Machine and Deep Learning Techniques for Wireless IoT Big Data Analytics

In the age of the IoT, an extensive number of sensing devices acquire or generate a lot of sensory data at a definite period of time for a broad range of applications. Based on the applicability of the application, these objects result in big data streams. Applying data analytics over such big data streams to identify and explore new types of information, forecast upcoming perceptions, and make decisions accordingly in a fair manner is considered to be a crucial process. This makes IoT a notable prototype for both home and industrial applications and is therefore found to be of higher value resulting in technology that improves the quality of life. This section is an overview of using a class of advanced machine learning techniques [6], namely deep learning, to smoothen the analytics and learning in the IoT area. In this chapter, a brief review of machine and deep learning for IoT is presented. Besides, the applications of IoT with big data analytics in wireless mode are also explained.

  • 4.4.1 Introduction to Machine and Deep Learning
  • 4.4.1.1 Design Considerations in Machine Learning

One of the applications of artificial intelligence is machine learning. Machine learning provides systems the potentiality to learn from activity or objects in an automatic manner and upgrade from the current state without being exceptionally programmed. On the other hand, machine learning [7] concentrates on the design and development of computer programs with the objective of accessing the data and learning from them. The learning process starts with the observations for the purpose of observing the patterns in data and therefore ensuring decisions in a better manner in the near future according to the circumstances provided by the user. The main objective in the design of machine learning remains in allowing the computers learn in an automatic manner without the involvement of humans and make actions accordingly. Figure 4.7 shows the schematic representation of machine learning using a flower as an object.

As illustrated in the figure, with an image as input, the machine learning model initially performs manual feature extraction. Manual extraction of features results in a considerable amount of time being consumed and therefore involves computational complexity. Following feature extraction by human beings, classification of whether the given flower represents a rose, lotus, or sunflower is performed. Flere, classification is performed via machine learning.

Schematic diagram of deep neural network

Figure 4.8 Schematic diagram of deep neural network.

4.4.1.2 Design Considerations of Deep Learning

Deep learning on the other hand is a subgroup of machine learning. Deep learning is similar to machine learning with the only difference being the capabilities with which it is said to be performed. Conventional machine learning methods are progressive but require a certain direction. But with deep learning methods [8], the algorithms decide on whether the prediction involved in deep learning is correct or not. The basic structure of deep learning method is designed in such a manner that it draws conclusions regarding the objects of resultant prediction just like a human being. In order to achieve this objective, deep learning utilizes a layered structure of algorithms referred to as the artificial neural network (ANN), analogous to the neural network of the human brain. This makes deep learning more effective than conventional machine learning methods. Figure 4.8 shows the schematic representation of deep learning using a flower as an object.

As illustrated in the figure, features are not extracted manually; rather features are learned by applying the convolutional neural network. With the learned features and by applying deep neural network, the images are classified.

4.4.2 Machine and Deep Learning Methods

Machine learning uses the following four methods.

Supervised Machine Learning Method. Here, past events or occurrences are observed using labeled examples. These past events are applied to the new data to predict the future course of action or events.

Unsupervised Machine Learning Method. In unsupervised machine learning method, the information to be used is neither classified nor labeled. Here, a hidden structure is described from unlabeled data; however, it is said that the correct output is not obtained. Only inferences are said to be obtained.

Semi-Supervised Machine Learning Method. This method lies between supervised machine learning method and unsupervised machine learning method. Here, only a small amount of labeled data and a large amount of unlabeled data are observed.

Reinforcement Machine Learning Method.-. This is a learning method that communicates with the environment resulting in actions and identifies the presence of errors. One of the most pertinent characteristics of this method is learning by trial and error.

Besides the machine learning methods given above, some of the deep learning methods are listed below:

■ Convolutional neural network

■ Recurrent neural network

■ Long short-term memory

■ Autoencoders

■ Sparse autoencoders

■ Stacked autoencoder

■ Backpropagation

■ Stochastic gradient descent

■ Learning rate decay.

4.4.3 Utilization of Learning Methods for IoT Big Data Analysis

Compared to conventional machine learning methods, deep learning methods have gained popularity in the recent years. Though ANNs have gained popularity in the past decades, it is difficult to apply them if the number of layers is increased. Besides, with higher computational complexity involved in the implementation of feedforward neural networks, the implementation process is found to be cumbersome. With these limitations involving computational complexity and more data, deep learning techniques have gained from progressions in efficient training algorithms of deep networks.

