Oxygen, crampons, data: Can statistics help us climb Everest?

This article focuses on using statistical and machine learning models to predict the success of Mount Everest expeditions based on historical data from the Himalayan Database, which includes 2,214 expeditions since 1921. Various predictive approaches—such as logistic regression, K-nearest neighbors (KNN), bagged trees, random forests, and TreeNets—are compared, with tree-based methods achieving the highest classification accuracy in forecasting expedition outcomes. Key factors influencing success include the use of supplemental oxygen, the chosen route, the number of teams on the mountain, and the duration of the summit push. The analysis highlights how these models can guide expedition planning by estimating probabilities of success and exploring interactions among critical variables, while acknowledging limitations like unmeasured factors and the complexity of human traits.