Text
Kinerja algoritma Kmeans++ pada pengelompokkan dokumen teks pendek pada abstrak di jurusan teknik elektro fakultas teknik UNJ
KINERJA ALGORITMA KMEANS++ PADA PENGELOMPOKKAN DOKUMEN TEKS PENDEK PADA ABSTRAK DI JURUSAN TEKNIK ELEKTRO FAKULTAS TEKNIK UNJ
CATUR RAHMA SISTIANI
ABSTRAK
Pengelompokkan pada dokumen teks pendek masih sulit ini dikarenakan di sparsity kata. Tujuan penelitian ini adalah untuk mengetahui kinerja algoritma kmeans++ pada teks pendek dan untuk mengetahui proses pengelompokkan algoritma k-means++ pada teks pendek di abstrak skripsi Jurusan Teknik Elektro Fakultas Teknik UNJ dilaksanakan pada semester genap tahun ajaran 2014-2015. Penelitian ini menggunakan metode penelitian eksperimen. Data abstrak yang digunakan sebanyak 200 abstrak. Penelitian meneliti 4 data yaitu Data pertama adalah abstrak ilmiah di jurusan Teknik Elektro, Universitas Negri Jakarta pada paragraf 1 sampai paragraf 3. Data kedua adalah paragraf 1 pada abstrak ilmiah di jurusan Teknik Elektro, Universitas Negri Jakarta. Data ketiga adalah paragraf 2 pada abstrak ilmiah di jurusan Teknik Elektro, Universitas Negri Jakarta. Data keempat adalah paragraf 3 pada abstrak ilmiah di jurusan Teknik Elektro, Universitas Negri Jakarta. Pengujian kinerja algoritma k-means++ menggunakan matrix confusion. Berdasarkan hasil penelitian, didapatkan kesimpulan bahwa keakurasian pada abstrak, paragraf 1 di abstrak, paragraf 2 di abstrak, dan paragraf 3 di abstrak mencapai lebih dari 80% . Didapatkan juga kesesuaian antar data yang diprediksi dengan hasil yang benar dari data yang sebenarnya(presisi) pada abstrak, paragraf 1 di abstrak, paragraf 2 di abstrak, dan paragraf 3 di abstrak mencapai lebih dari 50% . Didapatkan juga peluang munculnya data relevan yang diambil sesuai dengan query (recall) pada abstrak, paragraf 1 di abstrak, paragraf 2 di abstrak, dan paragraf 3 di abstrak mencapai lebih dari 80%.
Kata kunci: Algoritma kmenas++, Teks Pendek, Matrix Confusion
KMEANS++ ALGORITHM PERFORMANCE SHORT TEXT DOCUMENT GROUPING IN THE ABSTRACT OF ELECTRICAL ENGINEERING FACULTY OF ENGINEERING UNJ
CATUR RAHMA SISTIANI
ABSTRACT
The purpose of this research is to know the performance of kmeans++ algorithm to group short text in undergraduate thesis abstract of electrical engineering faculty of engineering UNJ held on even semester academic year 2014-2015. This research uses experimental research methods. There are 200 abstracts used in this research. The research examined the data first Data i.e. 4 is the scientific abstracts in the Department of electrical engineering, University of Negri Jakarta on paragraphs 1 to 3 paragraphs. The second is paragraph 1 data on scientific abstracts in the Department of electrical engineering, University of the Country. The third paragraph is 2 data on scientific abstracts in the Department of electrical engineering, University of the Country. The fourth paragraph is 3 data on scientific abstracts in the Department of electrical engineering, University of the Country. Performance testing algorithm for k-means ++ using the confusion matrix. Based on the results of the study, obtained on abstract accuracy by 90%, accuracy in paragraph 1 in the abstract of 88%, accuracy in paragraph 2 in the abstract of 84%, accuracy in paragraph 3 in the abstract of 91%. The conclusion is a good degree of accuracy is in paragraph 3 in the abstract thesis Department of electrical engineering faculty of engineering UNJ amounted to 91%. Kmeans algorithm uses classification accuracy ++ on the abstract thesis Department of electrical engineering faculty of engineering UNJ, paragraraf 1pada abstract thesis Department of electrical engineering faculty of engineering UNJ, paragraph 2 on the abstract thesis Department of electrical engineering faculty of engineering UNJ and paragraph 3 on the abstract thesis Department of electrical engineering faculty of engineering UNJ can reach more than 80%.
Keywords: algorithm kmenas ++, short text, the confusion matrix
Bibliografi : lembar 49-51
SS00006534 | SK 6534 | UPT Perpustakaan UNJ | Tersedia |
Tidak tersedia versi lain