MethodologyIn a world where machines capable ofthinking on their own are a dream of many to create, Machine learning has madeit achievable to make smarter and intelligent decisions in the IoT, which turnsout to be one of the reasons for the success of the Internet of Things. Figure 2: Learning from the availabledata without being explicitly programmedSource: Gartner(January 2017)Machine learning Algorithms areessentially characterized into three major categories. They are as follows:· Supervised Machine Learning Algorithms.
· Unsupervised Machine LearningAlgorithms.· Reinforcement Machine LearningAlgorithms. Supervised Machine learning algorithmsconsists of an input and as well as an output variable which are used for predictingfrom a given set of variables. Examples of Supervised Machine LearningAlgorithms are KNN, Decision Trees, and Logistic Regression etc. In unsupervised machine learningalgorithm, there are no input or output variables, instead it is used to figureout useful patterns and structure from the data. Some of the examples ofunsupervised learning algorithm are K-means algorithm, Apriori Algorithm etc. In the Reinforcement machine learningalgorithm, the machines are trained to make decisions by calculating how goodor bad a situation is. Examples of this type of algorithms include MarkovDecision Process etc.
TheMachine learning approaches and algorithms are essentially data driven and theyare proposed and enforced to make decisions or predictions from the possesseddata (Rutta, Scioscia, Losetoa, Pintoa & Di Sciascioa, 2017). Every machinelearning algorithm is based on certain values of parameters and for everymachine learning algorithm, there will be a distinct way of understanding it.If, for interpreting a problem of classification, a machine learning algorithmmakes use of a divergent set of parameters, then the validity of theclassification of the problem will differ greatly in each case.
One of themajor difficulties in Machine learning is to search the most satisfactory pairof values for parameters for an algorithm to solve a problem in order to getthe best performance metrics. Therefore, to get the best result for solving aproblem, the best suitable pair of parameter values must be used. A key feature that makes Machinelearning much more appealing is that much explicit programming is not requiredto achieve meaningful understanding of data by Machine learning unlike othertechnologies.