SFSahar Fatemi
← Research

USGS Earthquake Classification

From catalogue data to classification

This project treats earthquake magnitude as a multiclass classification problem: preparing historical catalogue records, constructing features and comparing models that assign events to magnitude categories.

The work connects data engineering with model evaluation. Temporal features describe when events occurred, while quality-related features capture information about the records. Class-imbalance handling addresses the challenge of learning from categories with unequal representation, making it an important part of the comparison alongside model choice.

Built a Spark pipeline for USGS earthquake data and compared models for multiclass magnitude classification.

Developed the data pipeline and contributed to model comparison.

Used Python, PySpark, and Spark ML.

Cleaned, processed, and analyzed USGS earthquake catalogue data.

Engineered temporal and quality related features for multiclass magnitude classification.

Compared Logistic Regression, Decision Tree, Random Forest, and Gradient Boosted Tree models.

Used imbalance handling strategies in the model comparison.

Discuss the pipeline

To discuss the Spark pipeline or magnitude-classification methods, get in touch.