Fires resulting from human activities, encompassing arson, electrical problems, smoking, cooking mishaps, and industrial accidents, necessitate understanding to facilitate effective prevention. This study investigates human-caused fires in Keelung City, Taiwan, employing geographic information system (GIS)-based dimensionality reduction techniques. By analyzing eleven diverse factors, including fire incident density, population-related, building-related and economic-related features, valuable insights are gained for enhancing fire prevention. Utilizing principal component analysis (PCA), factor analysis (FA), and out-of-bag (OOB) predictor importance, our algorithm identifies key factors explaining dataset variance. Results from three approaches reveal a significant link between fire incidents and the elderly population, buildings over 40 years old, and the tertiary sector in the economy, contributing to developing effective measures for mitigating and managing fire occurrences.