How vermashaurya/data-analysis works
Project Type & Purpose
• This is a monorepo of standalone data-analysis tools written in Python.
• It provides:
• Image annotation (YOLO-based object detection)
• Classification algorithms (ID3 Decision Tree)
• Performance metrics for linear & logistic regression
• Time-series analysis (AutoRegressive models)
• There is no traditional frontend/backend; each folder is a self-contained component that ingests data, processes it, and produces output (plots, CSVs, annotated images, dashboards).
File Structure & High‐Level Components
• Top‐level directories = major modules:
• Classification/DecisionTree
• Image-annotation
• Performance-Metrics/Linear-Regression
• Performance-Metrics/LogisticRegression
• TimeSeries-Analysis
• Common patterns: each module contains:
• Data inputs (CSV, cfg, weights, names)
• Processing scripts or notebooks (.py, .ipynb)
• Outputs (plots, .dot files, .pdf, Streamlit UI)
• Config/build files:
• model weights & config (yolov3.weights, .cfg, coco.names)
• dependencies called out in README (numpy, pandas, opencv, streamlit)
Key Architecture Insights from README
• Dependencies per module:
• Image-annotation: OpenCV, NumPy, Pillow, YOLO assets
• Linear/Logistic Regression: numpy, pandas, Streamlit (for logistic)
• Time-Series: pandas (inferred), statsmodels or similar
• Execution modes:
• CLI scripts (main.py, regression.py, logisticregression.py)
• Jupyter notebooks for exploratory analysis
• Streamlit app for interactive logistic regression metrics
• classification CSVs
• datasets for regression & time series
• images and YOLO config/weights
2. Preprocessing Layer (rectangles)
• image loader & annotator (main.py)
• data cleaners embedded in notebooks/scripts
3. Modeling Engines (rectangles)
• Decision Tree (decisiontree.py)
• Linear Regression (linearregression.py)
• Logistic Regression (logisticregression.py)
• AutoRegressive Model (notebook)
4. Evaluation & Metrics (rectangles)
• MSE, RMSE, MAE, R², Adj-R² (Linear)
• MSLE, MAPE (Logistic)
• Jupyter Notebooks
• Streamlit dashboard
• YOLO weight files (downloaded)
• pip-installed Python libraries
• Data → Preprocessing
• Preprocessing → Modeling
• Modeling → Evaluation Metrics
• Evaluation Metrics → Visualization/UI
• Visualization/UI → End user/Analyst
c. Architectural Patterns & Principles
• Modular, folder-per-component
• Separation of concerns: data ingestion vs. model vs. UI
• Reusable utilities (could be common library down the road)
d. Key Technologies & Frameworks
• Python scripts & Jupyter notebooks
• OpenCV, NumPy, pandas, statsmodels (inferred)
• Streamlit for interactive dashboards
• YOLOv3 object detection framework