From Raw Data to Insightful Decisions: A Framework for Environmental Data Management, Analytics, and Visualization for Smarter Cities

Abstract

Smart cities depend on continuous streams of environmental data to monitor air quality, detect emerging risks, and guide effective responses. Yet these data are frequently incomplete, noisy, and inconsistent, limiting their reliability for decision-making. Overcoming this challenge requires more than isolated algorithms. It demands an integrated framework that addresses data quality, predictive insight, and human-centered interpretation.

This thesis introduces a unified solution for environmental data analytics, demonstrated using five years of gas and weather data from Jubail Industrial City. The first contribution presents a dual-phase cleaning framework for multivariate time series. Statistical and machine learning techniques are used to detect anomalies, followed by context-aware interpolation to reconstruct missing values. This method preserves temporal structure and produces a high-resolution, analysis-ready dataset.

The second contribution applies advanced forecasting using representative models such as XGBoost to predict carbon monoxide concentrations with high accuracy. These predictions are enriched by descriptive and spatiotemporal analysis, revealing spatial patterns and seasonal trends that support a shift from reactive mitigation to proactive air quality management.

The third contribution translates analytical output into an interactive visual analytics platform. Designed for performance and usability, the system enables decision-makers to explore sensor data and forecasts through maps, heatmaps, and time series visualizations. Its human-in-the-loop design integrates expert judgment into the interpretation process, promoting transparency, trust, and context-aware decision-making.

Together, these contributions form a cohesive framework that transforms raw environmental data into actionable intelligence. By combining rigorous data preparation, predictive modeling, and intuitive visual tools, this research enhances the capacity of smart cities to make faster, more informed, and more effective decisions.