Predictive Data mining is a major technique, which is supported by Machine Learning (ML) and is the most important criteria for any kind of ML applications. The datasets instances used by ML algorithms are represented by using the similar group of characteristics. The characteristics might be continuous, categorical or binary. If the known labels are given to instances such kind of learning is called as supervised, similarly where the instances are not provided with labels then we call it as unsupervised learning. The main motto of supervised learning is that creating a concise model for the class labels distribution regarding predictor characteristics. During this a classifier, which is produced will map a class label to examine the instances of known values of the predictor features where a value of class label is unknown. The term Classification refers to the method of forecasting the same data based on the categorical target value or a categorical class variable. This might be purposeful for any form of statistical data. The paradigms are most useful for the image classification, techniques of data mining, Predictive modeling and so on. Hence many techniques were developed on the basis of Artificial Intelligence (Logic-based techniques, Perceptron-based techniques) and Statistics (Bayesian Networks, Instance-based techniques). This paper describes about the major kinds of ML algorithms along with an experimental study and its applications. The future scope of Machine Learning and its importance in different research domains are also mentioned.