We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Fft-asvr: an adaptive approach for accurate prediction of IoT data streams.
- Authors
Maurya, Manish Kumar; Singh, Vivek Kumar; Shaw, Sandeep Kumar; Kumar, Manish
- Abstract
In IoT applications, prediction models have fundamental challenges such as real-time processing, producing results with considerable/without delay, and taking action against pattern drift. While existing models can excel when data statistics remain relatively stable, real-time systems may encounter difficulties, particularly when confronted with dynamic shifts in data behavior. Analyzing data streams generated by different IoT applications and detecting complex pattern on the fly has become an open area of research. Complex event processing with adaptivity is a must to get desired features in such models. To address this issue, a comprehensive model for prediction has been proposed in this paper. It consists of two phases: (1) the basic model is constructed using historical data, (2) a fast Fourier transform-based adaptive support vector regression (FFT-ASVR) approach is proposed to predict events embedded in IoT data streams. FFT-ASVR predicts abnormal events by experiencing a change in data streams with real-time model updation. The performance of FFT-ASVR with a similar existing method SVM-RBF is presented using real-time traffic data of Madrid city. The proposed approach has significant improvement in terms of mean absolute percentage error (MAPE) for prediction, is adaptive in nature, and is also capable of handling the issue of pattern drift.
- Subjects
MADRID (Spain); INTERNET of things; PREDICTION models; FAST Fourier transforms
- Publication
Journal of Supercomputing, 2024, Vol 80, Issue 10, p13976
- ISSN
0920-8542
- Publication type
Article
- DOI
10.1007/s11227-024-05961-w