How guymizrahi1/nitrogen-status-by-spectroscopy works
Project Type Analysis
This is a machine learning project with multiple model implementations (RF, XGBoost, PLSR) for spectroscopy-based nitrogen status identification. The structure suggests a data processing pipeline with various modeling approaches.
Main Components
A. Data Layer:
• Raw Data Storage (datasets directory)
• Training data (train_data.parquet, train_data_plsr.parquet)
• Validation data (validation_data.parquet, validation_data_plsr.parquet)
• Test data (test_data.parquet, test_data_plsr.parquet)
B. Data Preparation Layer:
• Data preprocessing (data_preparation.py)
• Configuration settings (constants_config.py)
• Dataset handling (baseline_for_training/Dataset.py)
C. Model Layer:
Three main modeling approaches:
PLSR (Partial Least Squares Regression)
• Implementation (PLSR/PLSR.py, PLSR/PLSR_class.py)
• Model storage (PLSR/models)
Random Forest
• Implementation (RF/RF.py, RF/rf_main.py)
• Model storage (RF/models)
XGBoost
• Implementation (XGBoost/xgboost_multioutput_implement.py)
• Utilities (XGBoost/utils.py)
• Model storage (XGBoost/models)
D. Evaluation & Output Layer:
• Model outputs (PLSR/outputs)
• Performance metrics storage
• Visualization components (PLSR/Plots, RF/plots, XGBoost/figures)
Specific Visualization Elements
a. Use rectangles for:
• Data storage components
• Processing modules
• Model implementations
b. Use diamonds for:
• Decision points
• Validation checkpoints
• Model selection points
• Database/Storage components
• Model persistence
• Data flow direction
• Process sequence
• Model training flow
• Evaluation feedback