🎥Video 35 min2 / 8

Learning types: Supervised and Unsupervised

Learning Types: Supervised and Unsupervised

Machine learning has three main types, each suited to a different kind of problem. Understanding the difference is the first step toward choosing the right algorithm.

1. Supervised Learning

Definition: The model learns from labeled data — every example has a known correct answer.

When to use: When you have data with known correct answers.

Two subtypes:

Regression: Output is a continuous number (house price, temperature)
Classification: Output is a category (spam/not spam, cat/dog)

2. Unsupervised Learning

Definition: The model finds patterns in unlabeled data — no correct answers provided upfront.

When to use: When you want to discover hidden patterns in data.

Subtypes:

Clustering: Group similar items (customer segments)
Dimensionality Reduction: Compress features (PCA, t-SNE)
Anomaly Detection: Find outliers (fraud detection)

3. Reinforcement Learning

Definition: An agent learns by trial and error to maximize a reward signal.

Applications: AlphaGo, robotics, game AI, data center optimization.

How to Choose

Do you have labeled data?
├── Yes → Supervised
│   └── Numeric or category output?
│       ├── Numeric → Regression
│       └── Category → Classification
└── No → Unsupervised
    └── Group or compress?
        ├── Group → Clustering
        └── Compress → Dimensionality Reduction

python

import numpy as np
import matplotlib
matplotlib.use("Agg")
import matplotlib.pyplot as plt
from sklearn.datasets import make_classification, make_blobs, make_regression
from sklearn.linear_model import LinearRegression, LogisticRegression
from sklearn.cluster import KMeans
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score, mean_squared_error

print("=" * 55)
print("مقارنة: Supervised vs Unsupervised Learning")
print("=" * 55)

# ─────────────────────────────────────────
# 1. Supervised — Regression
# ─────────────────────────────────────────
print("\n1. SUPERVISED — Regression (التنبؤ بالأسعار)")
X_reg, y_reg = make_regression(n_samples=200, n_features=1,
                                noise=30, random_state=42)
X_train, X_test, y_train, y_test = train_test_split(
    X_reg, y_reg, test_size=0.2, random_state=42
)
reg_model = LinearRegression()
reg_model.fit(X_train, y_train)
y_pred = reg_model.predict(X_test)
rmse = np.sqrt(mean_squared_error(y_test, y_pred))
print(f"   RMSE = {rmse:.2f}  (كلما أصغر، أفضل)")

# ─────────────────────────────────────────
# 2. Supervised — Classification
# ─────────────────────────────────────────
print("\n2. SUPERVISED — Classification (تصنيف)")
X_clf, y_clf = make_classification(n_samples=300, n_features=2,
                                   n_redundant=0, random_state=42)
X_train, X_test, y_train, y_test = train_test_split(
    X_clf, y_clf, test_size=0.2, random_state=42
)
clf_model = LogisticRegression()
clf_model.fit(X_train, y_train)
accuracy = accuracy_score(y_test, clf_model.predict(X_test))
print(f"   Accuracy = {accuracy:.1%}")

# ─────────────────────────────────────────
# 3. Unsupervised — Clustering
# ─────────────────────────────────────────
print("\n3. UNSUPERVISED — Clustering (تجميع بدون تسميات)")
# بيانات بدون labels — النموذج يكتشف المجموعات بنفسه
X_cluster, _ = make_blobs(n_samples=300, centers=4,
                           cluster_std=0.8, random_state=42)
kmeans = KMeans(n_clusters=4, random_state=42, n_init="auto")
kmeans.fit(X_cluster)
labels = kmeans.labels_

# عدد العناصر في كل مجموعة
unique, counts = np.unique(labels, return_counts=True)
print("   المجموعات المكتشفة:")
for g, c in zip(unique, counts):
    print(f"     مجموعة {g}: {c} عنصر")

print("\n" + "=" * 55)
print("ملخص:")
print("  Regression  → يتنبأ بأرقام مستمرة (مع labels)")
print("  Classification → يصنّف لفئات (مع labels)")
print("  Clustering  → يجمّع بدون labels")
print("=" * 55)

PreviousNext

Learning Types: Supervised and Unsupervised

Machine learning has three main types, each suited to a different kind of problem. Understanding the difference is the first step toward choosing the right algorithm.

