Forecasting Continuous Air Pollution with Cloud-Based Big Data Platform using Spatial-Temporal Visualization

In today’s era of rapidly growing data, having a well-designed platform for analyzing and visualizing big data is becoming increasingly important. This study presents a cloud-based big data platform for analyzing and visualizing continuous spatial-temporal air pollution. The platform combines a Spark Hadoop-based machine learning environment with a TensorFlow-based deep learning framework for forecasting and a map for demonstration and visualization of the results. The platform uses the Inverse Distance Weighted (IDW) method to estimate the concentration of particulate matter in the atmosphere and a cloud based non-SQL database to store the pre-calculated air quality images. The platform also features a sliding window mechanism, which allows users to easily visualize the results over time. This work provides an instant overview of air quality in Taiwan and evaluates the performance of the platform, concluding that it is efficient and capable of providing continuous spatial-temporal results. This platform represents a significant step forward in the field of air quality monitoring and forecasting.