Perbaikan Performansi Klasifikasi Dengan Preprocessing Iterative Partitioning Filter Algorithm

Abstract

Preprocessing data and preprocessing performance analysis are crucial in data
mining. Those two points have great impact to data mining process success rate, because a
quality decisions must be based on quality data. Preprocessing is useful to increase the quality
of data and to reduce the noise data. Our experiment show that the performance iterative
partitioning filter algorithm is tested by using some dataset from University of California, Irvine
(UCI) Machine Learning Repository and is simulated by using modified iterative partitioning
filter's parameter variation. This experiment also explained how to analyze classification result
from a preprocessed dataset using Backpropagation, so that it can identify best accuracy from
multiple datasets that have been tested. Final result from this experiment is table of data consist
of training time, classification accurarcy, classification error, Kappa statistic, Mean Absolute
Error (MAE) or average of iterations error, Root mean squared error and confusion matrix. This
final result is presented in ratio chart between experiment result and modified iterative
partitioning filter's parameter