Główne pojęcia
Designing an energy-efficient ensemble approach to mitigate data incompleteness in IoT applications.
Streszczenie
The paper introduces ENAMLE, an energy-aware technique to address data incompleteness due to sensor failures in IoT systems. It focuses on optimizing energy consumption while maintaining accuracy by using an ensemble of sub-models. The study evaluates SECOE's energy bottlenecks and proposes ENAMLE as a proactive solution. By dynamically adjusting the number of models based on missing data rates, ENAMLE effectively reduces energy consumption and optimizes throughput. Experimental studies demonstrate the efficiency of ENAMLE compared to SECOE and the base model across different datasets.
Statystyki
"ENAMLE substantially reduces energy consumption by 7.7%, 25.6%, and 12.3% compared to SECOE-MinM+4, SECOE-MinM+8, and Base model."
"ENAMLE boosts throughput by 35.7% and 77.1% compared to SECOE-MinM+4 and SECOE-MinM+8."
"ENAMLE cuts energy by 19.6% and 38.5% less than SECOE-MinM+4 and SECOE-MinM+8."
Cytaty
"ENAMLE adapts its model selection strategy based on missing data rate, optimizing both energy consumption and accuracy."
"Using a combination of MinM+2 for low and high missing data rates provides a significant decrease in averaged energy compared with individual ensembles by SECOE."
"Adaptive switching between two different ensembles alleviates the impact of concurrent sensor failures while efficiently optimizing energy."