COLLABORATIVE FILTERING RECOMMENDATION SYSTEM USING APRIORI AND COSINE SIMILARITY FOR IMAGE COMBINATION ON MICROSTOCK

DWI, ANJAS (2020) COLLABORATIVE FILTERING RECOMMENDATION SYSTEM USING APRIORI AND COSINE SIMILARITY FOR IMAGE COMBINATION ON MICROSTOCK. S1-Sarjana thesis, Universitas Mercu Buana Jakarta.

Text (JURNAL MAHASISWA) 41515110104 Anjas Dwi.pdf Restricted to Registered users only Download (1MB)

Abstract

Google trend statistical data over the past 15 years indicates that the need for infographics increases and is directly proportional to technological development. This has led to the emergence of many sites that provide services for graphic design needs such as microstock. Infographics or graphic design requires a right combination of materials to produce clear delivery and to attract people's attention. This research was carried out in the hope that it can assist customers in determining a combination of good images or material design for graphic design projects according to the trend of most microstock customers, by focusing on the collaborative filtering recommendation system using the Apriori algorithm and Cosine similarity applied directly to microstock. The trial data used were 64 transaction data and tags from each image item collected directly from several designers who had selected the combination of image items for their graphic design projects. By setting a minimum support value of 50% and a minimum confidence value of 75%, 4 association rules were formed with the highest confidence value reaching 87.76%. The average execution time of forming an association rules is 0,0036 seconds. System performance, in performing cosine calculations with minimum similarity of 50% and recommending the results of image items, requires an average time of 1.862 seconds. Keywords: Recomendation System, Collaborative Filtering, Association Rule, Apriori, Cosine Similarity Data statistik Google Trend selama 15 tahun terakhir menunjukan bahwa kebutuhan inforgrafis meningkat dan berbanding lurus dengan perkembangan teknologi. Ini menyebabkan munculnya banyak situs yang menyediakan layanan untuk kebutuhan desain grafis seperti microstock. Infografis atau desain grafis membutuhkan kombinasi material yang tepat untuk menghasilkan penyampaian yang jelas dan menarik perhatian orang-orang. Penelitian ini dilakukan dengan harapan dapat membantu pelanggan dalam menentukan kombinasi gambar atau material desain yang bagus untuk proyek desain grafis sesaui trend sebagian besar pelanggan microstock dengan berfokus pada sistem rekomendasi Collaborative Filtering menggunakan algoritma Apriori dan Cosine similarity yang diterapkan langsung pada Microstock. Data ujicoba yang digunakan adalah 64 data transaksi yang dikumpulkan langsung dari beberapa desainer yang telah memilih kombinasi item gambar untuk proyek desain grafis mereka. Dengan menetapkan nilai minimum support 50% dan nilai minimum Confidence 75%, terbentuk sebanyak 4 Association rule dengan perolehan nilai Confidance tertinggi mancapai 87.76%. Performa sistem dalam pembentukan association rule membutuhkan waktu kurang lebih 0,0036 detik. Performa sistem dalam melakukan perhitungan cosine dengan minimum similarity 50% dan merekomendasikan hasil item gambar membutuhkan waktu rata-rata 1,862 detik. Kata kunci: Recomendation System, Collaborative Filtering, Association Rule, Apriori, Cosine Similarity

Actions (login required)

View Item