# Machine learning Core algorithms, honest evaluation and real-world applications of machine learning, in one map. ## Overview Machine learning (ML) is the practice of building systems that learn patterns from data instead of following hand-written rules. Given examples, such as past transactions labelled as fraud or not, an ML model learns to make predictions on new cases. A new employee learns to spot suspicious invoices by reviewing hundreds of past cases with a mentor; machine learning does the same with data, at scale and consistently. This page maps the foundations, the methods that work best on business data, how to evaluate honestly, and where ML creates value across industries. ## Foundations Every practitioner should understand a handful of classic methods: [Linear regression](/tools/linear-regression) and [Logistic regression](/tools/logistic-regression) as interpretable baselines, [Decision trees](/tools/decision-trees) and [Gradient boosting](/tools/gradient-boosting) for tabular data, [k-nearest neighbors](/tools/k-nearest-neighbors) and [Naive Bayes](/tools/naive-bayes) as simple, fast learners, and [Principal component analysis](/tools/principal-component-analysis) and [t-SNE](/tools/t-sne) for understanding high-dimensional data. Broader families are covered in [Supervised learning](/tools/supervised-learning), [Unsupervised learning](/tools/unsupervised-learning) and [Reinforcement learning](/tools/reinforcement-learning). ## How models learn Almost every model is trained the same way: pick a [loss function](/tools/loss-functions) that scores mistakes, then lower it step by step with [gradient descent](/tools/gradient-descent). Good results depend as much on the data as on the algorithm: [data preprocessing](/tools/data-preprocessing) and [feature engineering](/tools/feature-engineering) shape what the model sees, [regularization](/tools/regularization) keeps it from memorising, the [bias-variance trade-off](/tools/bias-variance-trade-off) explains under- and overfitting, [hyperparameter tuning](/tools/hyperparameter-tuning) finds good settings, and [evaluation metrics](/tools/evaluation-metrics) say whether it actually works. ## Classic models [SVM](/tools/svm), [Random forest](/tools/random-forest), [Clustering](/tools/clustering) and [Isolation forest](/tools/isolation-forest) remain everyday tools. [Ensemble learning](/tools/ensemble-learning) explains why combining many models, as boosting libraries do, so often wins. ## Learning paradigms How much labelled data you have decides the approach: [Supervised](/tools/supervised-learning), [Semi-supervised](/tools/semi-supervised-learning), [Self-supervised](/tools/self-supervised-learning) and [Unsupervised learning](/tools/unsupervised-learning), learning from rewards in [Reinforcement learning](/tools/reinforcement-learning), and reusing knowledge with [Transfer learning](/tools/transfer-learning), [Few-shot](/tools/few-shot-learning) and [Zero-shot learning](/tools/zero-shot-learning). ## Tabular data and foundation models Most business data lives in tables. Boosted trees such as [XGBoost](/tools/xgboost), [LightGBM](/tools/lightgbm) and [CatBoost](/tools/catboost) have long dominated; tabular foundation models such as [TabPFN](/tools/tabpfn) and the large tabular foundation model documented in [this site's docs](/model) now predict on new tables in context, without training a model per table. The detailed guide is [Tabular foundation models](/tools/tabular-foundation-models), with [In-context learning](/tools/in-context-learning) explaining the mechanism. ## Evaluating honestly Good models are proven, not assumed. [Cross-validation](/tools/cross-validation) estimates generalisation; held-out reports, leakage checks and calibrated metrics keep results honest; explainability tools such as [SHAP](/tools/shap) and [LIME](/tools/lime) show why a model decides as it does. See also [Evaluation](/model/evaluation) in the model docs. ## Where ML is applied Industrial and business