Machine Learning Algorithms

To many, it might seem that all of a sudden the world has awakened to AI and machine learning, but few, know that it all started when the British scientist Alan Turing, invented the "Turing Test" while helping Britain decipher war messages sent from Nazi boardrooms to German troops . The test, by definition, identifies the smart machine from a machine. The identification is done by humans which, when interacting with machines are not able to find whether they are interacting with machines or fellow humans. The machines exhibiting such cognitive abilities (thinking like humans) are considered smart.

The field which provides machines the ability to learn and improve from experience without being explicitly programmed is machine learning. This field is ever expanding and has applications in everything under the Sun

Broadly, there are three types of Machine Learning Algorithms:-

1. Supervised Learning

• This algorithm consist of a Target Variable (or dependent variable) which is to be predicted from a given set of predictors (independent variables). Using these set of variables, we generate a function that map inputs to desired outputs. The training process continues until the model achieves a desired level of accuracy on the training data.

•Examples of Supervised Learning: Regression, Decision Tree, Random Forest, KNN, Logistic Regression etc. 

2. Unsupervised Learning

•This happens when the machines learn from examples without any labeled response. In this algorithm, we do not have any target or outcome variable to predict / estimate. This type of algorithm tends to restructure the data into something else, such as new class or a new series of uncorrelated values. They are quite useful in providing humans with insights into the meaning of data and new useful inputs to supervised machine learning algorithms.

• Examples of Unsupervised Learning: Apriori algorithm, K-means. 

3. Reinforcement Learning:

Using this algorithm, the machine is trained to make specific decisions. It works this way: the machine is provided positive and negative feedback with examples and is then exposed to an environment where it trains itself using trial and error. The machine learns from past experience and tries to capture the best possible knowledge to make accurate business decisions.

Approaches to Machine Learning

•Regression Algorithms- are the algorithms which try to predict dependent variable by forming its relationship with independent variables.

•Linear -Predict a numeric value. 

•Logistic –Predict a variable which has only two values(binary)

•Clustering Algorithms- are the unsupervised algorithms which assign the given set of observations into clusters such that observations in a cluster are homogenous.

•Decision Tree Algorithms are the supervised algorithms which make a predictive model by building from a variable’s observations to its target value.

•Bayesian Algorithms- are used for classification and are based on Bayes' theorem of conditional probability . It would consider all predictors as independent of each other which will independently contribute to the probability of an event.

•Instance-based Algorithms- creates a database of the stored observations and then compare new data to the database using a similarity measure to find the best match and make prediction at that very instant.

•Association Rule Learning Algorithms make rules that best explain the observed relationship between variables such as IF Then Association.

Part 2 of this article will have the comparison matrix and their applications