The Skill Explained
What Is Machine Learning?
Machine learning (ML) is the part of AI where a computer learns patterns from examples instead of being programmed rule by rule. You show it data, such as past sales or labelled images, and it builds a model that makes predictions on new data it has never seen.
Why AI Employers Ask for It
Many business problems are predictions: will this customer leave, is this transaction fraud, how many units will we sell. Machine Learning answers them from data, and it is also the foundation under deep learning and language models. Employers ask for it because people who understand training, evaluation and overfitting build AI that works on new data, not only on the examples it has seen.
How It Works
In supervised learning, each example has inputs (features) and a known answer (the label). The algorithm adjusts its internal parameters to make its predictions match the labels. Common methods include linear and logistic regression, decision trees, random forests and gradient boosting. Unsupervised learning finds structure without labels, for example grouping similar customers.
The core discipline is honest evaluation. Data is split into a training set and a test set the model never sees while learning. A model that scores well on training data but poorly on test data has overfitted: it memorised instead of learning. Metrics are chosen for the problem, such as accuracy, precision and recall for classification or mean absolute error for numbers.
In practice most of the effort goes into data: collecting it, cleaning it, choosing features and checking that the data used for training matches the data the model will meet in real use.
What You Do With It at Work
- Frame a business question as a prediction with a clear target and metric.
- Prepare features from raw data and split it into training and test sets.
- Train and compare several models, for example with scikit-learn.
- Explain which features drive predictions to the people who will use them.
- Monitor a deployed model and retrain it when the data changes.
Tools and Terms You Will Meet
- Features and labels
- The inputs a model learns from, and the answers it learns to predict.
- Overfitting
- When a model memorises training data and fails on new data.
- Cross-validation
- Testing a model on several different splits of the data for a fairer score.
- scikit-learn
- The standard Python library for classic machine learning.
- Gradient boosting
- A strong method for tables of data that combines many small decision trees.
Industry Example: Pharma
A manufacturing team trains a model on past batch data (temperatures, pressures, raw material lots) to predict which batches risk failing a quality test, so operators can act before the batch is finished.
A Common Mix-Up
Machine Learning is not the same as AI. AI is the broad goal of machines doing tasks that need intelligence; Machine Learning is one way to get there, and today the main one.
How Deep Do You Need to Go?
- Know itExplain training versus testing, overfitting and what a model's metric means.
- Use itBuild, evaluate and compare models on real data with scikit-learn, and explain errors.
- Master itDesign features and validation for messy real problems, and run models in production.