| Summary: | There are few studies using the data mining approach to assess the quality of water, especially for Selangor rivers. This study assesses the water quality using data mining techniques and identified the most significant variables that affect water quality. Machine learning techniques used are Decision Tree (Gini) and Decision Tree (Entropy), Logistic Regression Enter, Backward Elimination and Forward Selection and Artificial Neural Network with 4 and 8 hidden nodes. This study revealed that Logistic Regression Enter is the best model since it is neither underfit nor overfit with the sensitivity, specificity, accuracy, mean squared error and misclassification rate values of 92.51%, 97.45%, 96.36%, 0.028 and 3.64% respectively. There are other two best models: Decision Tree (Gini) and Artificial Neural Network with 4 hidden nodes. According to the variable importance output based on Decision Tree (Gini), the most important variable effect on the water quality is Biochemical Oxygen Demand (BOD) with the highest value of 0.2284, followed by Chemical Oxygen Demand with a value 0.1471 respectively. © Penerbit UMT
|
|---|