4.4.4 Applications of IoT with Big Data Analytics in Wireless Mode

The demands of big data and analytics in IoT have aggressively improved over the years and have thus resulted in dramatic advancements in decision-making processes. Due to this, the need and demand for applying data analytics to big data in IoT have expanded as well, thereby altering the way data collection, storage, and analysis are being made. Besides, big data and analytics have higher prospects for acquiring information in a meaningful form from the data collected by sensor devices. The widespread need for big data and IoT interprets the functional and nonfunctional descriptions for data analytics. Some of the applications of IoT with big data analytics in wireless mode are given below.

4.4.4.1 Smart Cities

In the recent future, approximately 70% of the world’s population is expected to reside in cities. To accommodate the new demand, rise in the number of cities with the increase in the population (i.e., huge data) residing in these cities, municipalities around the globe are turning toward IoT [9] to improve their services, cut costs, and increase the awareness for a safer and greener environment. With big data analytics, these are said to be done by implementing efficient water supply via smart meters that have the potentiality to improve leak detection and also ensuring consumers have real-time access to information pertaining to consumption. Next, IoT in smart cities can pave the way toward providing solutions for congestion-free traffic. These are to be handled by fixing smart signals, by alerting drivers before a possible accident, and suggesting diversions accordingly. Finally, energy-efficient buildings are also the focus nowadays with the objective of minimizing the emission of carbon monoxide.

4.4.4.2 Transport Sector

IoT big data have even made swift changes in the transportation sector. In the current era, the field of transportation has also evolved as a blooming sector with the availability of automatic cars with sensors and road traffic signals to sense the traffic and make changes in an automatic manner. Besides, several sensors fixed in the vehicle even provide information pertaining to the ongoing status of the vehicle, so that any congestion is intimated to the users in a consistent manner.

4.4.4.3 Weather Forecasting

There are several applications of IoT in environmental monitoring. Some of them include protection of environment, continuous monitoring of weather, and monitoring of water safety. In all the above-mentioned applications, the purpose of sensors lies in detecting and measuring environmental changes on a regular basis. Through several sensors, big data on pollution in air and water are said to be easily collected via repeated sampling. This assists in averting considerable pollution and associated calamities. The applications of IoT big data permit operations to reduce human involvement in farming system analysis and regularly monitor the same. With the introduction of IoT big data in the field of air and water pollution, the systems possess the potentiality to recognize changes in irrigation, soil analysis, and environment accordingly. This enables us to avert considerable contamination and related calamities. In the area of weather monitoring, though strong and advanced systems in the present day permit deep monitoring, they do not support widely used instruments like radar and satellites. However, with the introduction of IoT, big data technologies in the area of weather monitoring support fine-grained data and provide better accuracy and flexibility. By efficient and effective weather monitoring, early detection and response to calamities and prevention of loss of life and property are said to be assured.

4.4.4.4 Agriculture

In the field of agriculture, the advent of Internet and several things connected together paves way for both IoT and big data. With the inception of big data in the field of agriculture, involving information regarding soil type, irrigation type, crops to be cultivated, and IoT involved, precision in farming is said to be achieved by introducing smart instruments to measure the drones for field monitoring, water management, soil monitoring, and process monitoring. Besides, livestock monitoring is said to be performed deliberately by using cow monitoring solutions, installing feasibility of rancher for cattle location, and so on.

4.4.4.5 Healthcare

IoT appliances have been advantageous in health and wellness applications. To monitor the health condition of a patient continuously, several wearable devices have found a profound place in the industry. With the introduction of IoT, health applications [10] have been made available for the elderly and patients with critical health conditions. In the current era, with huge amounts of data (i.e., big data) regarding the patient to be analyzed, IoT sensors are utilized to monitor and record the readings of a patient’s health conditions pertaining to blood pressure, sugar level, temperature, and so on and transmit warnings in case any abnormal indicators are said to be identified. Figure 4.9 shows the block diagram of IoT in a

healthcare system. There are a number of applications of IoT in the healthcare sector. With the introduction of IoT in the healthcare applications [11], patients’ lives and living conditions are said to be improved. This is said to be achieved via smart medical sensors to monitor different body parameters such as breathing, blood pressure, and sugar level pertaining to different patients (i.e., big data). Pressure monitoring and several other fitness monitoring sensors have resulted in the increase in the frequency of health monitoring.

