← Cheatsheets

Tags: machine-learning, ml, ai, supervised, unsupervised, scikit-learn, algorithms, beginnerLast updated: 2026-07-01

Machine Learning Cheatsheet

What Is Machine Learning?

Machine Learning (ML) is a approach to programming where instead of writing explicit rules, you show a computer examples and let it figure out the patterns itself.

Types of ML

When ML vs Traditional Programming

Do you know all the rules?
  └── No  → Do you have examples?
       └── Yes → Use supervised learning

Rule of thumb: If a human can't do the task in 1 second, ML is probably needed.

Core Terminology

Algorithm Selection

For Regression (predicting a number)

For Classification (predicting a category)

For Clustering (grouping without labels)

For Dimensionality Reduction

Key Algorithms Explained

The ML Pipeline

Raw Data → Clean → Explore → Engineer → Split → Train → Evaluate → Deploy → Monitor

Step-by-Step with scikit-learn

import pandas as pd
import numpy as np
from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import accuracy_score, classification_report

# 1. Load
df = pd.read_csv("data.csv")

# 2. Clean
df = df.dropna()
df["feature"] = df["feature"].astype(float)

# 3. Feature engineer
df["new_feature"] = df["col_a"] / (df["col_b"] + 1e-6)

# 4. Split
X = df.drop("target", axis=1)
y = df["target"]
X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.2, random_state=42
)

# 5. Train
model = RandomForestClassifier(n_estimators=100)
model.fit(X_train, y_train)

# 6. Evaluate
y_pred = model.predict(X_test)
print(f"Accuracy: {accuracy_score(y_test, y_pred):.2f}")
print(classification_report(y_test, y_pred))

Evaluation Metrics

Classification

Regression

Overfitting & How to Prevent It

Overfitting is when the model memorises training noise instead of learning the pattern. Great on training data, poor on new data.

Signs of Overfitting

• Training accuracy >> validation accuracy
• The gap grows as training progresses
• Model weights are very large
• Decision boundary is overly complex

How to Fix It

Python ML Stack

Your First ML Project (Iris)

from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import accuracy_score

iris = load_iris()
X, y = iris.data, iris.target
X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.2, random_state=42
)
model = RandomForestClassifier(n_estimators=100, random_state=42)
model.fit(X_train, y_train)
y_pred = model.predict(X_test)
print(f"Accuracy: {accuracy_score(y_test, y_pred):.2f}")

Common Pitfalls

Philosophy

• Start simple. Linear regression + baseline tells you if the problem is solvable.
• More data beats a better algorithm. Spend time getting more data before tuning.
• A deployed model beats a perfect one. 80% in production > 99% stuck in a notebook.
• Garbage in, garbage out. Bias in the data becomes bias in the model.
• ML is a tool, not a solution. Sometimes a simple if/else rule is all you need.

History & Milestones

The best model is the one you can actually use.