Wireless sensor networks are susceptible to failures of nodes and links due to various physical or computational reasons.Some physical reasons include a very high temperature,a heavy load over a node,and heavy rain.Co...Wireless sensor networks are susceptible to failures of nodes and links due to various physical or computational reasons.Some physical reasons include a very high temperature,a heavy load over a node,and heavy rain.Computational reasons could be a third-party intrusive attack,communication conflicts,or congestion.Automated fault diagnosis has been a well-studied problem in the research community.In this paper,we present an automated fault diagnosis model that can diagnose multiple types of faults in the category of hard faults and soft faults.Our proposed model implements a feed-forward neural network trained with a hybrid metaheuristic algorithm that combines the principles of exploration and exploitation of the search space.The proposed methodology consists of different phases,such as a clustering phase,a fault detection and classification phase,and a decision and diagnosis phase.The implemented methodology can diagnose composite faults,such as hard permanent,soft permanent,intermittent,and transient faults for sensor nodes as well as for links.The proposed implementation can also classify different types of faulty behavior for both sensor nodes and links in the network.We present the obtained theoretical results and computational complexity of the implemented model for this particular study on automated fault diagnosis.The performance of the model is evaluated using simulations and experiments conducted using indoor and outdoor testbeds.展开更多
The problem of automatic and accurate forecasting of time-series data has always been an interesting challenge for the machine learning and forecasting community.A majority of the real-world time-series problems have ...The problem of automatic and accurate forecasting of time-series data has always been an interesting challenge for the machine learning and forecasting community.A majority of the real-world time-series problems have non-stationary characteristics that make the understanding of trend and seasonality difficult.The applicability of the popular deep neural networks(DNNs)as function approximators for non-stationary TSF is studied.The following DNN models are evaluated:Multi-layer Perceptron(MLP),Convolutional Neural Network(CNN),and RNN with Long Short-Term Memory(LSTM-RNN)and RNN with Gated-Recurrent Unit(GRU-RNN).These DNN methods have been evaluated over 10 popular Indian financial stocks data.Further,the performance evaluation of these DNNs has been carried out in multiple independent runs for two settings of forecasting:(1)single-step forecasting,and(2)multi-step forecasting.These DNN methods show convincing performance for single-step forecasting(one-day ahead forecast).For the multi-step forecasting(multiple days ahead forecast),the methods for different forecast periods are evaluated.The performance of these methods demonstrates that long forecast periods have an adverse effect on performance.展开更多
文摘Wireless sensor networks are susceptible to failures of nodes and links due to various physical or computational reasons.Some physical reasons include a very high temperature,a heavy load over a node,and heavy rain.Computational reasons could be a third-party intrusive attack,communication conflicts,or congestion.Automated fault diagnosis has been a well-studied problem in the research community.In this paper,we present an automated fault diagnosis model that can diagnose multiple types of faults in the category of hard faults and soft faults.Our proposed model implements a feed-forward neural network trained with a hybrid metaheuristic algorithm that combines the principles of exploration and exploitation of the search space.The proposed methodology consists of different phases,such as a clustering phase,a fault detection and classification phase,and a decision and diagnosis phase.The implemented methodology can diagnose composite faults,such as hard permanent,soft permanent,intermittent,and transient faults for sensor nodes as well as for links.The proposed implementation can also classify different types of faulty behavior for both sensor nodes and links in the network.We present the obtained theoretical results and computational complexity of the implemented model for this particular study on automated fault diagnosis.The performance of the model is evaluated using simulations and experiments conducted using indoor and outdoor testbeds.
文摘The problem of automatic and accurate forecasting of time-series data has always been an interesting challenge for the machine learning and forecasting community.A majority of the real-world time-series problems have non-stationary characteristics that make the understanding of trend and seasonality difficult.The applicability of the popular deep neural networks(DNNs)as function approximators for non-stationary TSF is studied.The following DNN models are evaluated:Multi-layer Perceptron(MLP),Convolutional Neural Network(CNN),and RNN with Long Short-Term Memory(LSTM-RNN)and RNN with Gated-Recurrent Unit(GRU-RNN).These DNN methods have been evaluated over 10 popular Indian financial stocks data.Further,the performance evaluation of these DNNs has been carried out in multiple independent runs for two settings of forecasting:(1)single-step forecasting,and(2)multi-step forecasting.These DNN methods show convincing performance for single-step forecasting(one-day ahead forecast).For the multi-step forecasting(multiple days ahead forecast),the methods for different forecast periods are evaluated.The performance of these methods demonstrates that long forecast periods have an adverse effect on performance.
