notebookpkg 1.4.0

A package manager for Jupyter notebook templates

notebookpkg

A notebook template manager for ML students.
One command installs a ready-to-run Jupyter notebook — already wired to your dataset, with your column names, your target, and your drop columns injected automatically.

No more writing the same boilerplate for every assignment. Just pick a template, point it to your CSV, and open Jupyter.

---

Installation

pip install notebookpkg

Requirements: Python 3.7+, pandas, scikit-learn, matplotlib, seaborn, nbformat, click

---

How It Works

1. You run one command with your CSV file
2. The tool reads your dataset and detects all column names and types
3. It injects your dataset path, column names, target column, and drop columns into the template
4. A .ipynb file is created in your current folder
5. Open it in Jupyter and run all cells — everything is pre-filled

---

Quick Start

# Step 1: See all available templates
notebookpkg list

# Step 2: Install a template for your CSV
notebookpkg install linear-regression --dataset Salary_Data.csv --target Salary

# Step 3: Open the notebook
jupyter notebook linear-regression_notebook.ipynb

---

Commands

notebookpkg list

Lists all available templates with their descriptions.

notebookpkg list

Output:

📦 Available Templates:

  decision-tree                       Decision Tree: criterion=entropy, max_depth=5, plot_tree, accuracy, report
  eda-basic                           Basic EDA: head, shape, info, describe, nulls, dtypes, nunique
  eda-full                            Full EDA: visual + outliers, skewness, duplicates, value counts
  eda-visual                          Visual EDA: pairplot, heatmap, distributions
  kmeans-clustering                   KMeans Clustering: StandardScaler, elbow method, silhouette score, cluster plot
  knn-classifier                      KNN Classifier: StandardScaler, fit, accuracy, confusion matrix, report
  lasso-ridge                         Linear + Lasso + Ridge Regression with StandardScaler and coefficient plots
  linear-regression                   Linear Regression: EDA, fit, predict, visualize, MSE, R²
  logistic-regression                 Logistic Regression: StandardScaler, fit, accuracy, confusion matrix, report
  multi-model-compare                 LR + KNN + Naive Bayes on same dataset with accuracy comparison
  naive-bayes                         Gaussian Naive Bayes: StandardScaler, fit, accuracy, confusion matrix heatmap
  polynomial-regression               Polynomial Regression: PolynomialFeatures, smooth curve plot, MSE, R²
  random-forest-classifier            Random Forest Classifier: model1, accuracy, confusion matrix, feature importance
  random-forest-regressor             Random Forest Regressor: RFR, fit, MSE, R², Actual vs Predicted scatter
  svm-classifier                      SVM: Linear kernel, then RBF kernel with AgeSalary feature engineering

---

notebookpkg install

Installs a template wired to your dataset.

notebookpkg install <template-name> --dataset <path-to-csv> [options]

All options:

Option	Required	Default	Description
--dataset	Yes	—	Path to your CSV file
--target	No	Last column	Target/label column name
--drop	No	None	Columns to drop, comma-separated
--degree	No	2	Polynomial degree — only for polynomial-regression
--clusters	No	3	Number of clusters — only for kmeans-clustering
--output	No	<template>_notebook.ipynb	Custom output filename

---

---

notebookpkg syntax

Prints the complete code of a template — every cell in order — directly in your terminal. Use this to preview exactly what will be generated before installing.

notebookpkg syntax <template-name>

Example:

notebookpkg syntax logistic-regression

Output:

============================================================
  Template : logistic-regression
  Logistic Regression: StandardScaler, fit, accuracy, confusion matrix, report
  Total cells: 16
============================================================

── Cell 1 ──────────────────────────────────────────────────
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import accuracy_score, confusion_matrix, classification_report

── Cell 2 ──────────────────────────────────────────────────
df = pd.read_csv('{{DATASET_PATH}}')
df.head()

── Cell 3 ──────────────────────────────────────────────────
{{DROP_CODE}}

... (all remaining cells shown in full)

============================================================
  Install this template:
  notebookpkg install logistic-regression --dataset yourdata.csv
============================================================

You can run syntax for any of the 15 templates:

notebookpkg syntax eda-basic
notebookpkg syntax eda-visual
notebookpkg syntax eda-full
notebookpkg syntax linear-regression
notebookpkg syntax polynomial-regression
notebookpkg syntax logistic-regression
notebookpkg syntax knn-classifier
notebookpkg syntax naive-bayes
notebookpkg syntax lasso-ridge
notebookpkg syntax decision-tree
notebookpkg syntax random-forest-regressor
notebookpkg syntax random-forest-classifier
notebookpkg syntax svm-classifier
notebookpkg syntax kmeans-clustering
notebookpkg syntax multi-model-compare

---

Templates

EDA Templates

eda-basic

Basic Exploratory Data Analysis. Covers the essential checks every notebook needs.

