Intelligent agriculture, digital health, or smart cities are only a few out of multiple uses of the Internet of Things. The limited energy supply of the IoT nodes, specifically battery-run sensor nodes, prevents them from maintaining consistent work and hinders the network’s functioning. In this regard, a smart framework that utilizes the progressive machine learning models with multi-criteria decision-making should ensure higher energy efficiency in the IoT networks. While the existing researches have attempted to decrease energy levels in the IoT networks, most of them apply primitive concepts: clustering, routing, and node sleep, and do not use the most efficient machine learning algorithms for energy prediction. Indeed, several people have tried using machine learning algorithms, like decision trees, linear regression, and elementary ANN for energy prediction. However, most of these algorithms are efficient if they considered as individual ones, and people almost never combine energy prediction and node priority. As a result, we propose a complex system of several designed operations that result in increased energy efficiency. First, the data on energy consumption are gathered at regular intervals and preprocessed: normalized, denoised, and empty value-imputed. Then the LSTM model is used to find temporal patterns and predict the future changes. After node ranking, various dynamic strategies like routing, some of the nodes put to sleep, and traffic are optimized. As a result, the lifetime of the network increases by 35 % whereas the energy consumption decreases by 23 %.
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号