Perbandingan Kinerja Model Prediksi Algoritma Regresi Linear dan K-Nearest Neighbors dalam Kasus Jumlah Pelanggan Telekomunikasi

Aldewo Dillon Perkasa, - (2023) Perbandingan Kinerja Model Prediksi Algoritma Regresi Linear dan K-Nearest Neighbors dalam Kasus Jumlah Pelanggan Telekomunikasi. S1 thesis, Universitas Pendidikan Indonesia.

Abstract

Dalam ilmu komputer dan data science, perbandingan antara algoritma - algoritma memiliki peran penting. Penelitian ini membandingkan performa dua algoritma untuk prediksi atau forecasting, dengan data dari Badan Pusat Statistik mengenai Jumlah Pelanggan Wired dan Wireless Telekomunikasi. Evaluasi dilakukan dengan metrik seperti Mean Absolute Deviation (MAD) untuk regresi linear dan k-nearest neighbors. Hasilnya, MAD regresi linear: 1.520 (wired) dan 46.140 (wireless), serta k - nearest neighbors: 827 (wired) dan 35.723 (wireless). Mean Square Error (MSE) regresi linear: 27.850 (wired) dan 24.907.799 (wireless), k-nearest neighbors: 10.282 (wired) dan 21.165.042 (wireless). Root Mean Square Error (RMSE) regresi linear: 1.668 (wired) dan 49.907 (wireless), k-nearest neighbors: 1.014 (wired) dan 46.005 (wireless). Mean Absolute Percentage Error (MAPE) regresi linear: 15,10% (wired) dan 13,48% (wireless), k-nearest neighbors: 8,57% (wired) dan 9,87% (wireless). Hasil menunjukkan k-nearest neighbors lebih unggul daripada regresi linear dalam prediksi. Algoritma ini, yang utamanya berfungsi sebagai klasifikasi, dapat mengungguli regresi linear dalam tugas prediksi atau forecasting.
-----
In computer science and data science, the comparison between algorithms plays a crucial role. This research compares the performance of two algorithms for prediction or forecasting using data from the Central Statistics Agency (Badan Pusat Statistik) regarding the number of Wired and Wireless Telecommunication Customers. Evaluation is conducted using metrics such as Mean Absolute Deviation (MAD) for linear regression and k-nearest neighbors. The results show that for linear regression, MAD is 1,520 (wired) and 46,140 (wireless), while for k-nearest neighbors, it is 827 (wired) and 35,723 (wireless). Mean Square Error (MSE) for linear regression is 27,850 (wired) and 24,907,799 (wireless), and for k-nearest neighbors, it is 10,282 (wired) and 21,165,042 (wireless). Root Mean Square Error (RMSE) for linear regression is 1,668 (wired) and 49,907 (wireless), and for k-nearest neighbors, it is 1,014 (wired) and 46,005 (wireless). Mean Absolute Percentage Error (MAPE) for linear regression is 15.10% (wired) and 13.48% (wireless), while for k-nearest neighbors, it is 8.57% (wired) and 9.87% (wireless). The results indicate that k-nearest neighbors outperform linear regression in prediction. This algorithm, primarily functioning as classification, can surpass linear regression in prediction or forecasting tasks.

	Text S_SISTEL_1905342_Title.pdf Download (416kB)
	Text S_SISTEL_1905342_Chapter1.pdf Download (141kB)
	Text S_SISTEL_1905342_Chapter2.pdf Restricted to Staf Perpustakaan Download (656kB)
	Text S_SISTEL_1905342_Chapter3.pdf Download (210kB)
	Text S_SISTEL_1905342_Chapter4.pdf Restricted to Staf Perpustakaan Download (385kB)
	Text S_SISTEL_1905342_Chapter5.pdf Download (114kB)
	Text S_SISTEL_1905342_Appendix.pdf Restricted to Staf Perpustakaan Download (229kB)

Item Type:	Thesis (S1)
Additional Information:	Link Google Scholar: Ichwan Nul Ichsan : https://scholar.google.co.id/citations?user=uIGskD0AAAAJ&hl=en Hafiyyan Putra Pratama : https://scholar.google.com/citations?user=tQe1410AAAAJ&hl=en ID SINTA: Ichwan Nul Ichsan : 6721201 Hafiyyan Putra Pratama : 6681148
Uncontrolled Keywords:	Machine Learning, Linear Regression, K-Nearest Neighbors, Prediction, Evaluation Metrics
Subjects:	L Education > LC Special aspects of education T Technology > T Technology (General)
Divisions:	UPI Kampus Purwakarta > S1 Sistem Telekomunikasi
Depositing User:	Aldewo Perkasa
Date Deposited:	01 Sep 2023 08:29
Last Modified:	01 Sep 2023 08:29
URI:	http://repository.upi.edu/id/eprint/100941

Actions (login required)

View Item