Prediction-Enabled Resource Allocation for Air Quality Monitoring Using Temporal Convolutional Network-Based PM2.5 Forecasting Model in Cloud Datacenters

The impact of air pollution on sustainable development and human health is a growing concern worldwide. To gauge air quality, monitoring stations are strategically positioned, but managing the vast amounts of sensor data they generate is challenging. Cloud data centers offer a promising solution for handling this big data, yet their resource allocation can be inefficient and energy-intensive. This paper introduces a novel prediction-enabled dynamic resource allocation system for air quality monitoring. Operating on a two-pronged approach, the system combines air quality prediction and resource allocation. It leverages a temporal convolutional network-attention mechanism-based model to forecast particulate matter (Pm2.5) levels from multi-variate time series data, specifically using the Beijing Pm2.5 dataset. The crossover boosted artificial lizard search optimization (CALO) algorithm optimizes resource allocation, minimizing energy consumption while maximizing CPU and memory utilization. The results achieve superior air quality forecasts with enhanced efficiency and reduced energy consumption. The developed approach achieves a significantly reduced error range, boasting an MAE of 0.198 and an RMSE of 4.365. Furthermore, it attains an [Formula: see text] value of 0.956, indicating a high predictive accuracy. Thus this system presents an efficient and effective solution for managing big data in air quality monitoring, offering significant improvements in resource utilization and energy efficiency. Its implementation has the potential to address the challenges associated with air quality monitoring in modern society.