With this, elderly patients have the potentiality of monitoring their own health frequently. Besides, even before the condition of a patient who arrives at the hospital is found to be critical, the hospital authorities have the technology to diagnose and start the treatment immediately.

Conclusion

In this chapter, ongoing research work on the IoT platform that allows a network of devices that communicate, analyze, and process information in a collaborative manner is presented. The IoT network generates huge amounts of data in several forms and formats that are stored and processed in cloud using different materials and methods, and this chapter provides its framework and working to understand certain issues concerning IoT devices, their interconnectedness, and services they may offer, including an efficient, effective, and secure analysis of the data. Besides, this chapter also provides challenges faced by IoT due to the large volume of data (i.e., big data) involved in wireless communication. Apart from the challenges faced, the design of modern ubiquitous wireless communication with IoT and its implementation to fit in several real-time applications are included. Also, with the big data involved in the design of IoT for wireless communication, to fit in real-time applications, machine learning methods and, specifically, deep learning methods with a detailed comparison of machine learning and deep learning are included. Finally, the chapter ends with the applications of IoT with big data analytics in the areas of smart city, transport, and agriculture.

References

  • 1. Huang-Chen Lee and Kai-Hsiang Ke, “Monitoring of Large-Area IoT Sensors Using a LoRa Wireless Mesh Network System: Design and Evaluation”, IEEE Transactions on Instrumentation and Measurement, Vol. 67, No. 9, 2177-2187, 2018.
  • 2. Soenke Ziesche, Innovative Big Data Approaches for Capturing and Analyzing Data to Monitor and Achieve the SDGs. Bangkok: United Nations, ESCAP, Economics and Social Commission for Asia and the Pacific East and North-East Asia Office.
  • 3. Dewan Md. Farid, Mohammad Abdullah Al-Mamun, Bernard Manderick, and Ann Now, “An Adaptive Rule-Based Classifier for Mining Big Biological Data”, Expert Systems with Applications, Vol. 64, 305-316, 2016.
  • 4. Nilanjan Dey, Aboul Ella Hassanein, Chintan Bhatt, and Amira S. Ashour, Internet of Things and Big Data Analytics Toward Next-Generation Intelligence. Vol. 30. Berlin: Springer.
  • 5. Neetesh Kumar and Deo Prakash Vidyarthi, “A Green Routing Algorithm for IoT- Enabled Software Defined Wireless Sensor Network”, IEEE Sensors Journal, Vol. 18, No. 22, 9449-9460, 2018.
  • 6. Feras A. Batarseh and Eyad Abdel Latif, “Assessing the Quality of Service Using Big Data Analytics with Application to Healthcare”, Big Data Research, Vol. 4, 13-24, 2015.
  • 7. Mohammad Saeid Mahdavinejad, Mohammadreza Rezvan, Mohammadamin Barekatain, Peyman Adibi, Payam Barnaghi, and Amit P. Sheth, “Machine Learning for Internet of Things Data Aanalysis: A Survey”, Digital Communications and Networks, Vol. 4, No. 3, 161-175, 2017.
  • 8. Mehdi Mohammadi, Ala Al-Fuqaha, Sameh Sorour, and Mohsen Guizani, “Deep Learning for IoT Big Data and Streaming Analytics: A Survey”, IEEE Communications Surveys & Tutorials, Vol. 20, No. 4, 2923—2960, 2018.
  • 9. Zaheer Khan, Ashiq Anjum, Kamran Soomro, and Muhammad Atif Tahir, “Towards Cloud Based Big Data Analytics for Smart Future Citi es”, Journal of Cloud Computing, Vol. 4, No. 1, 1—11, 2015.
  • 10. Maruf Pasha and Syed Muhammad Waqas Shah, “Framework for E-Health Systems in IoT-Based Environments”, Wireless Communications and Mobile Computing, Vol. 2018, 11,2018.
  • 11. David Windridge and Miroslaw Bober, “A Kernel-Based Framework for Medical Big- Data Analytics”. In: Interactive Knowledge Discovery and Data Mining in Biomedical Informatics: State-ofthe-Art and Future Challenges. Lecture Notes in Computer Science, 8401. Springer, pp. 197-208, 2014.
 
Source
< Prev   CONTENTS   Source   Next >