Full length article
Deep learning with LSTM based distributed data mining model for energy efficient wireless sensor networks

https://doi.org/10.1016/j.phycom.2020.101097Get rights and content

Abstract

Wireless sensor network (WSN) comprises a collection of sensor nodes employed to monitor and record the status of the physical environment and organize the gathered data at a central location. This paper presents a deep learning based distributed data mining (DDM) model to achieve energy efficiency and optimal load balancing at the fusion center of WSN. The presented DMM model includes a recurrent neural network (RNN) based long short-term memory (LSTM) called RNN-LSTM, which divides the network into various layers and place them into the sensor nodes. The proposed model reduces the overhead at the fusion center along with a reduction in the number of data transmission. The presented RNN-LSTM model is tested under a wide set of experimentation with varying number of hidden layer nodes and signaling intervals. At the same time, the amount of energy needed to transmit data by RNN-LSTM model is considerably lower than energy needed to transmit actual data. The simulation results indicated that the RNN-LSTM reduces the signaling overhead, average delay and maximizes the overall throughput compared to other methods. It is noted that under the signaling interval of 240 ms, it can be shown that the RNN-LSTM achieves a minimum average delay of 190 ms whereas the OSPF and DNN models shows average delay of 230 ms and 230 ms respectively.

Introduction

In general, Wireless Sensor Network (WSN) is a self-configured and infrastructure-less wireless networks that helps to observe the external and ecological status, like temperature, moisture, movements and pollutants to pass the information via network to sink from the data might be monitored as well as predicted. A sink or base station (BS) has been treated as interface among the network and user. By using such network, the user can able to derive essential data by inducing queries and collect the required details from BS. Generally, a WSN is composed of numerous sensor nodes. Here, sensors are capable of communicating with alternate nodes through the radio signals. It embeds processing units, storage, radio transceivers and power elements. A single node from WSN is composed of restricted computing speed, memory, communication bandwidth and so on. Once the sensor node has been injected, it is responsible to self-organize in a suitable network infrastructure along with multi-hop communication process within the system. Furthermore, wireless sensors acknowledge for queries provided from a “control site” in order to process only particular rules and sensing samples. Global Positioning System (GPS) as well as local positioning techniques could be applied to derive the position and related data. It is constrained with actuator which is considered as to be used only in specific situations. Sometimes, it is assumed to be Wireless Sensor and Actuator Networks.

WSN is capable to adopt novel techniques and acquires non-conventional method for a protocol development because of various conditions. Due to the need for minimum complexity and energy utilization for prolonged network lifespan, an appropriate balance among signal and data computing abilities should be identified. It leads to providing maximum energy in scientific events. Recently, various types of developments in WSN takes place in developing energy and computationally effective techniques, whereas the domain is limited to simple data-oriented and reporting fields. Moreover, a Cable Mode Transition (CMT) method that helps to compute the lower value of active sensor nodes to balance the K-coverage from terrain and K-connectivity of the system. In particular, it declares the time period of inactive cable sensors with no influence of coverage as well as connection requirements of network which is depend upon the local data.

Several energy efficient solutions for WSN operations based on optimization algorithms and deep learning (DL) models have been presented in the literature. In Cheng et al. [1], a delay-aware data collection network for WSN is deployed. It mainly aims in reducing the latency in data collecting process of WSN that tends to elaborate the network lifetime. In Rahman and Matin [2], it is assumed that more number of relay nodes has been adopted to decrease the network vulnerability as well as Particle Swarm Optimization (PSO) model is applied to place an optimized sink position in terms of relay nodes to resolve the lifespan issue.

The design of WSN includes several constraints. The most essential constraint is because of the fact that sensor nodes are placed in an adverse region and it should be often recharged with batteries. Hence, the sensor lifetime is dramatically reduced compared with quantity of power induced in battery and the way of conserving power. The management of power utilization of sensors has been developed as an active research area. The purpose of energy conservation is employed in data acquisition, computing, reduction, transmission, etc. [3], [4], [5]. The data transmission process is considered to be the initial step in saving energy. Therefore, such protocols could be treated as various layers in communication namely physical layer, MAC layer, routing layer as well as application layer. In case of MAC layer, maximum amount of power has been attenuated while retransmitting the data once the collision is completed, and control packet transmission in the absence of applicable data or determining packets to attain alternate sensor nodes. Only few protocols have been presented to be treated on energy consumption. The S-MAC protocol [6], applies time synchronization from sensors to isolate in a cycled manner.

