Energy Consumption model for performance Analysis in WSN using Machine Learning Techniques

Abstract

Wireless Sensor Network (WSN) is one of the most promising technologies for real-time applications because of its size, cost-effective and easily deployable nature. The job of WSN is to monitor a field of interest and gather certain information and transmit them to the base station for post data analysis. Some of the WSN applications consists of a large number of sensor nodes. Therefore managing such a large number of nodes requires a scalable and efficient algorithms. In addition, due to the external causes or intended by the system designers, the WSNs may change dynamically. Therefore it may affect network routing strategies, localization, delay, cross-layer design, coverage, QoS, link quality, fault detection, etc. Because of the highly dynamic nature, it may require depreciating dispensable redesign of the network, but the traditional approaches for the WSNs are explicitly programmed, and as a result, the network does not work properly for the dynamic environment. newlineMachine Learning is the process that automatically improves or learns from the study or experience, and acts without being explicitly programmed. Machine learning is an efficient way for automatic computer systems building which is enhanced through experience as well as fundamental laws governing entire learning processes. The learning is done by historical data patterns. Machine learning solution has spread in wide areas such as sales, search Engines, Transportation, and now it is promising in auto- mobile industry. Improved predictions can be done by accumulating multiple learners by models ensemble which in turn accomplished by capturing data underlying distribution in precise manner. Various classification algorithms are available in the existing systems and the output variable can be either nominal or numeric. Therefore, the datasets loose the inbuilt order amid class values. This problem is mitigated by proposing a hybrid ensemble classification and regression learning algorithm (HECL) methodology which includes bagging besides adaboost algorithms as a key to solve nominal and numeric problem in auto- motive industry. We implemented different ensemble methods of machine learning such as bagging boosting and stacking. These methods finds the best tradeoff with bias and variance. In this study a systematic approach hybrid ensemble classification and learning algorithm is proposed .Furthermore, this is a nested based optimisation algorithm which tunes hyperparameters and finds the optimal weights to combine the ensembles. In addi- tion HECL algorithm is designed for speeding up newlineii newlineoptimisation process. The algorithm is run on real time auto mpg datasets. I am grateful to the university of Irvine and kaggle for providing the dataset. The prediction accuracy of HECL algorithm is compared with base learners of the state-of-art approaches (Random Forest, KNN, XGBoost, Decision Tree). The suggested HECL approach yields improved classification performance than traditional approach which is substantiated by the results. In this research work we built a model that can predict mileage of the car and also identifies the possibility of accident. newlineThe model which we build is trained with new car dataset and hence it is used to predict future mileage for all the upcoming cars, This allows companies to resource on R andD and make more efficient, and make more popular vehicles to outshine the competitors. In today s world, there are many inventions in the automobile industry, the major invention in automobile sector is towards to design and build a module for safety measures in automobiles. As we are aware that traffic accidents are unavoidable. Most of the accidents occur in rural and urban areas. Based on different accidental scenarios different patterns are generated under different situations and accurate prediction models might be utilized for detection which has the ability for separating several accident circumstances. K-Means clustering algorithm is been used to prevent accidents and develop safety measures in predicting the accident possibility. We try to attain maximum likelihoods of accident reduction through low budget resources and utilizing few scientific measures. newline