Machine Learning

Machine learning is the field of building systems that improve at a task by learning statistical patterns from data instead of following rules written by hand. A developer defines the inputs, the target, and a loss function; an optimisation procedure then fits parameters that minimise error on training examples, and the resulting model is judged on data it has never seen. The main families are supervised learning, which maps labelled inputs to outputs for tasks like fraud detection and demand forecasting; unsupervised learning, which finds structure such as customer segments without labels; and reinforcement learning, which learns from rewards. Deep learning is a subset that uses many-layered neural networks, but much production work still runs on simpler methods, including XGBoost and LightGBM for tabular data and logistic regression where interpretability is required. The hard part is rarely the algorithm. Data quality, leakage between training and test sets, class imbalance, and drift after deployment cause more failures than model choice does, which is why teams invest in evaluation and monitoring. A model that scores well offline can still degrade quietly once the world it was fitted to changes.