Improving prediction accuracy for WSN data reduction by applying multivariate spatio-temporal correlation.

Area of Science:

• Computer Science
• Electrical Engineering
• Network Engineering

Background:

• Data reduction techniques in wireless sensor networks (WSN) aim to conserve energy by minimizing data transmission.
• Current prediction methods for data reduction in WSNs often lack sufficient accuracy.
• Existing approaches frequently rely on temporal correlation, which may not capture complex data relationships.

Purpose of the Study:

• To propose and evaluate a novel data reduction method for WSNs utilizing multivariate spatial and temporal correlation.
• To enhance prediction accuracy compared to existing methods.
• To improve energy efficiency in WSNs through more effective data reduction.

Main Methods:

• Development of a prediction model based on multivariate spatial and temporal correlation.
• Assessment of the proposed method using simulations with simple and multiple linear regression.
• Comparison of prediction accuracy against time-based prediction and existing WSN data reduction solutions.

Main Results:

• The proposed method demonstrates higher correlation between gathered inputs compared to using time as the sole independent variable.
• Multiple linear regression yielded the most accurate predictions within the evaluated models.
• The proposed approach achieved approximately 50% improvement in humidity prediction and 21% in light prediction over current solutions.

Conclusions:

• Multivariate correlation offers a superior approach for prediction in WSN data reduction compared to traditional methods.
• The developed method significantly improves prediction accuracy, leading to enhanced WSN performance.
• This work pioneers the application of multivariate correlation for WSN data reduction, setting a new benchmark for accuracy and efficiency.