Cells generated:

1. Imports
2. df.read_csv() + df.head()
3. Drop columns cell (optional)
4. df.shape
5. df.info()
6. df.describe()
7. df.isnull().sum()
8. df.dtypes
9. df.nunique()

notebookpkg install eda-basic --dataset data.csv

---

eda-visual

EDA with all key visualizations.

Cells generated: Everything in eda-basic, plus:

• sns.pairplot(df)
• Correlation heatmap (df.corr() + sns.heatmap())
• Histogram for each numeric column

notebookpkg install eda-visual --dataset data.csv

---

eda-full

Complete EDA including outlier detection and categorical analysis.

Cells generated: Everything in eda-visual, plus:

• df.duplicated().sum()
• Boxplot for each numeric column
• Skewness: df.skew(numeric_only=True)
• IQR outlier count for each numeric column
• value_counts() for each categorical column

notebookpkg install eda-full --dataset data.csv

---

Regression Templates

linear-regression

Standard Linear Regression pipeline on your CSV.

Cells generated:

1. Imports
2. Load dataset + head
3. Drop columns cell
4. shape, info, describe, isnull
5. pairplot
6. Correlation heatmap
7. X / y split (iloc)
8. train_test_split (test_size=0.2, random_state=0)
9. regressor = LinearRegression() + fit
10. Predict
11. Visualize training data (scatter + regression line)
12. Visualize testing data
13. Coefficient and intercept
14. MSE
15. R²

notebookpkg install linear-regression --dataset Salary_Data.csv --target Salary

---

polynomial-regression

Polynomial Regression with smooth curve visualization.

Cells generated:

1. Imports (includes PolynomialFeatures)
2. Load dataset + head
3. Drop columns cell
4. info, describe, pairplot, heatmap
5. X / y split
6. PolynomialFeatures(degree=N) + transform
7. train_test_split
8. plr = LinearRegression() + fit
9. Smooth curve plot using X_gride
10. Predict
11. MSE
12. R²

notebookpkg install polynomial-regression --dataset hw.csv --target Price
notebookpkg install polynomial-regression --dataset hw.csv --target Price --degree 3

---

lasso-ridge

Linear Regression + Lasso + Ridge, all on the same dataset with comparison.

Cells generated:

1. Imports
2. Load + EDA (info, describe, columns, shape)
3. Drop columns cell
4. X / y split
5. train_test_split
6. StandardScaler
7. Linear Regression (lm) + coefficient barh plot
8. Lasso (alpha=0.1) + MSE + R² + coefficient barh plot
9. Ridge (alpha=0.1) + MSE + R²

notebookpkg install lasso-ridge --dataset BostonHousing.csv --target medv

---

Classification Templates

logistic-regression

Logistic Regression with StandardScaler.

Cells generated:

1. Imports
2. Load dataset + head
3. Drop columns cell
4. shape, info, describe, isnull
5. Correlation heatmap
6. X / y split
7. train_test_split (test_size=0.3, random_state=0)
8. sc = StandardScaler() + fit_transform / transform
9. lr = LogisticRegression() + fit
10. Predict
11. Accuracy score
12. Confusion matrix
13. Classification report

notebookpkg install logistic-regression --dataset Day5.csv --target Purchased
notebookpkg install logistic-regression --dataset Day5.csv --target Purchased --drop "User ID,Gender"

---

knn-classifier

K-Nearest Neighbors Classifier with StandardScaler.

Cells generated:

1. Imports
2. Load dataset + head
3. Drop columns cell
4. shape, info, describe, isnull, duplicated
5. Correlation heatmap + pairplot
6. X / y split
7. train_test_split (test_size=0.2, random_state=42)
8. StandardScaler
9. knn = KNeighborsClassifier() + fit
10. Predict
11. Accuracy, confusion matrix, classification report

notebookpkg install knn-classifier --dataset Day5.csv --target Purchased

---

naive-bayes

Gaussian Naive Bayes with StandardScaler and confusion matrix heatmap.

