International Journal of Futuristic Machine Intelligence & Application (IJFMIA) Vol. 1 Issue 2,ISSN 2395-308x 2015 ©IJFMIA 1 www.ijfmia.in A Survey on Energy Efficient Data Prediction Techniques in Wireless Sensor Network ThakerMaulik B. 1 ,Prof. Uma Nagaraj 2 , Prof. Pramod D. Ganjewar 3 M.E. Student,Dept. of Computer Engineering, MIT Academy of Engineering,Pune,India 1 Prof. and Head of Dept. of Computer Engineering, MIT Academy of Engineering,Pune,India 2 Assistant Prof. of Dept. of Computer Engineering, MIT Academy of Engineering,Pune,India 3 Abstract: Wireless sensor network is a network which contains sensing and routing via nodes mainly called sensorsand it transmits the data to sink. WSN contains star and mesh topologies or single and multi-hop network. WSN used for monitor physical and environment condition and deals with many protocols according of various applications. In this paper, we mainly describe various data prediction techniques toremove redundant data and increase the energy of sensor nodes. Data prediction is one of the data pre-processing techniques use to predict data at sensor or sink node and aims to remove unnecessary data while transmission occur from source node to base station. We describe here recent techniques based on data prediction in this survey which helps to increase energy as well as life of network. Index Terms –Data Prediction, Data redundancy, Energy efficiency, Network lifetime,WSN. I. INTRODUCTION Wireless sensor networks (WSN) consist of a large number of low-cost, spatially distributed, varieties of sensors in environment for monitoring or sensing purpose. The sensed data by each sensor node is transmitted to the sink using single-hop or multi-hop scheme. Each sensor node use the power supply to sensed and transmit the data and energy of each node is limited hence it will decrease when more and more transmission occur by each node. Wireless sensor networkscan be deployed in a geographical area for monitoring temperature, pressure ,humidity [1] etc. WSN are also use the applications such as environmental monitoring [2], agriculture, medical and process control [3].It requires sensor nodes to collect data and transmit to the sink at a specific rate. The data generated by sensornodes usually contain a large portion of redundant data and transmission of data may cause unnecessary energy consumption. Data Reduction techniques are helpful to remove data redundancy in WSN. It divides in to several parts like data compression, data prediction, In-network processing [4]and data aggregation. Data compression is used to reduce the amount of data sent by sensor nodes. This scheme involves encoding strategy at sensor node and decoding strategy at sink if the application doesn’t require the measurements. In- network processing is useful while data is routed towards the sink node and use various data aggregation schemes. The goal is to transmit the sensor data into less voluminous refined data using the functions like minimum, maximum, average, entropy etc. But some application require accurate measurements, in that cases it can degrade accuracy [5]. Data prediction is a technique in which values of data are predicted at sensor for transmitting or at sink for receiving and sometimes also called dual prediction in wireless sensor network. Generally, data prediction is depending on physical phenomenabeing monitored, prediction models and algorithms. Data prediction is an efficient approach to reduce the number of redundant data in sensor network [6].Dataprediction uses history data sample bysensors to predict the future data, and not needto transmit the newly sampled data to the sink. If the difference betweensampled data and its prediction value iswithin some given error then only transmission is required. It mainly divided into time- series approach, stochastic approach and algorithmic approach. Stochastic approach is used when sensed data value is considered as a random process by probabilitydensity function.Moving Average (MA), Least Mean Square (LMS), Auto-Regressive (AR) or Auto Regressive Moving Average (ARMA) models are based on time-series approach. Here, in this survey, we mention various prediction approaches in WSN like in-network prediction, cluster based approach, linear prediction, multiple linear prediction and dual prediction models in order to increase energy of WSN by removing data redundancy. II. DATA PREDICTION TECHNIQUES 1. In-network prediction: In this approach data are collected from sensor nodes and while transmitting, the values are predicted that called in- network prediction. It uses regression models. Data collected by sensor nodes can be considered as temporalseries, where the time is instant when a givenvalue is collected, sensed data