In order to save power and eliminate collision, PC-MAC protocol [7] has been deployed. In network layer, routing protocols helps to reduce the power application by the mechanism of packet delivery. Besides, it attempts to grab merits of higher density of WSN. By developing novel WSN routing protocols is a most promising issue. It is divided using the mechanism applied. Hence, routing protocols might be geocentric, data-centric, applies network topology as well as link states. In case of data-centric protocols, GKAR is assumed to be the instance of K-any cast protocol. EASPRP [8] seeks for shortest path including energy efficiency and EERT protocol [9] manages the Quality of Service (QoS). In real-time applications, REFER protocol [10] employs Kautz graphs. Only some of the routing protocols could be applied in a combined manner in order to encircle a particular region. Few routing protocols [11] apply the motion ability of mobile sensor nodes. The GAROUTE protocol [12] utilizes genetic algorithm (GA) to develop group of sensors to minimize communications. Here, cluster heads (CH) could be deployed to compute local information as well as to divide the respective data. LEACH protocol is an applied cluster to manage the network traffic. It has been altered [13] to enhance the delivery time and to reduce the interferences. EEDR protocol [14] mainly concentrates in transmitting packet to decrease the power application. Alternate models has been projected to minimize the power utilization and to improve the lifetime of WSB. Huang J.-W [15] followed sensor coverage is mainly applied for reducing the power application. The same operation has been repeated in [16] that applied an SCC (Sponsored Coverage Calculation) simulating model. Some other protocols alleviate data transmissions such as SEPSen protocol [17] that applies data processing throughout the system. The traffic as well as resource management is utilized in [18]. Major types of protocols apply wireless network simulators [19] respectively.

WSN has been employed in diverse applications namely land cover classification [20], SCR node forecasting in vehicular system, fault analysis, estimating the quality of groundwater [21]. Conventionally, such type of applications helps to determine the sample data from fusion center. In case, WSN is comprised of numerous sensor nodes, the function of computing sampling data has been restricted by using fusion center’s hardware, which is assumed to be costlier and more complex in upgrading frequently. Therefore, data communication absorbs more amount of energy, specifically for wireless relaying nodes. Data mining (DM) methods are induced to obtain applicable data from numerous data in the last a decade, which is assumed to be more efficient tool applied in predicting larger data. For past decades, shallow DM techniques namely, Support Vector Machine (SVM), boosting, as well as Logistic Regression (LR) have been presented [22]. Also, by applying these shallow DM models, it tends to enhance the fusion center’s function; however, the issue in energy consumption remained as unchanged. The solution to resolve these problems is by implementing the techniques in sensors to minimize data transmission, which is more tedious to be executed in WSN. In addition, Hinton and Salakhutdinov [23] developed a deep DM technique named called Deep Neural Network (DNN) that is used in extracting the inner representation as well as to alleviate data dimensionality. The DNN based DMM model has been presented in [24].

Though several models have been available in the literature, it is noted that there is still a need to enhance the fusion performance of the WSN. At the same time, there is a requirement to achieve minimum energy consumption, signaling overhead, average delay with maximum throughput. Sample data of WSN has grown in a rapid way owing to the existence of massive number of sensor nodes, a centralized data mining solution in a fusion center has come across the issue of minimizing the load of the fusion center as well as reducing the overall energy utilization. In this view, this paper presents a DL based distributed data mining (DDM) model with LSTM to achieve energy efficiency and optimal load balancing at the fusion center of WSN. The presented DMM includes a recurrent neural network with LSTM (RNN-LSTM) model which divides the network into various layers and place them into the sensor nodes. Using the RNN-LSTM model, the overhead of the fusion center in WSN is greatly reduced. At the same time, the amount of energy needed to transmit data by RNN-LSTM model is considerably lower than energy needed to transmit actual data. The presented RNN-LSTM model undergoes a wide set of experimentation under varying number of hidden layer nodes and signaling intervals. The experimental outcome stated that the RNN-LSTM reduces the energy consumption, signaling overhead, average delay and maximizes the overall throughput compared to other methods. The advantages of the paper contribution are listed here.

