Forest Fire Burn Area Predictions in Northeast Portugal

517 multivariate instances of forest fire data from northeast Portugal, donated in 2008. The dataset contains 13 real-valued attributes for predicting burned area as a regression task. It was created by Paulo Cortez and Aníbal Morais from the University of Minho for a data mining study published in 2007.

Use Cases

• Predict burned area of forest fires based on meteorological conditions.
• Benchmark regression algorithms like Gaussian SVM on a real-world environmental task.
• Analyze the relationship between weather variables (temp, RH, wind, rain) and fire severity.
• Study the performance of models on predicting small fires, which are described as the majority.

Strengths

• 517 instances provide a substantial sample for model training and validation.
• 13 attributes include direct weather conditions like temperature, relative humidity, wind, and rain.
• The dataset is associated with a published research paper (Cortez and Morais, 2007) detailing methodology and results.
• A specific regression metric result (MAD: 12.71 ± 0.01 for a Gaussian SVM) is provided for benchmarking.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.
• The data may reflect geographic bias inherent to its collection in northeast Portugal.

Provenance

Source

Paulo Cortez and Aníbal Morais, Department of Information Systems, University of Minho, Portugal.

Collection Method

Likely collected from meteorological and other sources for a data mining research project.

Geography

Northeast region of Portugal.