📊 Machine Learning Basics with Python (Jupyter + Scikit-Learn)

Based on notes and examples from the following tutorial: 👉

---

🚀 Machine Learning Workflow

🔁 Standard ML Pipeline

ex:

---

📦 Required Libraries

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn import tree

---

⚙️ Environment Setup

Install Anaconda + Jupyter

After installing Anaconda, run:

jupyter notebook

➡️ This will open Jupyter in your browser.

---

📂 Importing Dataset

🔍 Source for datasets

• https://www.kaggle.com

---

📥 Load CSV File

import pandas as pd

df = pd.read_csv('vgsales.csv')
df.shape  # (rows, columns)

---

📊 Basic Exploration

df.describe()
df.values

---

⌨️ Jupyter Shortcuts

Shortcut	Description
H	Show all shortcuts
Shift + Tab	Show function documentation
Ctrl + /	Comment/uncomment

---

🎯 Example Project: Music Recommendation

📌 Goal

Predict music genre based on:

• Age
• Gender

---

📥 Load Data

import pandas as pd

music_data = pd.read_csv('music.csv')
music_data.head()

---

🧹 Data Cleaning

Typical steps:

• Remove duplicates
• Handle missing values
• Fix inconsistencies
• Drop irrelevant data

---

🧠 Prepare Data

Define Features (X) and Target (y)

X = music_data.drop(columns=['genre'])
y = music_data['genre']

---

🤖 Create & Train Model

from sklearn.tree import DecisionTreeClassifier

model = DecisionTreeClassifier()
model.fit(X, y)

---

🔮 Make Predictions

predictions = model.predict([[21, 1], [22, 0]])
print(predictions)

Expected Output:

['HipHop', 'Dance']

---

📏 Model Evaluation

Split Data

from sklearn.model_selection import train_test_split

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)

---

Measure Accuracy

from sklearn.metrics import accuracy_score

model.fit(X_train, y_train)
predictions = model.predict(X_test)

score = accuracy_score(y_test, predictions)
print(score)

---

💾 Persisting Models (Save & Load)

Save Model

import joblib

joblib.dump(model, 'music-recommender.joblib')

---

Load Model

model = joblib.load('music-recommender.joblib')
predictions = model.predict([[21, 1]])

---

🌳 Visualizing Decision Trees

Export Tree

from sklearn import tree

tree.export_graphviz(
    model,
    out_file='music-recommender.dot',
    feature_names=['age', 'gender'],
    class_names=sorted(y.unique()),
    label='all',
    rounded=True,
    filled=True
)

---

📌 Visualization Steps

1. Install Graphviz
2. Install VSCode extension: Dot / Graphviz
3. Open .dot file
4. Click Preview

---

📊 Final Output

The model creates a decision tree showing:

• Splits based on age & gender
• Gini impurity
• Class prediction (HipHop, Jazz, Classical, etc.)

---

🧠 Key Concepts Summary

Concept	Explanation
Features (X)	Input variables
Target (y)	Output variable
Train/Test Split	Prevent overfitting
Accuracy	Model performance metric
Model Persistence	Save trained models
Decision Tree	Interpretable ML model

---

🎯 Key Takeaways

• ML is a step-by-step pipeline
• Data preparation is critical
• Simple models like Decision Trees are powerful
• Always evaluate using test data
• Save models to reuse in production

---

📎 Resources

• Kaggle datasets: https://www.kaggle.com
• Tutorial video: https://youtu.be/7eh4d6sabA0

---

💡 Pro Tip (for interviews)

When explaining ML projects:

👉 Always structure like this:

Problem → Data → Model → Evaluation → Business Impact

---

Here is a clean, GitHub-ready explanation you can paste directly under your code:

---

🌳 tree.export_graphviz() – Explanation of Parameters

This function exports a trained Decision Tree model into a .dot file, which can be visualized using Graphviz.

tree.export_graphviz(
    model,
    out_file='music-recommender.dot',
    feature_names=['age', 'gender'],
    class_names=sorted(y.unique()),
    label='all',
    rounded=True,
    filled=True
)

---

📌 Parameters Breakdown

Parameter	Description
model	The trained Decision Tree model that you want to visualize.
out_file='music-recommender.dot'	Output file where the tree structure is saved. This .dot file can be rendered using Graphviz.
feature_names=['age', 'gender']	Names of input features used in the model. These appear in the tree nodes to make splits understandable.
class_names=sorted(y.unique())	Names of target classes (labels). y.unique() gets all unique values, and sorted() ensures consistent ordering.
label='all'	Displays detailed information in each node (e.g., Gini index, samples, value).
rounded=True	Rounds the corners of the boxes in the tree for better visual appearance.
filled=True	Colors the nodes based on class prediction (helps quickly interpret the model).

---

🎯 What This Produces

• A .dot file describing the tree structure

• Each node shows:

  • Split condition (e.g., age <= 30)
  • Gini impurity
  • Number of samples
  • Class distribution
  • Predicted class

---

📊 Example Node Output

age <= 30.5
gini = 0.5
samples = 10
value = [5, 3, 2]
class = HipHop

---

🚀 How to Visualize

1. Install Graphviz
2. Open .dot file in VSCode
3. Use Graphviz preview extension

---