Title: COMPARATIVE STUDIES OF ARTIFICIAL INTELLIGENCE (AI) TECHNIQUES

Year of Publication: 2013
Page Numbers: 313-319
Authors: Lubabatu Sada Sodangi, Suleiman Isah Sani
Conference Name: The Third International Conference on Digital Information Processing and Communications (ICDIPC2013)
- United Arab Emirates

Abstract:


Determining the accuracy and duration of AI techniques is important for the reason that the Application of such techniques in Data mining has become wider. There is a range of techniques used in AI for finding hidden or unknown information in data sets (or data groups), and most of these techniques have their own sub fields. Artificial intelligence techniques are extensive and too numerous to list. In this paper, a machine learning algorithm that combines two widely used algorithms Decision trees [J48] and Neural networks [Multilayer perceptron] are compared and studied using Waikato Environment for Knowledge Analysis(Weka) and Iris dataset from UCI machine learning repository. These two algorithms determine and compare the classifier errors and time taken to build the model in the explorer window and compare the Error rate of each algorithm in experimenter window. The result shows that Decision Trees [J48] runs faster but produce high error rate and low accuracy whereas Neural Network [Multilayer perceptron] is more preferable in Data mining although it takes longer training time, yet it yields low error rate and high accuracy.