applications range from [Predictive maintenance](/tools/predictive-maintenance) and [Digital twins](/tools/digital-twins) to credit and fraud in [FinTech](/tools/fintech), network operations in [Telecom](/tools/telecom), forecasting in [Logistics](/tools/logistics) and personalisation in [E-commerce](/tools/e-commerce). ## The full map ### ML foundations - [Linear regression](https://www.ozercemkelahmet.com/tools/linear-regression.md) - [Logistic regression](https://www.ozercemkelahmet.com/tools/logistic-regression.md) - [Decision trees](https://www.ozercemkelahmet.com/tools/decision-trees.md) - [Gradient boosting](https://www.ozercemkelahmet.com/tools/gradient-boosting.md) - [k-nearest neighbors](https://www.ozercemkelahmet.com/tools/k-nearest-neighbors.md) - [Naive Bayes](https://www.ozercemkelahmet.com/tools/naive-bayes.md) - [Principal component analysis](https://www.ozercemkelahmet.com/tools/principal-component-analysis.md) - [t-SNE](https://www.ozercemkelahmet.com/tools/t-sne.md) - [Cross-validation](https://www.ozercemkelahmet.com/tools/cross-validation.md) ### Learning paradigms - Supervised learning - Unsupervised learning - Semi-supervised learning - Self-supervised learning - Reinforcement learning - Transfer learning - Few-shot learning - Zero-shot learning - Meta-learning ### Training fundamentals - [Gradient descent](https://www.ozercemkelahmet.com/tools/gradient-descent.md) - [Loss functions](https://www.ozercemkelahmet.com/tools/loss-functions.md) - [Regularization](https://www.ozercemkelahmet.com/tools/regularization.md) - [Bias-variance trade-off](https://www.ozercemkelahmet.com/tools/bias-variance-trade-off.md) - [Evaluation metrics](https://www.ozercemkelahmet.com/tools/evaluation-metrics.md) - [Hyperparameter tuning](https://www.ozercemkelahmet.com/tools/hyperparameter-tuning.md) - [Feature engineering](https://www.ozercemkelahmet.com/tools/feature-engineering.md) - [Data preprocessing](https://www.ozercemkelahmet.com/tools/data-preprocessing.md) - [Synthetic data](https://www.ozercemkelahmet.com/tools/synthetic-data.md) ### Classic models - SVM - Random forest - [Ensemble learning](https://www.ozercemkelahmet.com/tools/ensemble-learning.md) - XGBoost - LightGBM - CatBoost - Clustering - Isolation forest ### Industries and applications - [Predictive maintenance](https://www.ozercemkelahmet.com/tools/predictive-maintenance.md) - [Digital twins](https://www.ozercemkelahmet.com/tools/digital-twins.md) - [Industry 4.0](https://www.ozercemkelahmet.com/tools/industry-4-0.md) - [Smart manufacturing](https://www.ozercemkelahmet.com/tools/smart-manufacturing.md) - [Smart city](https://www.ozercemkelahmet.com/tools/smart-city.md) - [Urban intelligence](https://www.ozercemkelahmet.com/tools/urban-intelligence.md) - [Telecom](https://www.ozercemkelahmet.com/tools/telecom.md) - [FinTech](https://www.ozercemkelahmet.com/tools/fintech.md) - [Logistics](https://www.ozercemkelahmet.com/tools/logistics.md) - [E-commerce](https://www.ozercemkelahmet.com/tools/e-commerce.md) - [Online travel](https://www.ozercemkelahmet.com/tools/online-travel.md) - [Real-time systems](https://www.ozercemkelahmet.com/tools/real-time-systems.md) - [Augmented reality](https://www.ozercemkelahmet.com/tools/augmented-reality.md) ## Related topics - [Deep learning](/topics/deep-learning) - [Big data](/topics/big-data) - [MLOps and AI infrastructure](/topics/mlops-ai-infrastructure) ## Frequently asked questions ### What is the difference between AI and machine learning? AI is the broad goal of building intelligent systems; machine learning is the main approach used today, in which systems learn from data. ### Which ML method works best on tables? Gradient-boosted trees are a strong default; tabular foundation models are increasingly competitive and need no per-table training. ## Summary Machine learning turns data into predictions. Strong foundations, methods suited to tabular data, honest evaluation and clear business problems are what make it work in practice.