How attiqrafiq/aqi_prediction_model works
Project type and purpose
• This is an end-to-end machine-learning pipeline focused on Air Quality Index (AQI) data for Pakistan.
• It ingests raw time-series pollutant measurements, applies Pakistan’s AQI standard, trains multiple models, evaluates them, and produces a comparison report.
a. Data Sources & Reference Materials
• “final AQI 19-23.csv” / “air_quality_with_AQI.txt” – raw pollutant measurements
• “AQI Standards in Pakistan.docx” – rulebook for AQI calculation
b. Preprocessing Stage
• Component: AQI_preprocessing.ipynb
• Inputs: raw CSV/TXT, standards doc
• Output: “air_quality_with_AQI.csv”
c. Model Training & Evaluation Stage
• Component: model_training_and_evaluation.ipynb
• Inputs: “air_quality_with_AQI.csv”
• Actions: feature engineering, classical and deep-learning model training
• Outputs: trained model artifacts, metrics, “model_comparison_results.csv”
d. Dependency Management
• requirements.txt – Python libraries (pandas, sklearn, tensorflow, etc.)
e. Final Reports
• model_comparison_results.csv
• any charts/notebook outputs
Architectural patterns & design principles
• Pipeline/Micro-batch: Each notebook is a pipeline stage.
• Separation of concerns: preprocessing logic separate from modeling logic.
• Reproducibility: requirements.txt captures environment.
• Modularity: easy to swap in/out different model algorithms.
Relevant technologies
• Python + Jupyter Notebooks
• pandas, numpy for ETL
• scikit-learn and deep-learning frameworks (e.g., TensorFlow/Keras) for modeling
• CSV as data interchange format