  • (a)

    No requirement of labeling quantity of training data in a manual process in case of various domains, where it is completely done automatically.

  • (b)

    Internal representations are integrated with alternate DM technique and enhanced the models to attain optimal results.

  • (c)

    The dimensionality of RNN-LSTM helps to minimize the data transmission by WSN as well as to conserve the energy of WSN.

  • (d)

    The distributed estimation decreases the overload in fusion center that leads to several benefits like less hardware requirement and delay.

The upcoming portions of the paper are organized as follows. Section 2 elaborates the RNN-LSTM model. Section 3 validates the experimental validation of the proposed model and Section 4 concludes the paper.

Section snippets

Proposed method

The presented RNN-LSTM for DDM is intended to achieve energy efficiency and load balancing at the fusion center of the WSN. The presented DMM includes a recurrent neural network with LSTM (RNN-LSTM) model which divides the network into various layers and place them into the sensor nodes.

Experimental validation

For validating the results of the RNN-LSTM model for DDM in WSN, a set of experiments were carried out in MATLAB R2014a. Here, the results are validated under varying number of hidden layer nodes and signaling interval. The number of hidden layer nodes ranges from 5 to 40 and the signaling interval lies between 240–260. The set of measures used to analyze the performance are signaling overhead, average throughput and average delay. A comparative analysis is also made with OSPF and DNN models.

Conclusion

This study has developed a new RNN-LSTM model for DDM in WSN to achieve energy efficiency and optimal load balancing at the fusion center of WSN. Using the RNN-LSTM model, the overhead of the fusion center in WSN is greatly reduced. At the same time, the energy consumption for processed data transmission by RNN-LSTM model is considerably lower than the transmission of actual data. Here, the results are validated under varying number of hidden layer nodes and signaling interval. The number of

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

Dr. K. Shankar sincerely acknowledge the financial support of RUSA–Phase 2.0 grant sanctioned vide Letter No. F. 24-51/2014-U, Policy (TNMulti-Gen), Dept. of Edn. Govt. of India, Dt. 09.10.2018.

Prof. Dr. Sachi Nandan Mohanty, received his Ph.D. from IIT Kharagpur, India in the year 2014, with MHRD scholarship from Govt of India. He has recently joined as Associate Professor in the Department of Computer Science & Engineering at Gandhi Institute for Technology Bhubanewar. His research areas include Data mining, Big Data Analysis, Cognitive Science, Fuzzy Decision Making, Brain-Computer Interface, Cognition, and Computational Intelligence. Prof. S.N. Mohanty has received 2 Best Paper

References (24)

  • RahmanM.N. et al.

    Efficient algorithm for prolonging network lifetime of wireless sensor networks

    Tsinghua Sci. Technol.

    (2011)
  • GongB. et al.

    An artificial immune network approach to multi-sensor land use/land cover classification

    Remote Sens. Environ.

    (2011)
  • ChengC.T. et al.

    A delay-aware data collection network structure for wireless sensor networks

    IEEE Sensors J.

    (2010)
  • ElhosenyM. et al.

    Energy efficient optimal routing for communication in VANETs via clustering model

  • ElhosenyM. et al.

    Reliable data transmission model for mobile ad hoc network using signcryption technique

    IEEE Trans. Reliab.

    (2019)
  • DuttaA.K. et al.

    An efficient hierarchical clustering protocol for multihop internet of vehicles communication

    Trans. Emerg. Telecommun. Technol.

    (2019)
  • AmmarI. et al.

    An improved s-MAC protocol based on parallel transmission for wireless sensor networks

  • HuJ. et al.

    Energy-efficient mac protocol designed for wireless sensor network for iot

  • DoohanN.V. et al.

    Energy aided shortest path routing protocol (EASPRP) for highly data centric wireless sensor networks

  • MazinaniS.M. et al.

    An energy-efficient real-time routing protocol for differentiated data in wireless sensor networks

  • LiZ. et al.

    A kautz-based real-time and energy-efficient wireless sensor and actuator network

  • FalconR. et al.

