ATMS 597 Project 5 Group A

Group Members: Chu-Chun Chen, Bowen Fang, David Lafferty

ASOS Station: Dubuque, Iowa (KDQB)

Task:
Using Automated Surface Observation System data from NCDC, build a supervised classification algorithm to predict frozen vs. liquid precipitation type.

Notebooks:

• DataProcessing: Parses the METAR data and stores readings for all yeares in a single csv file.
• LogisticRegression: Baseline model; scikit-learn's LogisticRegression with default parameters.
• GradientBoostedDecisionTree: Implementation of sckikit-learn's GradientBoostingClassifier including hyperparameter optimisation.
• RandomForest: Implementation of sckikit-learn's RandomForestClassifier including hyperparameter optimisation.

Results:
(All results are for test data)
Baseline model Brier Skill Score: 0.8607
Gradient Boosted Decision Tree Brier Skill Score: 0.9821
Random Forest Brier Skill Score: 0.9845

The best model was the Random Forest, with an improvement of 0.1238 in Brier Skill Score over the basline model!
Our presentations slideshow can be found here