🤖 Automated Data Analysis & ML Diagnostics Agent for Jupyter Notebook

Jupyter Notebook ke andar Exploratory Data Analysis (EDA) aur Machine Learning (ML) Model Diagnostics ke liye ek all-in-one AI assistant.

Aapko bas CSV file provide karni hai — agent aapse Step 1: First Column select karwayega, fir Step 2: Second Column, fir Step 3 mein 3 organized tabs ke andar 23 different plot types offer karega, aur ek click mein Interactive Visualizations (Plotly), Statistical & ML Insights, aur Python Code generate karke dega!

🌟 Naye Features (Multi-Plot & Direct Download)

1. ➕ Ek Se Zyada Plots Banayein (Active Plots Gallery)

• Pehla plot banne ke baad purana plot delete ya overwrite nahi hoga.
• Plot ke theek niche "➕ Create Another Plot (Naya Plot Banayein)" button milta hai.
• Click karte hi aap naye columns aur naya plot choose kar sakte hain — aur naya plot neeche append ho jayega!
• Is tarah aap ek hi notebook session mein kitne bhi plots (Plot #1, Plot #2, Plot #3...) stack karke compare kar sakte hain.

2. 📥 1-Click Direct Download (PNG & HTML)

• Har generated plot ke theek upar dedicated download buttons milte hain:
  • 📥 Download PNG (High-Res Image) — Publication-ready high resolution 2x scale image directly download hoti hai.
  • 🌐 Download Interactive HTML — Standalone browser file jisme zoom, pan, hover tooltips sab preserved rehte hain.
• Files aapke local exports/ folder mein bhi automatically save hoti hain.

🌟 Master Catalog of 23 Supported Plot Types

📊 Category A: Exploratory Data Analysis & Statistical Plots (16 Plots)

1. 📈 Line Plot (line_plot) — Trends, sequential progression, aur time-series.
2. 📊 Bar Plot (bar_plot) — Categories ka comparison (Mean / Sum aggregation ya counts).
3. 📋 Horizontal Bar Plot (horizontal_bar) — Categories ki clean horizontal ranking.
4. 📶 Grouped Bar Plot (grouped_bar) — Multiple categories ka side-by-side comparison.
5. 🧱 Stacked Bar Plot (stacked_bar) — Proportions aur percentage composition across groups.
6. 🥧 Pie Chart (pie_chart) — Circular percentage distribution.
7. 🍩 Donut Chart (donut_chart) — Modern center-hole donut chart.
8. 📊 Histogram (histogram) — Distribution spread, frequency bins aur KDE / box marginals.
9. ✨ Scatter Plot (scatter_plot) — Correlation, relationships, regression trendline aur $R^2$.
10. 🏔️ Area Plot (area_plot) — Cumulative filled volume under curve.
11. ⛰️ Stacked Area Plot (stacked_area_plot) — Multiple groups ka cumulative trend over sequence.
12. 🔥 Heatmap (heatmap) — 2D cross-tabulation frequency matrix aur 2D density heatmap.
13. 🌡️ Correlation Heatmap (correlation_heatmap) — Poore dataset ke numeric columns ka Pearson correlation matrix.
14. 📦 Box Plot (box_plot) — Quartiles (Q1, Median, Q3) aur IQR outlier markers.
15. 🎻 Violin Plot (violin_plot) — Kernel density estimation + embedded box plot.
16. 🔲 Pair Plot (pair_plot) — Multivariate pairwise scatter matrix across dataset features.

🤖 Category B: Machine Learning & Diagnostic Plots (7 Plots)

17. 🎯 Residual Plot (residual_plot) — Fitted values vs residuals ($y - \hat{y}$), homoscedasticity check, RMSE, MAE, $R^2$.
18. 🌲 Feature Importance Plot (feature_importance) — Random Forest model se top predictive features ki ranking.
19. 🔮 SHAP Summary Plot (shap_summary) — Explainable AI (SHAP) se features ka positive/negative impact.
20. 🧮 Confusion Matrix Heatmap (confusion_matrix) — Classification True vs Predicted labels, Accuracy, Precision, Recall, F1.
21. 📉 ROC Curve (roc_curve) — Receiver Operating Characteristic curve, False Positive vs True Positive rate, ROC-AUC score.
22. 🎯 Precision-Recall Curve (precision_recall_curve) — PR curve with Average Precision (AP) score, imbalanced classification ke liye best.
23. 📚 Learning Curve (learning_curve) — Sample size ke against Train vs Validation scores, Overfitting / Underfitting diagnosis.

📦 Python Library Installation

Yeh agent ab ek standard Python library (data-agent) ban chuka hai. Aap ise apne system environment mein install kar sakte hain taaki kisi bhi folder ya notebook se seedha import ho sake:

# Workspace folder mein jaakar install karein (editable mode):
pip install -e .

Ek baar install hone ke baad, aap apne computer par kahin se bhi bina kisi path setting ke seedha likh sakte hain:

from data_agent import AutoDataAgent

agent = AutoDataAgent()
agent.analyze("path/to/any_data.csv")

from data_agent import AutoDataAgent

# Agent initialize karein
agent = AutoDataAgent()

# CSV file provide karein
agent.analyze("sample_sales_data.csv")

• Step 1: 1-click se pehla column (X-axis / Target) select karein.
• Step 2: 1-click se doosra column (Y-axis / Feature) chunein ya Single Column / Dataset Mode click karein.
• Step 3: 3 Tabs mein se koi bhi plot select karein.
• Plot Output:
  • Upar Download PNG ya Download HTML par click karke save karein.
  • Niche ➕ Create Another Plot dabakar naya plot add karein!

2. Programmatic Export (Direct Code):

fig, code, insights = agent.quick_plot("Sales", "Profit", plot_id="scatter_plot")

# Save directly as PNG & HTML
png_path, _ = agent.export_plot(fig, filename="sales_vs_profit", format="png")
html_path, _ = agent.export_plot(fig, filename="sales_vs_profit", format="html")

print("Saved PNG:", png_path)
print("Saved HTML:", html_path)
fig.show()

🧪 Testing

Test suite ko run karein:

python3 test_agent.py

Sabhi 25 tests pass hote hain:

.........................
Ran 25 tests in 15.967s (OK)