ML Workflow Overview

Let's think about it this way for a second

Data Collection/Cleaning is the very first step of the pipeline — real-world data usually comes with missing values, duplicates, and inconsistent formats, so it needs cleaning before you feed it to a model. Train/Test Split means splitting the data into two parts (training data to 'teach' the model, and test data to check the model's performance on unseen data, like an exam paper it hasn't seen before). Model Training means 'running' the algorithm on the training data so it can learn the patterns. Evaluation means measuring the model's accuracy/performance using metrics (covered in the Intermediate chapter).

Let's connect it to a real scenario

Running the house price prediction project (covered in the Project chapter) through this pipeline looks like: (1) collect house data (size, location, price), (2) clean up missing values/outliers, (3) split the data into 80% training / 20% test, (4) train an algorithm (Linear Regression) on the training data, (5) check accuracy against the test data, (6) deploy to production once you're satisfied — you'll see this same 6-stage flow come up again and again throughout the tutorial.

ML pipeline: Data Collection/Cleaning → Train/Test Split → Model Training → Evaluation → Deployment
Train/Test Split = splitting data into two parts (one for training, one held out as unseen test data)
Evaluation = measuring the model's performance using metrics

Let's look at it together

text

ML Workflow
============

1. Data Collection    -> gather raw data
2. Data Cleaning       -> handle missing values, remove duplicates
3. Train/Test Split    -> e.g. 80% train, 20% test
4. Model Training       -> algorithm learns from training data
5. Evaluation           -> measure performance on TEST data (unseen)
6. Deployment           -> put the trained model into production use

You should see

You'll be able to explain the 6 stages of the ML workflow in order.

5-minute try-it

For a Netflix movie rating prediction project, write out how this 6-stage pipeline would apply, using example data (user, movie, rating).

A quick word of caution

Don't assume the deployment step is 'automatically done once the model finishes training' — in production, you need to monitor the model's performance, and if the data patterns shift over time (data drift), you'll need to retrain the model.