Predicting Intensity of Hurricane
This script analyze Risk-Score calculated by USC-ANRG team to determine if the Risk-Score of different regions in LA county is trending up or down.
Note: These files are developed and tested in Python 3.7.6.
The following packages are required to run the codes: -jupyter noteboook -numpy -scipy -pandas -sklearn -datetime -matplotlib -re -timeit
================================================================================= "RI Prdiction Model"
Required input files: “Dataset_SHIPS_RII_ATL.csv"
================================================================================= ============================= Results ==================================
- Evaluation Table
- A table listing all different models, parameters, sampling methods, and features used to predict Hurricane Intensity Column Description: - Normalized: Shows whether the feature data is normalized or not - Predictor Set: The name of the feature set used to predict hurricane intensity - ML Name: Types of machine learing algorithm used - Sampling Method: Shows which sampling method used - True Positive: Number of correctly predicted high-intensity hurricane - True Negative: Number of correctly predicted low-intensity hurricane - False Negative: Number of incorrectly predicted low-intensity hurricane - False Positive: Number of incorrectly predicted high-intensity hurricane - PSS: Peirce Skill Score (PSS) - FAR: False Alarm Ratio (FAR) - POD: Probability of Detection (POD)/True Positive Rate(TPR)/Recall/Sensitivity - F1: weighted average of precision and recall - Precision: Positive Predictive Value - Brier Score: a score function that measures the accuracy of probabilistic predictions
================================================================================= ================================= GitHub Repository ============================= We will also make the raw python scripts available via our GitHub repository: http://localhost:8888/tree/Guided%20Project/Hurricane-Prediction-Model
================================================================================= ================================= Contact Info ================================== Kooha Kwon: [email protected]