How dreamywarrior/comparative-study-of-stellar-classification works
Project type and purpose
• This is a research-oriented machine-learning classification project implemented as Jupyter notebooks. It focuses on classifying celestial objects (stars, galaxies, quasars) using Sloan Digital Sky Survey (SDSS) DR-16 and DR-17 data.
• The goal is to compare five self-sufficient ML algorithms, evaluate their accuracies on each dataset, and identify the top performer.
File structure
Top-level files:
• “CoSSC_DR16.ipynb” – notebook for DR-16 data
• “CoSSC_DR17.ipynb” – notebook for DR-17 data
• “CoSSC – Comparative Study of Stellar Classification (2)-2.pptx” – presentation of results
• “README.md” – project overview and data & methodology links
There are no separate “src”, “lib”, or “backend/frontend” directories. All processing, modeling, and visualization steps live inside the two notebooks.
README for additional insights
• Confirms use of SDSS DR-16 & DR-17 datasets
• Emphasizes a comparative study of five ML algorithms
• Describes end-goals: efficiency, accuracy, and trend analysis
Main components
a. Data Sources
• SDSS DR-16 (external dataset, raw FITS/CSV files)
• SDSS DR-17 (external dataset)
b. Data Ingestion & Preprocessing
• Notebook cells that load raw data, filter, clean, and transform into feature tables
c. Feature Engineering
• Steps in notebook that select or derive input features (e.g., magnitudes, redshifts)
d. Model Training
• Five parallel algorithm components (e.g., Random Forest, SVM, Neural Network, etc.)
e. Evaluation & Comparison
• Common evaluation module computing accuracy, precision, recall for each model
f. Result Aggregation & Visualization
• Generation of plots, tables, and summary statistics
g. Presentation Layer
• PowerPoint file capturing key findings
Relationships and interactions
• Preprocessing feeds into Feature Engineering
• Features split to the five Model Training components in parallel
• Each model outputs to the Evaluation & Comparison component
• Comparison flows into Result Aggregation & Visualization
• Final outputs link to the Presentation Layer
Architectural patterns & design principles
• Modular pipeline: clear separation between data acquisition, preprocessing, modeling, and reporting
• Parallel model training: five self-contained algorithm modules
• Reproducibility: notebooks capture each pipeline stage sequentially
Relevant technologies
• Python & Jupyter Notebooks
• pandas, NumPy for data handling
• scikit-learn (or equivalent) for ML algorithms
• Matplotlib/seaborn for visualization
• SDSS data formats (FITS/CSV)
Detail
• Logical step from raw data ingestion through final PPT outputs.
• Both DR-16 and DR-17 flows side by side, converging at the comparison stage.