Aulia, Rya Sofi (2017) Penentuan Panjang Optimal Data Deret Waktu Bebas Outlier Dengan Metode Window Time. Undergraduate thesis, Institut Teknologi Sepuluh Nopember.
Preview |
Text
1313100064_Undergraduate_Theses.pdf - Published Version Download (3MB) | Preview |
Abstract
Data outlier sering kali mempengaruhi model data secara umum sehingga pengaruh dari data outlier tersebut harus dikurangi atau dihilangkan. Namun, di sisi lain outlier merupakan data yang sangat informatif apabila penyebab adanya outlier tersebut diketahui sehingga beberapa penelitian merekomendasikan untuk tidak menghilangkan outlier namun mengganti model awal dengan model baru yang disisipkan dengan model outlier. Kemunculan outlier dapat menyebabkan bias yang cukup serius dalam estimasi parameter. Atas dasar penelitian-penelitian yang dilakukan sebelumnya maka pada penelitian ini dilakukan metode baru untuk mendeteksi outlier. Tujuan dari metode ini adalah untuk mendapatkan panjang data optimum yang bisa digunakan untuk mendeteksi data outlier. Penelitian ini terfokus pada pendeteksian outlier pada data deret waktu dengan jumlah data yang banyak. Dari hasil simulasi data dan implementasi yang dilakukan pada data riil didapatkan hasil bahwa window time 500 dan 1000 memberikan nilai akurasi deteksi outlier lebih baik dibandingkan dengan window time 100. Selain itu, metode deteksi menggunakan window time memberikan hasil yang lebih baik dibandingkan metode deteksi outlier biasa.
==================================================================================================================
Data outliers often affects the common data model so that the effect of the outlier data is to be reduced or eliminated. However, on the other hand outlier is the data that is very informative if the cause of the outlier is known that several studies recommend not eliminate outliers but replace the initial model with a new model that is inserted with the model outliers. The emergence of outliers can cause quite serious bias in the estimation of parameters. Based on the studies conducted previously, this research carried out new methods for detecting outliers. The purpose of this method is to obtain the optimum length of the data that can be used to detect the data outliers. This research is focused on outlier detection in time series data with large amounts of data. From the simulation results and the implementation of data on real data showed that window time 500 and 1000 deliver the accuracy of outlier detection is better than the window time 100. In addition, the detection method using the window time gives better results than usual outlier detection method.
| Item Type: | Thesis (Undergraduate) |
|---|---|
| Uncontrolled Keywords: | Data Bebas Outlier, Free Data Outliers, Outlier, Window Time |
| Subjects: | H Social Sciences > HA Statistics |
| Divisions: | Faculty of Mathematics and Science > Statistics > 49201-(S1) Undergraduate Thesis |
| Depositing User: | - RYA SOFI AULIA |
| Date Deposited: | 13 Apr 2017 08:07 |
| Last Modified: | 08 Mar 2019 04:18 |
| URI: | http://repository.its.ac.id/id/eprint/3531 |
Actions (login required)
 |
View Item |