Employing both descriptive and predictive algorithms toward improving prediction accuracy

Abstract

The research describes the use of both descriptive and predictive algorithms for better accurate prediction. The current research has focused on the use of either descriptive or predictive algorithm for prediction, but this research work employed the two  algorithms. Clustering technique was used in the descriptive stage while classification technique was used in the predictive stage. K-Means and Expected Maximization (EM) were used for clustering while models from three classifiers (Decision Stump, M5P and RepTree) were used for classification. The result of using each of the two algorithms individually was presented as well as the result of combination of both algorithms. It was discovered that utilizing both algorithms for prediction provided more accurate result.

Keywords: Data Mining, Clustering, Classification, Expected Maximization, M5P