How apodwikat/nato-sps-programme works
Project type and purpose
• This is a data-science/ML research repository built under the NATO SPS program. Its goal is to ingest Mediterranean SST and weather data, explore and engineer features (DS-Practice), fit classical time-series and regression models (ClassicalModels), then package results into a final NATO report (NATO-Report).
File structure
• Top-level directories:
• ClassicalModels – Jupyter notebooks implementing decision trees, AR/ARMA, linear regression on SST and related features
• DS-Practice – notebooks for raw data ingestion (.nc → CSV), factor analysis, Prophet forecasts, early data exploration
• NATO-Report – the clean dataset, the final report notebooks, PDF and PPT deliverables
• Config files: .gitignore, LICENSE, README.md
• No microservices or web frontend—this is a single‐machine data-pipeline/research process.
README
• Confirms: “DataScience and Machine Learning models on SST and weather forecasts in the Mediterranean region.”
• No additional architectural details; everything lives as Jupyter notebooks operating over CSV inputs.
a. Main components
1. Data Sources
– Raw CSV files (e.g. SST-Antalya-2007-2023-Cleaned.csv, Adana_Data(csv).csv)
– Raw netCDF data (ingested by Open(.nc into csv).ipynb)
2. Data Preprocessing & Feature Engineering
– DS-Practice notebooks: conversion, cleaning, factor analysis
3. Modeling Engines
– ClassicalModels notebooks: decision trees, AR, ARMA, linear regression
4. Model Outputs & Artifacts
– Generated CSV results (e.g. NATO-RESULTS.csv)
5. Reporting & Delivery
– NATO-Report folder: final report notebook, PDF, PPT
b. Relationships & interactions
• DS-Practice → ClassicalModels (cleaned data feeds modeling)
• Both DS-Practice and ClassicalModels feed into Model Outputs
• Model Outputs → NATO-Report (final analysis, notebook → PDF/PPT)
c. Architectural patterns/design principles
– Linear pipeline: Ingestion → Cleaning → Modeling → Reporting
– Clear separation of concerns by folder: exploratory vs. classical models vs. final deliverables
– Reproducibility via Jupyter notebooks
d. Key technologies & libraries
– Python & Jupyter Notebooks
– pandas, numpy for data manipulation
– statsmodels, Prophet for time-series modeling
– CSV and netCDF data formats
– Output formats: CSV, PDF, PPT