Building's Fire in Gimhae
Last updated
Last updated
Problem: Predictive Modeling of Building's Fire in Gimhae, South Korea 2021.
Question: How can we make a model, in the situation where the trend of data differs by season.
Approach: We make 4 models for each season. Because a seasonal effect was so strong that it changes whole covariance structure of X variable.
Model: We used a bayesian logistic regression. In the bayesian setting, we assumed coefficients are random variable. It made possible to estimate the confidence interval of coefficients leading to more deep analysis about effect of explanatory variables.
Evaluation: This project doesn't suggest the actual interpretation, just it interprets the meaning of the coefficients. It was the obvious limitation of this project.
Period: 2021.05-2021.06
Prize: First out of 3 team in ESC club.