Cells generated:

1. Imports
2. Load + shape, describe, isnull
3. Drop columns cell
4. Correlation heatmap
5. X / y split
6. train_test_split with stratify=y
7. StandardScaler (fit only on train)
8. nb = GaussianNB() + fit
9. Predict
10. Accuracy
11. Classification report
12. Confusion matrix as sns.heatmap

notebookpkg install naive-bayes --dataset Day5.csv --target Purchased

---

decision-tree

Decision Tree Classifier with tree visualization.

Cells generated:

1. Imports (includes from sklearn import tree)
2. Load + EDA
3. Drop columns cell
4. Distribution plot + heatmap + pairplot
5. X / y split
6. train_test_split
7. StandardScaler
8. DecisionTreeClassifier(criterion='entropy', max_depth=5, random_state=0)
9. Predict
10. Accuracy score
11. Confusion matrix
12. Classification report
13. tree.plot_tree() — full visual tree diagram

notebookpkg install decision-tree --dataset SNP.csv --target Purchased

---

svm-classifier

SVM with both Linear and RBF kernels, plus feature engineering.

Cells generated:

1. Imports (includes SVC)
2. Load + EDA (info, describe, isnull, value_counts)
3. Drop columns cell
4. Scatter plot of features
5. X / y split
6. train_test_split
7. StandardScaler
8. model = SVC(kernel='linear') + fit + predict + accuracy + CM + heatmap
9. Feature engineering: df['AgeSalary'] = df['Age'] * df['EstimatedSalary']
10. Re-split with new feature
11. model1 = SVC(kernel='rbf') + fit + predict + accuracy + CM + heatmap

notebookpkg install svm-classifier --dataset SNP.csv --target Purchased

---

multi-model-compare

Runs Logistic Regression, KNN, and Naive Bayes on the same dataset and compares accuracy.

Cells generated:

1. Imports
2. Load + EDA
3. Drop columns cell
4. X / y split
5. train_test_split
6. model_lr = LogisticRegression() → fit → predict → accuracy → report
7. model_knn = KNeighborsClassifier() → fit → predict → accuracy → report
8. model_nb = GaussianNB() → fit → predict → accuracy → report
9. Comparison dict with all three accuracy scores printed together

notebookpkg install multi-model-compare --dataset Day5.csv --target Purchased

---

Ensemble Templates

random-forest-regressor

Random Forest Regressor with actual vs predicted scatter plot.

Cells generated:

1. Imports
2. Load + isnull, duplicated, info, describe
3. Drop columns cell
4. Correlation heatmap
5. X / y split
6. train_test_split (test_size=0.2, random_state=42)
7. RFR = RandomForestRegressor(n_estimators=100, random_state=42) + fit
8. Predict
9. MSE
10. R²
11. Scatter plot: Actual vs Predicted

notebookpkg install random-forest-regressor --dataset housing.csv --target Price

---

random-forest-classifier

Random Forest Classifier with feature importance bar chart.

Cells generated:

1. Imports
2. Load + EDA
3. Drop columns cell
4. X / y split
5. train_test_split
6. StandardScaler
7. model1 = RandomForestClassifier(n_estimators=100, random_state=42) + fit
8. Predict
9. Accuracy
10. Classification report
11. Confusion matrix heatmap
12. Feature importance: model1.feature_importances_
13. Bar chart of feature importance

notebookpkg install random-forest-classifier --dataset iris.csv --target species

---

Clustering Templates

kmeans-clustering

KMeans Clustering with elbow method and silhouette score. No target column needed.

Cells generated:

1. Imports (includes KMeans, silhouette_score)
2. Load + shape, info, describe, isnull, duplicated
3. Drop columns cell
4. pairplot
5. Correlation heatmap
6. StandardScaler on numeric columns
7. Elbow method loop (k=1 to 9) + inertia plot
8. KMeans(n_clusters=N) + fit
9. Cluster labels added to df
10. Cluster scatter plot with centroids marked in red
11. Silhouette score

notebookpkg install kmeans-clustering --dataset Mall_Customers.csv
notebookpkg install kmeans-clustering --dataset Mall_Customers.csv --clusters 5

---

The --drop Option

Many real datasets have ID columns, name columns, or other columns that should not go into the model. Use --drop to remove them before anything is processed.

