Penggalian Top-K Frequent Closed Constrained Gradient ltemsets pada Basis Data Retail

Absari, Dhiani Tresna (2009) Penggalian Top-K Frequent Closed Constrained Gradient ltemsets pada Basis Data Retail. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 5106201012-Master_thesis.pdf Restricted to Repository staff only Download (9MB)

Abstract

Analisis asosiasi tanpa penggunaan parameter minimum support antar itemsets dalam satu set transaksi pada basis data retail telah banyak dilakukan dalam penelitian. Penggalian top-k frequent closed itemsets dengan algoritma TFP tanpa menggunakan minimum support merupakan salah satu penelitian yang menarik untuk diaplikasikan dalam analisis asosiasi. Di sisi lain, dalam dunia retail, penggunaan parameter batasan gradient perlu dilibatkan dalam analisis asosiasi agar dapat melakukan penyaringan itemset dengan mengacu pada suatu gradient tertentu. Gradient yang dimaksudkan di sini dapat berupa suatu ukuran tertentu, seperti nilai keuntungan, sebagaimana dilakukan dalam algoritma Frequent Closed Constrained Gradient Mining (FCCGM). Namun demikian, FCCGM masih mengharuskan pengguna untuk menentukan nilai minimum support, sehingga penggabungan algoritma TFP dan FCCGM menjadi menarik untuk dilakukan. Dalam penelitian ini, dilakukan modifikasi terhadap algoritma TFP agar top-k frequent closed itemsets dengan batasan gradient tanpa penentuan nilai minimum support dapat dibangkitkan. Hasil modifikasi algoritma, yang disebut top-k frequent closed constrained gradient itemsets, melibatkan beberapa langkah tambahan terhadap algoritma TFP. Tambahan langkah tersebut antara lain berupa pembangkitan nilai minimum support secara dinamis untuk membebaskan pengguna dalam menginisialisasi nilai minimum support dan langkah perhitungan gradient untuk masing-masing item yang terdapat dalam basis data. Selain itu, modifikasi terhadap tabel global header dilakukan sehingga dapat menyimpan hasil perhitungan gradient untuk masing-masing item-nya. Langkah pemangkasan gradient pada FP-Tree yang terbentuk, yaitu pemangkasan item yang tidak memenuhi batasan gradient, dilakukan tepat sebelum proses penggalian frequent closed itemset dijalankan. Terakhir, modifikasi terhadap langkah pembangkitan hasil penggalian frequent closed itemset dilakukan agar hasil akhir dari frequent closed itemsets dapat disimpan secara terurut berdasarkan nilai gradient dan support tertinggi. Algoritma yang dikembangkan telah berhasil diimplementasikan dalam lingkungan sistem operasi Windows. Hasil uji coba menunjukkan bahwa program yang dibuat mampu membangkitkan top-k frequent closed itemset dengan nilai gradient dan support tertinggi tanpa batasan minimum support pada basis data retail. Waktu komputasi yang diperlukan untuk membangkitkan top-k frequent closed itemset bergantung pada kedalaman FP-Tree yang terbentuk dari transaksi yang dianalisis; semakin dalam FP-Tree terbentuk, maka semakin besar waktu komputasi yang diperlukan untuk melakukan proses penggalian.
==================================================================================================================================
Association analysis without a minimum support threshold among itemsets within a transaction set in a retail database has been the topic of many research works. The investigation of top-k frequent closed itemsets mining using the TFP algorithm without a minimum support threshold is an interesting application in association analysis. Meanwhile, in the retail domain, incorporating a gradient threshold is necessary to filter itemsets based on a specific gradient measure. This gradient can represent a particular metric, such as benefit value, as implemented in the Frequent Closed Constrained Gradient Mining (FCCGM) algorithm. Nevertheless, since FCCGM still requires users to determine a minimum support value, combining the TFP and FCCGM algorithms becomes an intriguing research endeavor. In this research, the TFP algorithm is modified to extract top-k frequent closed itemsets with gradient constraints that require no minimum support value. The resulting modified algorithm, termed top-k frequent closed constrained gradient itemsets, employs several additional steps over the original TFP algorithm. These additional steps include the dynamic generation of a minimum support value—freeing users from manual initialization—and the calculation of a gradient for each item in the database. Furthermore, the global header table is modified to store the gradient calculation results for each item. Gradient pruning on the constructed FP-Tree, which removes items failing to meet the gradient constraints, is conducted prior to extracting the frequent closed itemsets. Finally, the step for extracting frequent closed itemsets is modified so that the final outcome is stored in descending order based on gradient and support values. The developed algorithm has been successfully implemented and tested under the Windows operating system. Experimental results demonstrate that the modified algorithm is capable of extracting top-k frequent closed itemsets with the highest gradient and support values without minimum support constraints in retail databases. The computing time required for mining top-k frequent closed itemsets depends on the depth of the resulting FP-Tree constructed from the analyzed transactions; the deeper the FP-Tree, the longer the computing time required.

Item Type:	Thesis (Masters)
Additional Information:	RTIf 005.1 Abs p
Uncontrolled Keywords:	bisnis retail, analisis asosiasi, top-k frequent closed itemsets, batasan gradient, retail business, association analysis, top-k frequent closed itemsets, gradient constraints
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5105.546 Computer algorithms
Divisions:	Faculty of Information Technology > Informatics Engineering > 55101-(S2) Master Thesis
Depositing User:	magang .
Date Deposited:	18 Sep 2026 06:47
Last Modified:	18 Sep 2026 06:47
URI:	http://repository.its.ac.id/id/eprint/144681

Actions (login required)

View Item