Authors of this paper belong to Computer Science and Information Technology Department of University of Azad Jammu and Kashmir, Muzaffarabad, Pakistan Author email addresses: Adil Aslam Mir: adilmir300@gmail.com Rabia Riaz: rabiaiqbal18@gmail.com Sobia Kazmi: srdc2012@gmail.com Sajjad Ahmed Nadeem: msajjadnadeem@yahoo.com Sharjil Saeed: rajasharjeel@gmail.com Corresponding Author email address : rabiaiqbal18@gmail.com Comparative Analysis of Boosting with other Machine Learning Methods for classifying Labeled Wireless Sensor Network Data Adil Aslam Mir, Rabia Riaz, Sobia Kazmi, Sajjad Ahmed Nadeem, Sharjil Saeed. Abstract- Wireless sensor networks (WSNs) are deployed in order to monitor the physical environment with the help of very low-priced entrenched devices. With the advances in technology, the nodes in a wireless sensor network are getting smaller and produces large amount of data. To cope with this huge amount of data, machine learning and data mining offers different standardized techniques and methods. The most important research problem regarding WSNs is the classification of sensory data for reducing the amount of data transmission. Various data mining and machine learning techniques are proposed and experimented for analyzing WSNs such as Decision Tree, Naive Bayes, and ZeroR. No comparative study among these and others methods (SVM, LDA, Boosting for example) has been made yet. This study aimed at investigating boosting (DTs as a base classifier) with other different machine learning methods such as Linear Discriminate Analysis (LDA), Support Vector Machines (SVMs), Decision Trees (DTs) for classification of WSNs data. The statistics below showed that when Decision Trees and boosting are used, there is a significant improvement in the accuracy for classification because an ensemble of classification rules is made and the final classification is based on the aggregation of these ensemble results. For analysis, we have used publically available single and multi-hop Labeled Wireless Sensor Network data containing different humidity and temperature readings. Index Terms— Boosting, Classification, Decision Trees, SVM, Wireless Sensor Networks 1. INTRODUCTION or the purpose of sensing, transmitting and processing of data from the environment, the major advances in the computer technology provided the ways to the fabrication of wireless enabled sensor nodes. These sensing devices are powered by battery and also have memory, processing unit, operating system, actuator, GPS sensors, cameras, sensor for light, audio and direction detection. The wireless sensor network is used to keep an eye on surroundings, habitat, real life applications, ambiance, traffic control and formation of a building [1]. Because of their low cost and benefits, the wireless sensor networks are getting larger and expectedly turn out to be common. In future these sensor networks could contain millions of sensor’s data [2]. For process monitoring in wireless sensor network, classification is mostly used in order to decrease the use of energy and traffic in network. The classification helps in lengthening the lifetime of sensor network [2-3]. Moreover, WSNs are installed in an unsupervised and unlocked environment where the physical communiqué is not probable and also generate a huge amount of data. The situation may happen when the abilities for the storage and collection of generated data are not in the hand of someone’s capacity to investigate, sum up and excavate knowledge from these data. So, the situation discussed above can be tackled by the use of different data mining and machine learning methods. The space complexity can be reduced in a way that each node in a wireless sensor network stores only important data, mainly the faulty data in order to send it to the fusion center. By using various data mining approaches, the information acquired can be helpful in reduction of energy consumption, network traffic and enhance the lifetime of sensor networks. Various data mining and machine learning techniques has been successfully applied in order to classify WSNs data such as J48 (Decision Tree), Naive Bayes, ZeroR [4], decision- tree-based hierarchical distributed classification [1], Neural Networks and ARIMA [5], K-Nearest Neighbours [6], SVM [7] etc. In this study we have investigated boosting (DTs as base classifier) with different machine learning methods such as Linear Discriminate Analysis (LDA), Support Vector Machines (SVMs), Decision Trees (DTs) for classifying sensor provided data and to investigate the accuracy of classification of different machine learning techniques in order to decrease the misclassification rate. F International Journal of Computer Science and Information Security (IJCSIS), Vol. 14, No. 9, September 2016 917 https://sites.google.com/site/ijcsis/ ISSN 1947-5500