fire-montesinho

Modeling Forest Fires in Montesinho Natural Park

Project for the subject Intelligent Data Analysis and Machine Learning I, from Potsdam University.

Author: Pedro Alonso Campana

Project structure:

Notebooks

Notebooks of the whole process:

• Fire_montesinho.ipynb: Contains all the parts put together.
• montesinho(summarised).ipynb: Contains the summary used for presenting

Notebooks used during the process:

• exploratory_analysis_preprocessing.ipynb: Contains the detailed data visualization related to data exploration and preprocessing
• model_selection.ipynb: Contains initial model comparaisons searching for a modeling approach.
• whole_model.py: Contains the comparison of the different models

other notebooks:

• Hyperparameter_tuning_classificators_accuracy.ipynb: Contains driver code for classification models and regression models for the splitted model
• modeling_second_halfth.ipynb: Contains driver code for creating the models for the subset of data from june to december
• Custom_SVM_Cutoff_squared_hinge.ipynb: Contains a customized version of the support vector machine. Wasn't able to tune it due to slow optimization/noisy data.

imports

• exploratory_analysis.py: Contains the code used for exploratory_analysis, mostly plots.
• preprocessing.py: Contains the code used for processing the data, normalization of variables, transformations...
• model_selection.py: Contains the code used for comparing different candidate models, hyperparameter tuning...
• model_end_to_end.py: Contains the classes for the end-to-end implementation of the model

API

• app.py: Contains a small API deployment of the model, can be run from the terminal with "python3 app.py"
• request.py: A small example request that can be run from the terminal with "python3 request.py"