BibTex Citation Data :
@article{JFMA11691, author = {Noviana Pratiwi and Yudi Setyawan}, title = {ANALISIS AKURASI DARI PERBEDAAN FUNGSI KERNEL DAN COST PADA SUPPORT VECTOR MACHINE STUDI KASUS KLASIFIKASI CURAH HUJAN DI JAKARTA}, journal = {Journal of Fundamental Mathematics and Applications (JFMA)}, volume = {4}, number = {2}, year = {2021}, keywords = {SVM, Kernel, linear, polynomial, Gauss}, abstract = { Abstrak. Penelitian ini difokuskan pada perbandingan beberapa fungsi kernel, cost dan proporsi data training pada Support Vector Machine terhadap akurasi pengklasifikasian curah hujan di Jakarta. Fungsi-fungsi kernel linier, Gauss dan polynomial digunakan untuk memodifikasi metode Support Vector Machine guna menyelesaikan kasus nonlinier yang sering terjadi pada kondisi real. Variabel yang digunakan dalam penelitian ini meliputi temperatur, kelembaban, penyinaran matahari dan kecepatan angin. Hasil analisis menunjukkan bahwa nilai support vector terkecil tidak memberikan akurasi yang tertinggi pada masing-masing fungsi kernel. Selain itu, proporsi dataset (training:testing) sebesar 90%:10% memberikan akurasi sedikit lebih tinggi dibandingkan dengan akurasi untuk proporsi 80%:20% untuk masing-masing fungsi kernel. Secara keseluruhan, akurasi tertinggi diperoleh pada proporsi 90%:10% oleh fungsi kernel linier dan polinom untuk cost 1 dan 1000 secara bersamaan yaitu 78,38%. Kata Kunci : Cost , Gauss, Kernel, linear, polynomial, Abstract . This research focuses on the comparison of several kernel functions, costs and proportions of data training on the Support Vector Machine to the accuracy of classifying rainfall in Jakarta. The linear, Gaussian and polynomial kernel functions were applied to modify the Support Vector Machine method to solve non-linear cases that often occur in actual conditions. The variables used in this study comprised of temperature, humidity, sunlight and wind speed. The analysis disclosed that the smallest support vector value did not provide the highest accuracy value for each kernel. In addition, the proportion of the dataset (training:testing) of 90%:10% provided a slightly higher accuracy compared to the accuracy for the proportion of 80%:20% for each kernel function. Overall, the highest accuracy attained at the proportion of 90%:10% by linear and polynomial kernel functions for cost 1 and 1000 simultaneously, which was 78.38%. }, issn = {2621-6035}, pages = {203--212} doi = {10.14710/jfma.v4i2.11691}, url = {https://ejournal2.undip.ac.id/index.php/jfma/article/view/11691} }
Refworks Citation Data :
Abstrak. Penelitian ini difokuskan pada perbandingan beberapa fungsi kernel, cost dan proporsi data training pada Support Vector Machine terhadap akurasi pengklasifikasian curah hujan di Jakarta. Fungsi-fungsi kernel linier, Gauss dan polynomial digunakan untuk memodifikasi metode Support Vector Machine guna menyelesaikan kasus nonlinier yang sering terjadi pada kondisi real. Variabel yang digunakan dalam penelitian ini meliputi temperatur, kelembaban, penyinaran matahari dan kecepatan angin. Hasil analisis menunjukkan bahwa nilai support vector terkecil tidak memberikan akurasi yang tertinggi pada masing-masing fungsi kernel. Selain itu, proporsi dataset (training:testing) sebesar 90%:10% memberikan akurasi sedikit lebih tinggi dibandingkan dengan akurasi untuk proporsi 80%:20% untuk masing-masing fungsi kernel. Secara keseluruhan, akurasi tertinggi diperoleh pada proporsi 90%:10% oleh fungsi kernel linier dan polinom untuk cost 1 dan 1000 secara bersamaan yaitu 78,38%.
Kata Kunci : Cost, Gauss, Kernel, linear, polynomial,
Abstract. This research focuses on the comparison of several kernel functions, costs and proportions of data training on the Support Vector Machine to the accuracy of classifying rainfall in Jakarta. The linear, Gaussian and polynomial kernel functions were applied to modify the Support Vector Machine method to solve non-linear cases that often occur in actual conditions. The variables used in this study comprised of temperature, humidity, sunlight and wind speed. The analysis disclosed that the smallest support vector value did not provide the highest accuracy value for each kernel. In addition, the proportion of the dataset (training:testing) of 90%:10% provided a slightly higher accuracy compared to the accuracy for the proportion of 80%:20% for each kernel function. Overall, the highest accuracy attained at the proportion of 90%:10% by linear and polynomial kernel functions for cost 1 and 1000 simultaneously, which was 78.38%.
Article Metrics:
Last update:
Authors who publish articles in this journal agree to the following terms:
For more detailed information about the copyright transfer, please refer to this page: COPYRIGHT TRANSFER FORM