What algorithms do you need to know to be a top algorithm expert?

There are two ways to categorize all the machine learning algorithms you encounter today.

• The first algorithm grouping is learning-style.

• The second algorithm is grouped by similarity of form or function.

In general, both methods can summarize the entire algorithm. However, we will focus on grouping algorithms by similarity.

Algorithms can model the problem in different ways, but we need data for whatever result we want. In addition, algorithms are popular in machine learning and artificial intelligence. Let’s look at three different ways of learning in machine learning algorithms:

Basically, in supervised machine learning, input data is called training data and has known labels or results, such as spam/non-spam or stock prices. Here, prepare the model through the training process. There are also predictions to be made. And correct those predictions when they are wrong. The training process continues until the model reaches the desired level.

• Example problem: classification and regression.

• Example algorithms: logistic regression and backpropagation neural networks.

In unsupervised machine learning, input data is unlabeled and there is no known result. We must prepare the model by deducing the structure that exists in the input data. This may be extracting general rules, but we can reduce redundancy through mathematical processes.

Example problems: clustering, reduction, and association rule learning.

Example algorithms: Apriori algorithm and K-means.

The input data is a mixture of tagged and untagged examples. There is the problem of expected predictions, but the model must learn to organize the data as well as the structure for making predictions.

• Example problem: classification and regression.

• Example algorithms: Extensions of other flexible methods.

• Algorithms grouped by functional similarity

ML algorithms are usually grouped according to their similarity in functionality. For example, tree-based approach and neural network approach. But there are still algorithms that can easily fit multiple categories. Such as learning vector quantization, which is a neural network approach and an instance – based approach.

Regression algorithms involve modeling the relationships between variables, which we improve by measuring errors generated in the predictions made using the model.

These methods are the workhorses of data statistics, and they have also been selected for statistical machine learning. The most popular regression algorithms are:

• Ordinary least squares regression (OLSR);

• Linear regression;

• Logistic regression.

• Stepwise regression;

• Multivariate adaptive regression spline (MARS);

• Local estimated scatter diagram smoothness (LOESS);

This kind of algorithm is used to solve the decision problem of training data. These methods build a database of sample data that needs to be compared with the new data. For comparison, we use similarity measures to find the best match and make predictions. For this reason, instance-based approaches, also known as winner-takes-all approaches and memory-based learning, focus on storing the representation of instances. Therefore, similarity measures are used between instances. The most popular instance-based algorithms are:

• K-nearest Neighbor (kNN);

• Learning vector quantization (LVQ);

• Self-organizing feature mapping (SOM);

• Local Weighted learning (LWL);

I have listed regularization algorithms here because they are popular and powerful. And often make simple changes to other methods. The most popular regularization algorithm is:

• Ridge regression.

• Minimum Absolute Contraction and selection operator (LASSO);

• Elastic net regression;

• Minimal Angular regression (LARS);

The decision tree approach is used to build a decision model based on the actual values of data attributes. The decision forks in the tree structure until a predictive decision is made for a given record. Decision trees are usually fast and accurate, which is a favorite algorithm of machine learning practitioners. The most popular decision tree algorithms are:

• Classification and regression tree (CART);

• Iterative Dichotomiser 3 (ID3);

• C4.5 and C5.0(different versions of powerful methods);

• Chi-square automatic interaction detection (CHAID);

• Decision stump;

• M5;

• Conditional decision tree;

These methods apply to problems of Bayes’ theorem, such as classification and regression. The most popular Bayesian algorithms are:

• Naive Bayes;

• Gaussian naive Bayes;

• Multinomial naive Bayes;

• Mean dependence estimator (AODE);

• Bayesian belief network (BBN);

• Bayesian network (BN);

Almost all clustering algorithms involve using inherent structures in the data, which require optimal organization of the data into groups of maximum commonality. The most popular clustering algorithms are:

• K – average;

• K – average;

• Expectation maximization (EM);

• Hierarchical clustering;

Association rule learning method extracts rules, which can perfectly explain the relationship between variables in the data. It is important that these rules can be found in large cubes. The most popular association rule learning algorithms are:

• The Apriori algorithm;

• Eclat algorithm;

Most of these algorithms are inspired by the structure of biological neural networks. They can be a class of pattern matching that can be used for regression and classification problems. It has a huge subfield because it has hundreds of algorithms and variations. The most popular artificial neural network algorithms are:

• Machine perception;

• Back propagation;

• Hopfield neural network;

• Radial Basis Function Neural Network (RBFN)

Deep learning algorithm is an update of artificial neural network. They are more concerned with building larger and more complex neural networks. The most popular deep learning algorithms are:

• Deep Boltzmann machine (DBM);

• Deep Belief Network (DBN);

• Convolutional Neural Network (CNN);

• Stacked autoencoder;

Like the clustering method, dimension reduction is also to seek the inherent structure of the data. In general, it is very useful to visualize dimension data. In addition, we can use it in supervised learning methods.

