48
Internet of Things (IoT)
3.5.2.1 Deep Learning Algorithms and Time Series Data
Considering that most sensor data collected from an IoT environment are time series in
nature, applying deep learning techniques for energy forecasting is gaining the attention
of researchers for developing a sustainable smart grid as a part of the future digital world
(Busseti et al., 2012; Connor et al., 1994; Hong, 2010). This high-dimensional dataset when
implemented with a deep neural network is found to perform better in comparison with
other existing approaches such as linear and kernelized regression techniques and more
importantly do not overfit.
3.5.2.2 IoT Implications
Since 2015, IoT has been emerging but the real impact is yet to be realized as the year
progresses with the development and deployment of a wide area network with 5G 2020
and beyond. Considering the availability of such a Tsunami of mostly time series data, it
will lead to an exponential demand for predictive analytic and for the development of a
viable model and address the challenges thereto.
Our prime focus in this chapter shall be on using deep learning techniques, as a practical solution to IoT data analytic.
3.5.2.3 Implications for Smart Cities
As one can see, smart cities are an application domain for IoT, where digital technologies are used to enhance the performance of the IoT system with a reduction in cost and
resource usage with an aim of active engagement of citizens for effective implementation
to find its benefit at large for well-being of self. The applications include but are not limited
to energy sector, health sector, transport sector, to name a few.
3.5.2.4 Deep Learning as a Solution to IoT
Deep learning basically concerns with neural network structure where many layers are
used to solve the complex situations built from the simpler ones. The ability of deep learning to solve the curse of dimensionality problems with no rules make it an attractive one
in IoT–big data scenario (Marino et al., 2016). The concept takes birth from the question of
learning to identify a cat to recognizing it from its behavior, shape, etc.
As IoT mostly depends on the resource-constraint computing devices, our aim shall be
to check whether deep learning can provide any interesting predictive analytic to obtain
meaningful observations.
3.6 Proposed Methodology and Datasets Used
The proposed methodology of using deep learning in IoT dataset shall address some of the
following questions:
• The applicability of deep learning in smart cities, smart health care, etc.
• What is the performance metrics for prediction?
The proposed methodology adopted in this chapter is shown in Figure 3.4.
Internet of Things (IoT)
3.5.2.1 Deep Learning Algorithms and Time Series Data
Considering that most sensor data collected from an IoT environment are time series in
nature, applying deep learning techniques for energy forecasting is gaining the attention
of researchers for developing a sustainable smart grid as a part of the future digital world
(Busseti et al., 2012; Connor et al., 1994; Hong, 2010). This high-dimensional dataset when
implemented with a deep neural network is found to perform better in comparison with
other existing approaches such as linear and kernelized regression techniques and more
importantly do not overfit.
3.5.2.2 IoT Implications
Since 2015, IoT has been emerging but the real impact is yet to be realized as the year
progresses with the development and deployment of a wide area network with 5G 2020
and beyond. Considering the availability of such a Tsunami of mostly time series data, it
will lead to an exponential demand for predictive analytic and for the development of a
viable model and address the challenges thereto.
Our prime focus in this chapter shall be on using deep learning techniques, as a practical solution to IoT data analytic.
3.5.2.3 Implications for Smart Cities
As one can see, smart cities are an application domain for IoT, where digital technologies are used to enhance the performance of the IoT system with a reduction in cost and
resource usage with an aim of active engagement of citizens for effective implementation
to find its benefit at large for well-being of self. The applications include but are not limited
to energy sector, health sector, transport sector, to name a few.
3.5.2.4 Deep Learning as a Solution to IoT
Deep learning basically concerns with neural network structure where many layers are
used to solve the complex situations built from the simpler ones. The ability of deep learning to solve the curse of dimensionality problems with no rules make it an attractive one
in IoT–big data scenario (Marino et al., 2016). The concept takes birth from the question of
learning to identify a cat to recognizing it from its behavior, shape, etc.
As IoT mostly depends on the resource-constraint computing devices, our aim shall be
to check whether deep learning can provide any interesting predictive analytic to obtain
meaningful observations.
3.6 Proposed Methodology and Datasets Used
The proposed methodology of using deep learning in IoT dataset shall address some of the
following questions:
• The applicability of deep learning in smart cities, smart health care, etc.
• What is the performance metrics for prediction?
The proposed methodology adopted in this chapter is shown in Figure 3.4.