Analytics Modeling From Scratch

Formula-first walkthroughs for building intuition, code, and judgment

Every walkthrough builds a modeling idea from the ground up. You see the question the method answers, the formula it relies on, the R code that makes it concrete, and the judgment calls that matter when the math meets messy data.

Get Started

Choose the path that matches how you want to learn today.

Goals

Start from the modeling job: classify, forecast, optimize, transform, or validate.

Topics

Jump directly to a formula-first walkthrough for a specific method.

Practice

Use focused drills to check recall, costs, and decision logic.

The Modeling Loop

Each page follows the same rhythm so you can move from intuition to implementation without losing the thread.

Question

What job is the method trying to do?

Formula

Which mathematical object carries the idea?

Code

How does the formula become executable R?

Judgment

Where can the model mislead you?

Choose by Goal

Pick the job you are trying to do, then jump straight into the walkthrough that matches the modeling problem.

Classify observations

Known labels, decision boundaries, and neighbor votes.

SVM KNN

Check model performance

Estimate how well a model handles data it did not see while training.

Cross-Validation

Group unlabeled data

Find structure from distances, centroids, and repeated assignment.

K-Means

Detect process shifts

Accumulate evidence until a small change becomes visible.

CUSUM

Forecast future values

Work through trend, memory, and volatility in ordered data.

Time Series

Predict or explain outcomes

Connect variables to outcomes with linear and logistic relationships.

Regression

Transform features

Reduce dimensions or reshape skewed variables before modeling.

PCA & Box-Cox

Build interpretable rules

Create decision paths from recursive splits and pruning.

CART

Choose variables

Balance signal, bias, variance, and shrinkage.

Variable Selection

Design data collection

Compare treatments intentionally with tests, factors, and adaptive choices.

Design of Experiments

Simulate uncertainty

Model queues, distributions, and steady-state behavior.

Simulation & Markov

Handle missing values

Reason about missingness mechanisms before choosing an imputation strategy.

Missing Data

Formulate decisions

Turn objectives, constraints, and variables into solvable models.

Optimization

Recognize advanced families

Build a practical map of methods you will see beyond the basics.

Advanced Topics

Topic Map

Use the full topic map when you want the complete formula-first walkthrough for a method.

Classification

Validation

Clustering

Change Detection

Time Series

Regression

Transformations

Trees

Variable Selection

Design of Experiments

Simulation & Markov

Missing Data

Optimization

Advanced Topics