Detection of Interviewer Falsification in Statistics Indonesia’s Mobile Survey
Yusep Rosmansyah, Ibnu Santoso, Ariq Bani Hardi, BSSN, Jakarta, Indonesia, Atina Putri, Sarwono Sutikno · International Journal on Electrical Engineering and Informatics · 2019
Interviewer falsification is an important issue faced by institutions conducting censuses and surveys around the world, including Statistics Indonesia.This study discusses several methods to systematically detect interviewer falsification and validation using data mining techniques so that human supervisors can take further actions.After analyzing relevant features and conducting experiments, the results showed that unsupervised classification algorithm using simple 2-means clustering achieved 70.5% accuracy, while the supervised classification using logistic regression improved the accuracy to 88.5%.A greater level of accuracy is still needed to be pursued in further research, but the current results are certainly better than the traditional method which has almost no falsification detection method at all.