Application of Data Mining Techniques to Identify Reducing Factor for Teff Yielding

Abstract

eff is an economically superior commodity in Ethiopia. It frequently controls a price level that is two to three times that of maize, the product with the highest production volume in the country, making teff an important cash crop for producers. The datasets for this study were obtained from the Central Statistical Agency of Ethiopia database, and the researcher trained and built models on a total of 7653 instances. Therefore, in order to construct a predictive model identifying reducing factors for teff yielding used Microsoft Excel and WEKA 3.9.3 version data mining tool was employed, respectively. Various experiments were carried out in order to achieve the goal of this research three classification algorithms to build the model that are J48 decision tree classifier, Random Forest classifier and REPTree classifier. In this regard, to select the best model/classifier, predictive performance evaluation and comparisons were employed using different model performance metrics Accuracy rate, TP, FP rate F-Measure, ROC Area, Confusion Matrix and Error rates. Based on this, among the three classifiers Random Forest Decision Tree classifier performs better accuracy and error rate compared to the other which is 97.844% and 2.156% respectively. As a result Random Forest classifier was selected for implementing the model to predict reducing factor for teff yielding. In this thesis, the experimental result shows that, the main determinant factors for reducing factor for teff yielding are main season (season type), use of field type, fertilizer type and crop d