    Controlled straight mobility and energy-aware routing in robotic wireless sensor networks

  • Cited by (0)

    Prof. Dr. Sachi Nandan Mohanty, received his Ph.D. from IIT Kharagpur, India in the year 2014, with MHRD scholarship from Govt of India. He has recently joined as Associate Professor in the Department of Computer Science & Engineering at Gandhi Institute for Technology Bhubanewar. His research areas include Data mining, Big Data Analysis, Cognitive Science, Fuzzy Decision Making, Brain-Computer Interface, Cognition, and Computational Intelligence. Prof. S.N. Mohanty has received 2 Best Paper Awards during his Ph.D. at IIT Kharagpur from International Conference at Benjing, China, and the other at International Conference on Soft Computing Applications organized by IIT Rookee in the year 2013. He has awarded Best thesis award first prize by Computer Society of India in the year 2015. He has published 15 International Journals of International repute and has been elected as Member of Institute of Engineers and IEEE Computer Society. He also the reviewer of IJAP, IJDM International Journals.

    Dr. E. Laxmi Lydia is an Professor of Computer Science Engineering at Vignan’s Institute of Information Technology(A). She is a big data analytics online trainer for the international training organisation and she has presented various webinars on big data analytics. She is certified by Microsoft Certified Solution Developer (MCSD). She published more than 100 research papers in international journals in the area big data analytics and data sciences and she published ten research papers in international conference proceedings. She is an author for the big data analytics book and currently she is working on government DST funded project and she holds a patents.

    Dr. Mohamed Elhoseny is currently an Assistant Professor at the Faculty of Computers and Information, Mansoura University where he is also the Director of Distributed Sensing and Intelligent Systems Lab. Besides, he has been appointed as an ACM Distinguished Speaker from 2019 to 2022. Collectively, Dr. Elhoseny authored/co-authored over 85 ISI Journal articles in high-ranked and prestigious journals such as IEEE Transactions on Industrial Informatics (IEEE), IEEE Transactions on Reliability (IEEE), Future Generation Computer Systems (Elsevier), and Neural Computing and Applications (Springer). Besides, Dr. Elhoseny authored/edited Conference Proceedings, Book Chapters, and 10 books published by Springer and Taylor & Francis. His research interests include Smart Cities, Network Security, Artificial Intelligence, Internet of Things, and Intelligent Systems. Dr. Elhoseny serves as the Editor-in-Chief of International Journal of Smart Sensor Technologies and Applications, IGI Global. Moreover, he is an Associate Editor of many journals such as IEEE Access (Impact Factor 3.5), IEEE Future Directions, PLOS One journal (Impact Factor 2.7), Remote Sensing (Impact Factor 3.5), and International Journal of E-services and Mobile Applications, IGI Global (Scopus Indexed). Also, he is an Editorial Board member in reputed journals such as Applied Intelligence, Springer (Impact Factor 1.9). Moreover, he served as the co-chair, the publication chair, the program chair, and a track chair for several international conferences published by IEEE and Springer.

    Majid Alotaibi received Ph.D. from The University of Queensland, Brisbane, Australia, in 2011. Currently, he is an assistant professor in the Department of Computer Engineering, Umm Al Qura University, Makkah, Kingdom of Saudi Arabia. His current research interests include Mobile Computing, Mobile and Sensor Networks, Wireless Technologies, Ad-hoc Networks, Computer Networks (Wired/Wireless), RFID, Antennas and Propagation, Radar, and Nano electronics.

    Dr. K. Shankar is currently a Post Doc Fellow in the Alagappa University, Karaikudi, India. Collectively, Dr. K. Shankar authored/co-authored over 45 ISI Journal articles (with total Impact Factor 115.117) and 150+Scopus Indexed Articles.  He has guest-edited several special issues at many journals published by Inderscience and MDPI. He has served as Guest Editor, Associate Editor in SCI, Scopus indexed journals like Elsevier, Springer, Wiley & MDPI. Dr. Shankar authored/edited Conference Proceedings, Book Chapters, and 2 books published by Springer. He has been a part of various seminars, paper presentations, research paper reviews, and convener and a session chair of the several conferences. He displayed vast success in continuously acquiring new knowledge and applying innovative pedagogies and has always aimed to be an effective educator and have a global outlook.  His current research interests include Healthcare applications, Secret Image Sharing Scheme, Digital Image Security, Cryptography, Internet of Things, and Optimization algorithms.

    View full text