Low-Overhead and High-Precision Prediction Model for Content-Based Sensor Search in the Internet of Things

Zhang, Puning; Liu, Yuanan; Wu, Fan; Liu, Suyan; Tang, Bihua

doi:10.1109/lcomm.2016.2521735

Cited by 37 publications

(17 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The prediction performances of proposed HCPM are verified here by being compared with the typical MPM [23] and MSE [21] methods which are based on the shallow learning theory. Simulation results are shown below.…”

Section: ) Validations On Prediction Methodsmentioning

confidence: 93%

“…In our previous work [21], based on the Grey System model, we proposed the sensor quantitative state prediction method, MSE, towards quantitative state based sensor search in IoT. Then in literature [22], [23] we further presented high-accuracy sensor quantitative state prediction method, MPM, which explored the change trend of sensor quantitative state by mining the time correlations between historical quantitative state data, thereby accurately predicting the sensor current quantitative state. Based on this, the current quantitative state based sensor search is performed, which largely reduce the communication overhead of sensor search process.…”

Section: Related Workmentioning

confidence: 99%

“…Moreover, the above mentioned research overlooks users search requirements for the sensor historical quantitative state which the user may utilize to perform historical data analysis tasks for making commercial or industrial decisions. Our previous work [23] proposed the segmented least squares polynomial approximating method, SLSPA, to approximately fit the original sensor data. The sensor only needed to upload the fitting polynomial, including a few polynomial coefficients and one constant term, rather than the original data series to the gateway, so as to largely reduce the data transmission energy consumptions.…”

Section: Related Workmentioning

confidence: 99%

“…Recent relevant research, like MAX [13], Microsearch [14], and Snoogle [15], drew on the idea of traditional Internet search engine that pre-stored the static keyword index information about the observed entity in the sensor, which perform an ergodic search for match sensors with dynamically varying state at the cost of large communication overhead. Aiming at addressing this issue, several search mechanisms with sensor state prediction methods were proposed in literature [16]- [23], where candidate entities were preliminarily chosen by predicting the current positions or observation content state of sensors in advance, and then the search system verified the actual state of candidates by accessing their associated sensors to find match entities. Thus only a fraction of sensors need to be accessed, so as to greatly reduce the communication overhead caused by existing ergodic search methods.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

SDU: State-Based Dual-Mode Sensor Search Mechanism Toward Internet of Things

Zhang

Liu

et al. 2019

IEEE Access

Self Cite

View full text Add to dashboard Cite

The wide-scale deployments and interconnections of massive diverse sensors in Internet of Things (IoT) enable the mapping of multimedia entity state data from the physical world into the cyberspace. Sensor search technologies further facilitate the access to multimedia entity state data. However, existing sensor search mechanisms overlook the search requirements for historical state information on entities. Besides, previous current state based sensor search mechanisms realize the low-cost sensor search through predicting the sensor current state based on shallow learning theory, which leads to limited prediction accuracy and higher communication overhead. Aiming at these shortages, an efficient dual-mode sensor search mechanism is firstly designed, including the current state based and historical state based search methods, to fulfill the search needs for current and historical state information on entities. Then based on deep learning theory, a high-accuracy data prediction method towards sensor current state is presented to improve the insufficient accuracy of existing prediction methods. Moreover, a lightweight data representation method is devised to fit the sensor historical state, thereby accomplishing space-saving and low computational overhead historical state based sensor search. Simulation results show that the proposed sensor search mechanism combined with the designed prediction method and approximate representation method can effectively enhance the communication overhead, recall ratio and precision ratio performances. INDEX TERMS Internet of Things, multimedia data, deep learning, sensor search, dual-mode.

show abstract

Section: ) Validations On Prediction Methodsmentioning

confidence: 93%

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

SDU: State-Based Dual-Mode Sensor Search Mechanism Toward Internet of Things

Zhang

Liu

et al. 2019

IEEE Access

Self Cite

View full text Add to dashboard Cite

show abstract

“…This model adopted a new data acquisition scheme ASBP, which used highly correlated spatio-temporal data in the network. Zhang et al proposed LHPM [25], which retained the reference value of sensor output during query and realized high-precision prediction, so as to reduce communication overhead as well as the cost of storage and energy. Jiang et al [26] believes that the performance of support vector machines depends on setting appropriate parameters.…”

Section: Related Workmentioning

confidence: 99%

An Efficient Preference-Based Sensor Selection Method in Internet of Things

et al. 2019

View full text Add to dashboard Cite

While it is well understood that the Internet of things (IoT) can facilitate numerous applications (e.g., environmental supervision, forest fire prevention and Intelligent farming), it also brings a significant challenge for efficiently selecting sensors that meet users' preference and specific requirement from millions of heterogeneous sensors. In this paper, we propose an improved fast nondominated sorting algorithm for efficiently preference-based sensor selection in IoT. Specifically, this proposed method mainly includes three parts: 1) Offline constructing R-tree to search sensor resources and narrowing the size of dataset according to user's preference; 2) Using an improved fast nondominated sorting approach to get nondominated front; 3) Employing TOPSIS to characterize every sensor option of the nondominated front. In order to illustrate the usability of the model, we conduct experiments on several simulation datasets. Experimental results show that this method outperforms several baselines in terms of both response time and accuracy.

show abstract