MACHINE LEARNING CLASSIFIERS

What separates the brain from the other mechanical parts of our body? It is its potential to learn and improve from its past experiences. Similar is the task and purpose of machine learning. It provides the ability to the system to learn from its past experiences without being explicitly programmed. 

Reference: https://becominghuman.ai/an-introduction-to-machine-learning-33a1b5d3a560

In machine learning and statistics, classification can be defined as the problem of identifying a pattern or defining a certain category amongst the set given of observations. This thought and statement can be defined by considering a simple example of classifying emails as “spam and non-spam” or sorting the list of voters on the basis of their gender as “male or female”. Basically, taking a complex set of data and distinguishing it on one or more criteria is classification.

An algorithm that implements this classification of materials via an electronic device is called a classifier. As of today, there are innumerable programs and codes available to individuals to classify materials, the onus is on the user as regards to which form of classification he/she desires from the algorithm. On a broad basis, classification in machine learning is of two types- when the outcome contains two options (classifying mail as spam or non-spam) or multistage classification(where there are more than two options to the situation). Here are a few of the endless machine learning classifiers:

Linear Regression and Logistic Regression

Linear Regression is the approach to model the relationship between a dependent and an independent variable while logistic regression uses  a logistic function to model a binary dependent variable.

Linear regression follows logistics regression when it comes to popularity as a machine learning algorithm. While in a lot of ways, these two are similar, the biggest difference lies in what they are used for. For tasks comprising of forecasting or predicting values, linear regression algorithm has an edge over logistic regression which is preferred for classification tasks. Spam or not classification, fraudulent or not, etc are some examples of usage of these algorithms in machine learning.

Reference: https://www.datacamp.com/community/tutorials/understanding-logistic-regression-python

Decision tree

Decision tree forms a certain classification or regression in the form of a tree structure, using the concept of branch, roots, tuples, and nodes. It uses an if-then rule set which is exclusive for classification. The rules are learned sequentially using the training data one at a time. Each time a rule is conceptualized, the tuples covered by the rules are removed. This process is continued on the training set until meeting a termination condition. The tree is constructed in a way where it is recursive in its approach. All the attributes should be categorical. Otherwise, they should be discretized in advance. Attributes at the top of the tree have more impact on the classification and they are identified using the information gain concept.

Reference: https://www.kdnuggets.com/2020/01/decision-tree-algorithm-explained.html

Naive Bayes

A probabilistic classifier that is heavily dependent on Bayes’ theorem. This theorem has a simple algorithm and can be easily scaled to larger datasets, rather than being dependent on approximation. It presumes that every variable in the dataset is independent of the other variables, irrespective of their influence on the outcome. Naive Bayes is a very simple algorithm to implement and good results have obtained in most cases. It can be easily scalable to larger datasets since it takes linear time, rather than by expensive iterative approximation as used for many other types of classifiers.

Reference: https://helloacm.com/how-to-use-naive-bayes-to-make-prediction-demonstration-via-sql/

The k-nearest-neighbours

The k-nearest-neighbours algorithm is a supervised classification technique that uses closeness as a measure for similarity. This algorithm labels points based on a bunch of previously labeled points. In order to label a new point, it looks at its nearest labeled members. Closeness is typically expressed in terms of a dissimilarity function. Once it checks with ‘k’ number of nearest neighbours, it assigns a label based on whichever label most of the neighbours have.

Geometric distance is an unreasonable or impractical measure to determine the nearest item. If the type of input is, for instance, a text, it is unclear as to how the variables are drawn in their geometric representation. Hence, calculation of distance, unless well-calibrated on a case-by-case basis, is vague and unreliable.

Reference: https://www.datacamp.com/community/tutorials/k-nearest-neighbor-classification-scikit-learn

 Random Forest

Random forests are an ensemble learning method, combining one or more algorithms of same or different kinds of classifying objects, for classification, regression, etc. It operated by constructing several decision trees. Each tree provides its mean prediction. The mode of the predictions of the individual trees is the output of the entire algorithm. Working in a manner similar to that of decision trees, it corrects for the latter’s habit of overfitting to the training set.

Reference: https://medium.com/@williamkoehrsen/random-forest-simple-explanation-377895a60d2d

Support Vector Machine

Support-vector machine models with associated learning algorithms that analyzes data used for regression and classification analysis. Given a group of coaching examples with each element marked as belonging to at least one or the opposite of two categories, an SVM training algorithm builds a model that assigns new examples to at least one category or the opposite, making it a non-probabilistic binary linear classifier. An SVM model may be a representation of the examples as points in space, mapped in order that the samples of the separate categories are divided by a transparent gap that's as wide as possible. New examples are then mapped into that very same space and predicted to belong to a category supported the side of the gap on which they fall.

Reference: https://en.wikipedia.org/wiki/Support-vector_machine

Authors: Laksh Maheshwari, Ankit Lad, Aditya Kulkarni, Ayush Mehta, Jayant Majji.