PERBANDINGAN KINERJA KOMBINASI METODE NORMALISASI DATA DAN FUNGSI KERNEL PADA SUPPORT VECTOR REGRESSION UNTUK PREDIKSI LAJU INFLASI INDONESIA

NUR ADZHANI, . (2026) PERBANDINGAN KINERJA KOMBINASI METODE NORMALISASI DATA DAN FUNGSI KERNEL PADA SUPPORT VECTOR REGRESSION UNTUK PREDIKSI LAJU INFLASI INDONESIA. Sarjana thesis, UNIVERSITAS NEGERI JAKARTA.

	Text COVER.pdf Download (2MB)
	Text BAB I PENDAHULUAN.pdf Download (259kB)
	Text BAB II LANDASAN TEORI.pdf Restricted to Registered users only Download (1MB) | Request a copy
	Text BAB III METODOLOGI.pdf Restricted to Registered users only Download (431kB) | Request a copy
	Text BAB IV HASIL DAN PEMBAHASAN.pdf Restricted to Registered users only Download (937kB) | Request a copy
	Text BAB V KESIMPULAN.pdf Restricted to Registered users only Download (251kB) | Request a copy
	Text DAFTAR PUSTAKA.pdf Download (262kB)
	Text LAMPIRAN.pdf Restricted to Registered users only Download (1MB) | Request a copy

Abstract

Support Vector Regression (SVR) merupakan salah satu metode machine learning yang banyak digunakan untuk memprediksi data kontinu, termasuk laju inflasi. Kinerja SVR dipengaruhi oleh pemilihan fungsi kernel dan metode prapengolahan data. Oleh karena itu, penelitian ini bertujuan membandingkan kinerja kombinasi metode normalisasi data dan fungsi kernel pada SVR dalam memprediksi laju inflasi Indonesia. Data yang digunakan berupa data inflasi bulanan Indonesia periode Mei 2016 hingga April 2026 sebanyak 120 observasi yang diperoleh dari Bank Indonesia. Data dibentuk menggunakan pendekatan lag sebanyak tiga periode, kemudian dibagi menjadi data pelatihan dan data pengujian dengan rasio 80:20. Penelitian membandingkan dua kondisi data, yaitu tanpa normalisasi dan menggunakan Min-Max Scaling, serta dua fungsi kernel, yaitu Radial Basis Function (RBF) dan Polinomial. Hiperparameter ditentukan menggunakan Grid Search Optimization, sedangkan kinerja model dievaluasi menggunakan Root Mean Square Error (RMSE) dan koefisien determinasi (R^2). Hasil penelitian menunjukkan bahwa kombinasi Min-Max Scaling dan kernel RBF memberikan kinerja terbaik dengan parameter optimal C = 100, ε = 0,01, dan γ = 0,01, menghasilkan RMSE pengujian sebesar 0,6040 dan R^2 sebesar 0,6098. Sebaliknya, penerapan Min-Max Scaling menurunkan kinerja kernel Polinomial. Hasil penelitian menunjukkan bahwa efektivitas normalisasi pada SVR bergantung pada karakteristik fungsi kernel untuk data inflasi Indonesia periode Mei 2016 hingga April 2026 yang digunakan dalam penelitian ini. ***** Support Vector Regression (SVR) is a machine learning method widely used for forecasting continuous data, including inflation rates. The performance of SVR is influenced by the choice of kernel function and data preprocessing technique. Therefore, this study aims to compare the performance of different combinations of data normalization methods and kernel functions in SVR for forecasting the Indonesian inflation rate. Monthly inflation data from Indonesia covering the period from May 2016 to April 2026, consisting of 120 observations obtained from Bank Indonesia, were used in this study. The time series data were transformed using a three-period lag approach and then divided into training and testing sets with an 80:20 ratio. Two data preprocessing schemes, namely no normalization and Min-Max Scaling, and two kernel functions, namely the Radial asis Function (RBF) and Polynomial kernels, were evaluated. Hyperparameters were otimized using Grid Search Optimization, while model performance was assessed using te Root Mean Square Error (RMSE) and the coefficient of determination (R^2). The results show that the combination of Min-Max Scaling and the RBF kernel achieved the best performance, with optimal parameters of C = 100, ε = 0.01, and γ = 0.01, yielding a testing RMSE of 0.6040 and an R2 of 0.6098. In contrast, Min-Max Scaling reduced the performance of the Polynomial kernel. These findings indicate that, for the Indonesian inflation data covering May 2016 to April 2026 used in this study, the effectiveness of data normalization depends on the characteristics of the selected kernel function.

Item Type:	Thesis (Sarjana)
Additional Information:	1. Drs. Sudarwanto, D.E.A., M.Si. ; 2. Dr. Lukita Ambarwati, S.Pd., M.Si.
Subjects:	Sains > Matematika
Divisions:	FMIPA > S1 Matematika
Depositing User:	Nur Adzhani .
Date Deposited:	18 Aug 2026 02:21
Last Modified:	18 Aug 2026 02:21
URI:	http://repository.unj.ac.id/id/eprint/71243

Actions (login required)

View Item