A Comprehensive Evolution for Applicability of Machine Learning Algorithms on Various Domains

Abstract

Machine learning algorithms are able to learn from data, make decision and improve what they learn by having experience without human intervention. Machine learning techniques are becoming increasingly important nowadays that everything is going to be fully automated. They have been used in various fields such as recommendation engines, self-driving cars, offering personal suggestions from retailers, cyber fraud detection, face recognition, and etc. This study presents some of the most commonly used machine learning techniques from supervised and unsupervised learning classes such as linear regression, logistic regression, neural networks, and self-organizing map. In linear regression technique, it is tackled to fit a linear function to user data in order to model the relationship between variables. It can be a useful technique to make weather estimation, to understand marketing effectiveness and to model consumer behavior. Logistic regression is a statistical model that uses a logistic function and is appropriate when dependent variable is binary. Neural networks mimic the operation of human brain to recognize patterns from the underlying data. They have wide range of application such as cancer diagnosis, e-mail spam filtering, and signal classification. Self-organizing map, a special type of neural networks, is utilized to achieve dimensionality reduction that generally used for seismic analysis, project prioritization, and image processing such as color reduction. Each implementation in this study shows that the success of the results obtained by applying machine learning techniques depends on using the right technique in the appropriate area.