Machine learning hascome into recognition in recent years.Recently, machinelearning has been a buzzword. Machine learning is a category of artificialintelligence which let software application to be rigorous to anticipate resulteasily. The output value is predicted within adequate range with the help ofinput data through algorithms and statistical analysis. This paper describesvarious learning algorithms along with the decision tree technique which isclassification algorithm under supervised learning. In decision trees,depending on parameters the data is divided constantly. Based on various inputvariable the objective is to predict target variables with the help of decisiontree algorithm.
Decision tree features and metrics are described. The usage ofthe decision tree in various fields is explained. Further, advantages andlimitations of decision trees are listed.Keywords: Machinelearning, supervised learning, decision treesINTRODUCTIONIn traditionalprogramming data and program are given as inputs to the computer and afterprocessing output is known. Machines are made learn by providing data andoutput in order to come up with the program from the computer as shown in fig.1 Fig. 1 Structure ofTraditional Programming and Machine LearningThe idea of makingmachines learn came up from different fields such as statistics, psychology,biology, and computer science. Artificial intelligence algorithms areconsidered as a fundamental to make a machine learn and work intelligently andeffectively depending upon situations.
The motive behind making machines learnis to come up with the best prediction or outcome. Machines are programmed insuch a manner that they produce an output learned from past experiences 1.Among different objectives of machine learning, the important one is to find asolution for a given problem by instructing computer for using the data orprevious actions. To yield knowledge from immense dispersed data is collectedover and over. The collected data thus help organizations to come up with newoutcomes which benefit their business along with prevailing market interest.Large data is collected from different platforms such as social networkingsites, stocks trading, e-commerce sites, surveys 2.
The foremost step is toformulate huge data accurately. Thus decision trees are required to come upwith useful data. The efficient working of decision trees is on discrete databut it also works on categorical data. As decision trees classify data from previousinformation it falls in supervised learning technique. To predict future eventssupervised learning constructs algorithm help in the production of ordinarypatterns and assumptions that takes place with the help of external instancesprovided. For a learning algorithm to besupervised training a set of data is provided including the precise knownoutput. To provide improved accuracy for classification of data theapplications of real-world decision trees as considered as finest algorithm 3.
The comparison of a decision tree has been made effectively with human expertsand computerized data. These trees have turned out to be a profitable tool forthe depiction, grouping, and generalization of information. Working on buildingdecision trees from existing data exists in different disciplines, for example,artificial neural network, machine learning, the theory of decision making andstatistics.