Machine learning is a relatively common and effective method in data analysis and data mining. Machine learning is divided into four categories, namely supervised learning, unsupervised learning, semi-supervised learning and reinforcement learning.
1. Supervised learning
Concept: Use samples containing feature and label information as training samples, and obtain an optimal model through training sample training, and then use this model to map all inputs to corresponding outputs, and make simple judgments on the outputs to achieve prediction and classification Purpose, it also has the ability to predict and classify unknown data.
Common algorithms: classification algorithms (KNN, Naive Bayes, SVM, decision trees, random forest, BP neural network algorithms, etc.) and regression algorithms (logistic regression, linear regression, etc.).
Application scenarios: classification and regression scenarios, such as spam classification, heart disease prediction, etc.
2. Unsupervised learning
Concept: The labeled information of the training samples is unknown. The goal is to reveal the inherent properties and laws of the data through the learning of unlabeled training samples, and provide a basis for further data analysis. The most studied and widely used of this type of learning task is "clustering". The purpose of clustering is to group similar things together, which is mainly obtained by calculating the distance between samples and the distance between groups.
Common algorithms: density estimation, anomaly detection, hierarchical clustering, EM algorithm, K-means algorithm, DBSCAN algorithm, etc.
Application scenarios: clustering scenarios, such as aggregate news websites.
3. Semi-supervised learning
Concept: Supervised learning requires a large amount of labeled data, which is a costly learning method. Unsupervised learning does not rely on the labeling of data, but it has some difficulties in solving classification and regression problems. Therefore, by partially labeling the samples, and then using machine learning algorithms, it is semi-supervised learning.
Common algorithms: generative model algorithms, self-training algorithms, joint training, semi-supervised support vector machines, methods based on graph theory, etc.
Application scenarios: Some scenarios where the tag data is difficult to obtain.
4. Reinforcement learning
Concept: The program or agent learns a mapping from environment to action by continuously interacting with the environment. The goal of learning is to maximize the cumulative return.
Reinforcement learning is a kind of trial and error learning, because it needs to try all possible actions in various states (environments), and judge the pros and The mapping relationship of the optimal action (ie strategy).
Common algorithms: Hidden Markov, Monte Carlo.
Application scenarios: For scenarios that require constant reasoning in the process, such as driverless car driving, Alpha Dog playing Go, etc.
references
[1]【Machine learning notes】Supervised learning and unsupervised learning
[2] Several semi-supervised learning algorithms
[3] Reinforcement learning for Python machine learning applications
[4] Li Bo. Practical application of machine learning [M]. Beijing: People's Posts and Telecommunications Press, 2017: 16-18.