Shulou(Shulou.com)06/03 Report--

This article mainly introduces "what kinds of machine learning algorithms are there". In daily operation, I believe many people have doubts about which kinds of machine learning algorithms. The editor consulted all kinds of materials and sorted out simple and easy-to-use operation methods. I hope it will be helpful for you to answer the doubts about "what kinds of machine learning algorithms are there?" Next, please follow the editor to study!

Machine learning is one of the most important sub-fields in data science. In 1959, IBM researcher ArthurSamuel first used the term machine learning. Since then, the field of machine learning has aroused great interest from many people.

When you start your journey to data science, the first sub-area you encounter may be machine learning. Machine learning is the name used to describe the set of computer algorithms, which are constantly learned and improved by collecting information during the operation.

Machine learning algorithms are based on some data. Initially, the algorithm uses some "training data" to establish intuition to solve specific problems. Once the algorithm has passed the learning stage, the similar problems based on different data sets can be solved through the acquired knowledge.

Generally speaking, machine learning algorithms are divided into four categories:

Supervision algorithm: the developer's supervision is needed in the process of running. To do this, developers can mark training data and set strict rules and boundaries for the algorithms to follow.

Unsupervised algorithm: an algorithm that is not directly controlled by developers. In this case, the expected result of the algorithm is unknown and needs to be defined by the algorithm.

Semi-supervised algorithm: this algorithm combines all aspects of supervised algorithm and unsupervised algorithm. For example, when initializing an algorithm, not all training data will be marked and no rules will be provided.

Reinforcement algorithms: this type of algorithm uses a technique called exploration / development. The technical content is simple; the machine performs one action, observes the results, and then considers those results when performing the next action, and so on.

Each of the above algorithms has a specific goal. For example, supervised learning aims to expand the scope of training data and predict future or new data accordingly. On the other hand, unsupervised algorithms are used to organize and filter data to make it meaningful.

Each category has a variety of specific algorithms designed to perform specific tasks. This article will introduce five basic algorithms that every data scientist must understand, covering the basic knowledge of machine learning.

1. Regress

The regression algorithm is a supervised algorithm, which is used to find the possible relationship between different variables in order to understand the influence of independent variables on dependent variables. Regression analysis can be regarded as an equation. For example, suppose that the equation y = 2x + zrecoery y is a dependent variable, then XPerry z is an independent variable. Regression analysis is to find out to what extent x and z affect the value of y.

The same logic applies to more advanced and complex problems. For all kinds of problems, there are many types of regression algorithms. The top five most commonly used are probably:

Linear regression: the simplest regression technique uses a linear method to describe the relationship between dependent variables (predicted values) and independent variables (values used for prediction).

Logistic regression: this type of regression is used for binary dependent variables and is widely used to analyze classified data.

Ridge regression: Ridge regression corrects the coefficients of the model when the regression model becomes too complex.

Lasso regression: Lasso (minimum absolute contraction selector operator) regression is used to select variables and regularize them.

Polynomial regression: this type of algorithm is used to fit nonlinear data. The best prediction when used is not a straight line, but a curve that tries to fit all data points.

two。 classification

Classification in machine learning is the process of classifying items based on pre-classified training data sets. Classification is considered as one of the supervised learning algorithms. These algorithms use the classification results of training data to calculate the probability of new items falling into one of the defined categories. A famous example of a classification algorithm is to classify incoming e-mail into spam or non-spam.

There are many types of classification algorithms, the most commonly used of which are:

K-nearest neighbor: KNN is an algorithm that uses training data sets to find k nearest data points in some data sets.

Decision tree: think of it as a flowchart, dividing each data point into two categories at a time, and then into two categories, and so on.

Naive Bayes: this algorithm uses conditional probability rules to calculate the probability that an item belongs to a particular category.

Support vector machine (SVM): in this algorithm, the data is classified according to the degree of polarity of the data, which may be beyond the X / Y prediction range.

Image source: Google

3. Integration

The ensemble algorithm obtains more accurate results by combining the predictions of two or more other machine learning algorithms. The results can be combined by voting or average results. Voting is usually used in the classification process, while averages are used in the regression process.

There are three basic types of integration algorithms: Bagging, Boosting, and Stacking.

Bagging: in Bagging, algorithms run in parallel on different training sets of the same size, then test all algorithms with the same dataset and vote to determine the overall results.

Boosting: in the case of Boosting, the algorithm runs sequentially, and then uses a weighted vote to select the overall result.

Stacking: as the name implies, Stacking consists of two levels, the primary learner is a combination of algorithms, and the secondary learner is a meta-algorithm based on the results of the basic level.

4. Clustering

Clustering algorithm is a group of unsupervised algorithms for grouping data points. The points in the same cluster are more similar to each other than the points in different clusters. There are four types of clustering algorithms:

Centroid-based clustering: this clustering algorithm organizes data into classes according to initial conditions and outliers. K-means is the most commonly used clustering algorithm based on centroid.

Density-based clustering: in this type of clustering, the algorithm connects high-density regions to the cluster to create a distribution of arbitrary shape.

Distribution-based clustering: this clustering algorithm assumes that the data is composed of a probability distribution, and then clusters the data into various versions of the distribution.

Hierarchical clustering: this algorithm creates a tree of hierarchical data clusters that can be changed by cutting the tree at the correct level.

5. Association

The association algorithm is an unsupervised algorithm, which is used to find the probability of some items appearing together in a specific data set, mainly for shopping basket analysis. The most commonly used association algorithm is Apriori. Apriori algorithm is a commonly used mining algorithm in transaction database. Apriori is used to mine frequent itemsets and generate association rules from those itemsets.

For example, if a person buys milk and bread, he may also buy some eggs. This can be seen from the previous purchase records of each customer. Then the algorithm calculates how frequently these items are purchased together and forms association rules according to the specific threshold for the confidence.

Image source: Google

Machine learning is one of the most famous and deeply studied sub-fields in data science. People have been developing new machine learning algorithms to achieve higher accuracy and faster execution speed. No matter which algorithm is used, it can usually be classified into one of the following four categories: supervised, unsupervised, semi-supervised and enhanced algorithms. The purpose of each algorithm is different.

These algorithms have been deeply studied and widely used, and you only need to know how to use it, not how to implement it. Most well-known Python machine learning modules, such as ScikitLearn, contain predefined versions of most, if not all, of these algorithms.

Once you understand how it works, master the usage and start using it.

At this point, the study of "what are the types of machine learning algorithms" is over. I hope to be able to solve your doubts. The collocation of theory and practice can better help you learn, go and try it! If you want to continue to learn more related knowledge, please continue to follow the website, the editor will continue to work hard to bring you more practical articles!

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.