TabularDates unconfirmedRichter's Predictor: Modeling Earthquake Damage
Can you predict the level of damage to buildings caused by the 2015 Gorkha earthquake in Nepal based on aspects of building location and construction?
Official page previewCaptured Oct 19 · not live
Open the competition on DrivenData↗
Open the competition on DrivenData↗- Effort
- 5–15 hours
- Difficulty
- Intermediate
- Compute
- CPU only
- Scoring
- Automatic
At a glance
Full rules ↗- Scoring
- “To measure the performance of our algorithms…” The official page's words
Is it for you?
Prepare before joining
Beginner fit60/100
Learning value82/100
Who it suits
Learners who want to practice multiclass classification on a large structured dataset with geographic and categorical features.
What makes it hard
The dataset is large with many categorical building-attribute features, the three damage grades have an ordinal relationship that standard classifiers do not exploit, and the micro-averaged F1 metric requires balanced performance across classes.
Getting started
Skills to bring
- Python
- pandas and data wrangling
- scikit-learn or gradient boosting libraries
- understanding of F1 score and multiclass metrics
From the organizer
- DataSurvey dataset on earthquake impacts, household conditions, and demographics
- TutorialBenchmark blog post for getting started ↗
First competition? Read No GPU? Pick Competitions That Still Teach You Good Habits