Analysis of classification learning algorithms

Abstract

The paper attempts to apply data mining technique, to estimate the teacher performance of college of Information Engineering (COIE) In Al Nahrain University in Baghdad/Iraq, Five classifications algorithms were used to build data they are (ZeroR, SMO, Naive Bayesian, J48 and Random Forest). The analysis implemented using WEKA (3. 8. 2) Data mining software tool. Information was collected from within the variety of form using “Referendum”; it was stored in Excel file CSV format then regenerate to ARFF (Attribute-Relation File Format). Many criteria like (Time is taken to create models, accuracy and average error) was taken to evaluate the algorithms. Random forest and SMOPredicts higher than alternative algorithms since their accuracy is the highest and have the lowest average error compared to others, “The teacher clarification, and wanting to be useful to students”, was the strongest attribute. Further, removing the bad ranked attributes (10, 11, 12, and 14) that have a lower contact on the Dataset can increase accuracies of algorithms.