• Principal component analysis (PCA);

• Principal component regression (PCR);

• Partial least squares regression (PLSR);

• Sammon Mapping;

• Multidimensional scaling (MDS);

• Projection tracking;

• Linear discriminant analysis (LDA);

• Gaussian mixture discriminant analysis (MDA);

• Quadratic discriminant analysis (QDA);

• Fisher discriminant analysis (FDA);

Often, classifying web pages, documents, and E-mail would be difficult and impossible. This is where naive Bayes classifier machine learning algorithms come in. A classifier is really a function that allocates the value of a population of elements. For example, spam filtering is a popular application of naive Bayes algorithms. Thus, a spam filter is a classifier that assigns labels “spam” or “non-spam” to all E-mail messages. Basically, it is one of the most popular learning methods for grouping by similarity. This applies to the popular Bayesian probability theorem.

Typically, K-means is an unsupervised machine learning algorithm for cluster analysis. In addition, k-means is a nondeterministic and iterative method. The algorithm operates on a given data set through a predetermined number of clusters K. Therefore, the output of the K-means algorithm is K clusters with input data separated between clusters.

Basically, it’s a supervised machine learning algorithm for classification or regression problems. SVM learns from data sets so that SVM can classify any new data. In addition, it works by sorting data into different classes through lookups. We use it to classify training data sets into several categories. Also, there are many such linear hyperplanes, and SVM tries to maximize the distance between the various classes, which is called marginal maximization.

SVM is divided into two categories:

• Linear SVM: In linear SVM, the training data must pass through a hyperplane separation classifier.

• Nonlinear SVM: In nonlinear SVM, it is not possible to separate training data using hyperplane.

This is an unsupervised machine learning algorithm. We use it to generate association rules from a given data set. Association rules mean that if item A occurs, item B also occurs with some probability, and most of the association rules generated are in IF_THEN format. For example, if people buy ipads, they also buy iPad cases to protect them. The basic principle by which the Apriori machine learning algorithm works: If the itemset occurs frequently, then all subsets of the itemset also occur frequently.

It shows the relationship between two variables, and it shows how changes in one variable affect the other.

A decision tree is a graphical representation that uses a branching approach to illustrate all possible outcomes of a decision. In a decision tree, internal nodes represent tests of attributes. Because each branch of the tree represents the result of the test, and the leaf node represents the specific class tag, the decision made after all the attributes have been computed. In addition, we must represent the classification by a path from the root node to the leaf node.

It is the machine learning algorithm of choice. We use bagging to create a bunch of decision trees with a random subset of data. We have to train the model many times on random samples of the dataset because we need to get good predictive performance from the random forest algorithm. Furthermore, in this integrated learning approach, we have to combine the output of all decision trees to make the final prediction. In addition, we poll the results of each decision tree to derive the final prediction.

The name of this algorithm can be a little confusing, Logistic regression algorithms are used for classification tasks rather than regression problems. Furthermore, the name “regression” here implies that linear models fit into feature Spaces. The algorithm applies logical functions to linear combinations of features, which are required to predict the results of classified dependent variables.

We studied machine learning algorithms and learned the classification of machine learning algorithms: Regression algorithm, the algorithm based on instance, regularization algorithm and decision tree algorithm, bayesian algorithm, clustering algorithm, association rule learning algorithm and artificial neural network algorithm, deep learning algorithm, dimension reduction algorithm and integrated algorithm, supervised learning and unsupervised learning, A semi-supervised learning algorithm, simple bayesian classifier, K means clustering algorithm, support vector machine (SVM) algorithm, A Priori algorithm, linear regression and Logistic regression. Being familiar with these algorithms will help you become an expert in machine learning!