With --drop, the generated notebook gets:

df = df.drop(columns=['User ID', 'Gender'], axis=1)
df.head()

Without --drop, the cell appears as a comment so you can still do it manually:

# No columns dropped
# To drop columns use: df = df.drop(columns=['col1','col2'], axis=1)

The profiler also respects the drop — column detection for NUMERIC_COLS, CAT_COLS, and FEATURE_COLS all happen after the drop, so the rest of the notebook is consistent.

# Drop one column
notebookpkg install knn-classifier --dataset Day5.csv --target Purchased --drop "User ID"

# Drop multiple columns
notebookpkg install logistic-regression --dataset Day5.csv --target Purchased --drop "User ID,Gender"

---

All Usage Examples

# ── EDA ──────────────────────────────────────────────────────────────────
notebookpkg install eda-basic   --dataset data.csv
notebookpkg install eda-visual  --dataset data.csv
notebookpkg install eda-full    --dataset data.csv

# ── Regression ───────────────────────────────────────────────────────────
notebookpkg install linear-regression      --dataset Salary_Data.csv --target Salary
notebookpkg install polynomial-regression  --dataset hw.csv --target Price --degree 3
notebookpkg install lasso-ridge            --dataset BostonHousing.csv --target medv

# ── Classification ────────────────────────────────────────────────────────
notebookpkg install logistic-regression    --dataset Day5.csv --target Purchased
notebookpkg install knn-classifier         --dataset Day5.csv --target Purchased
notebookpkg install naive-bayes            --dataset Day5.csv --target Purchased
notebookpkg install decision-tree          --dataset SNP.csv  --target Purchased
notebookpkg install svm-classifier         --dataset SNP.csv  --target Purchased
notebookpkg install multi-model-compare    --dataset Day5.csv --target Purchased

# ── Ensemble ─────────────────────────────────────────────────────────────
notebookpkg install random-forest-regressor   --dataset housing.csv --target Price
notebookpkg install random-forest-classifier  --dataset iris.csv    --target species

# ── Clustering ────────────────────────────────────────────────────────────
notebookpkg install kmeans-clustering --dataset Mall_Customers.csv
notebookpkg install kmeans-clustering --dataset Mall_Customers.csv --clusters 5

# ── With drop ─────────────────────────────────────────────────────────────
notebookpkg install logistic-regression --dataset Day5.csv --target Purchased --drop "User ID,Gender"

# ── Custom output filename ────────────────────────────────────────────────
notebookpkg install linear-regression --dataset Salary_Data.csv --target Salary --output my_analysis.ipynb

---

Project Structure

notebookpkg/
├── notebookpkg/
│   ├── cli.py          # CLI commands: install, list
│   ├── profiler.py     # Reads CSV, detects column types
│   ├── injector.py     # Replaces tokens in notebook cells
│   ├── registry.py     # Finds templates by name
│   └── templates/
│       ├── eda-basic/
│       ├── eda-visual/
│       ├── eda-full/
│       ├── linear-regression/
│       ├── polynomial-regression/
│       ├── logistic-regression/
│       ├── knn-classifier/
│       ├── naive-bayes/
│       ├── lasso-ridge/
│       ├── decision-tree/
│       ├── random-forest-regressor/
│       ├── random-forest-classifier/
│       ├── svm-classifier/
│       ├── kmeans-clustering/
│       └── multi-model-compare/
├── build_templates.py  # Regenerates all .ipynb template files
├── setup.py
├── MANIFEST.in
└── README.md

Each template folder contains:

• template.ipynb — the notebook with {{TOKEN}} placeholders
• meta.json — name, description, and whether a target column is needed

---

Dependencies

pandas
numpy
scikit-learn
matplotlib
seaborn
nbformat
click

These are installed automatically when you run pip install notebookpkg.

---

Author

Priyansu Pattanaik
B.Tech — Electronics & Telecommunication
PG Diploma in AI — CDAC Kharghar
[e-mail]

---

License

MIT License. Free to use, modify, and distribute.