1. Supervised Learning

Definition: The model learns from labeled data — every example has a known correct answer.

When to use: When you have data with known correct answers.

Two subtypes:

Regression: Output is a continuous number (house price, temperature)

Classification: Output is a category (spam/not spam, cat/dog)

2. Unsupervised Learning

Definition: The model finds patterns in unlabeled data — no correct answers provided upfront.

When to use: When you want to discover hidden patterns in data.

Subtypes:

Clustering: Group similar items (customer segments)

Dimensionality Reduction: Compress features (PCA, t-SNE)

Anomaly Detection: Find outliers (fraud detection)

3. Reinforcement Learning

Definition: An agent learns by trial and error to maximize a reward signal.

Applications: AlphaGo, robotics, game AI, data center optimization.

How to Choose

Do you have labeled data? ├── Yes → Supervised │ └── Numeric or category output? │ ├── Numeric → Regression │ └── Category → Classification └── No → Unsupervised └── Group or compress? ├── Group → Clustering └── Compress → Dimensionality Reduction

import numpy as np import matplotlib matplotlib.use("Agg") import matplotlib.pyplot as plt from sklearn.datasets import make_classification, make_blobs, make_regression from sklearn.linear_model import LinearRegression, LogisticRegression from sklearn.cluster import KMeans from sklearn.model_selection import train_test_split from sklearn.metrics import accuracy_score, mean_squared_error print("=" * 55) print("مقارنة: Supervised vs Unsupervised Learning") print("=" * 55) # ───────────────────────────────────────── # 1. Supervised — Regression # ───────────────────────────────────────── print("\n1. SUPERVISED — Regression (التنبؤ بالأسعار)") X_reg, y_reg = make_regression(n_samples=200, n_features=1, noise=30, random_state=42) X_train, X_test, y_train, y_test = train_test_split( X_reg, y_reg, test_size=0.2, random_state=42 ) reg_model = LinearRegression() reg_model.fit(X_train, y_train) y_pred = reg_model.predict(X_test) rmse = np.sqrt(mean_squared_error(y_test, y_pred)) print(f" RMSE = {rmse:.2f} (كلما أصغر، أفضل)") # ───────────────────────────────────────── # 2. Supervised — Classification # ───────────────────────────────────────── print("\n2. SUPERVISED — Classification (تصنيف)") X_clf, y_clf = make_classification(n_samples=300, n_features=2, n_redundant=0, random_state=42) X_train, X_test, y_train, y_test = train_test_split( X_clf, y_clf, test_size=0.2, random_state=42 ) clf_model = LogisticRegression() clf_model.fit(X_train, y_train) accuracy = accuracy_score(y_test, clf_model.predict(X_test)) print(f" Accuracy = {accuracy:.1%}") # ───────────────────────────────────────── # 3. Unsupervised — Clustering # ───────────────────────────────────────── print("\n3. UNSUPERVISED — Clustering (تجميع بدون تسميات)") # بيانات بدون labels — النموذج يكتشف المجموعات بنفسه X_cluster, _ = make_blobs(n_samples=300, centers=4, cluster_std=0.8, random_state=42) kmeans = KMeans(n_clusters=4, random_state=42, n_init="auto") kmeans.fit(X_cluster) labels = kmeans.labels_ # عدد العناصر في كل مجموعة unique, counts = np.unique(labels, return_counts=True) print(" المجموعات المكتشفة:") for g, c in zip(unique, counts): print(f" مجموعة {g}: {c} عنصر") print("\n" + "=" * 55) print("ملخص:") print(" Regression → يتنبأ بأرقام مستمرة (مع labels)") print(" Classification → يصنّف لفئات (مع labels)") print(" Clustering → يجمّع بدون labels") print("=" * 55)