complimentary Subscribers, Editors or Members of the Editorial Board of the International Journal of Nanotechnology (IJNT):
Login with your Inderscience username and password:

Username:        Password:          Forgotten your password?

Want to subscribe?
A subscription gives you complete access to all articles in the current issue, as well as to all articles in the previous three years (where applicable). See our Orders page to subscribe.

If you still need assistance, please email subs@inderscience.com