A Picture Shows You the Most Common Supervised Machine Learning Models
Supervised learning and unsupervised learning are two key types of machine learning problems. Supervised models are used to predict target values from labeled input data, while unsupervised models are used to find patterns in unlabeled input data.
In this article, I separate the common supervised machine learning models into two sub-categories, one for solving classification problems, and the other for solving regression problems. In each sub-category, all the corresponding models are grouped into three families — Linear family, KNN family, and Tree family.
Below is a diagram of the most common models used in supervised machine learning.
Here is a simple summary of the supervised learning models.
Linear models are easy to train and fast to make predictions. The nature of linear models is to find global relationships by estimating the model coefficient w and constant b. If your dataset has a clear global pattern, the linear model can make predictions quickly.
KNN models are easy to understand. Generally, a target value is estimated by the “majority vote” of its K nearest neighbors (data points). Unlike linear models, knn models do not make strong assumptions about the input data, and they can capture more local fluctuations and work well with sparse data.
Tree models are widely used in real-world applications. The decision tree model is easy to interpret and visualize. We can see how the decision is made by the model step by step, and we don’t need to do much feature tuning. Although the other two tree-based models (Random Forest and GBDT) are difficult to interpret, they are widely used due to their excellent prediction accuracy and ability to handle mixed feature types.
If you’d like to know more about how to apply different supervised machine learning models on real-life data to solve classification problems, check out my coding tutorials below on how to apply and evaluate different models on breast cancer data for benign and malignant classification.
