/Datathon-for-social-good-2019-UCB

This is for the DSS Datathon for Social Good in 2019 Fall at UC, Berkeley. Our Project is to forecast wildfire scale and duration by exploring historical wildfire dataset and establishing several prediction models.

Primary LanguageJupyter Notebook

Datathon-for-social-good-2019-UCB

This is for the DSS Datathon for Social Good in 2019 Fall at UC, Berkeley. Our Project is to forecast wildfire scale and duration by exploring historical wildfire dataset and establishing several prediction models.

Team Members : Tianyue Chen, Yongming Zhu, Yuyang Zhao, Hao Zhang(2020 IEOR Meng) Feicheng Qi(2020 MS Statistics)

Statement: Wildfire has caused great inconvenience for our daily life, especially in California. We want to forecast the impacted area and the expected duration of wildfire in order to make fully preparation for this disaster.

Solution: We explored the historical data from Kaggle(the wildfire records in U.S. from 1999 to 2015) and then checked the distribution of wildfire regarding to causes, scale and duration in different regions. In this competition, we mainly focused on the wildfire happened in California and then established two multi-class classification models based on XGBoost and SVM respectively for the prediction of wildfire scale and duration.

https://docs.google.com/presentation/d/14LV47Vjo9k09LQmbaFMvb9gwFk9D0Hif/edit#slide=id.p6