Publication: Beyaz eşyalar için yapay zeka algoritmaları ile fiyat-performans analizi
Abstract
Günümüzdeki teknolojik çalışmalara dikkat edildiğinde yapay zekanın ve makine öğrenmesinin önemi inkar edilemez derecededir. Bu tez çalışmasında da beyaz eşya özelinde yapay zekanın ve makine öğrenmesi algoritmalarının nasıl kullanılabileceği incelenmiştir. İlgili sitelerden çekilen güncel teknik veriler ile güncel teknik verileri çekilen ürünlere ait fiyatların belirtildiği sitelerden fiyat bilgileri elde edilerek, beyaz eşyaların sahip olduğu teknik özellikler ile fiyatları arasındaki ilişkiler optimize edilmiş, modellenmiş ve bazı çıkarımlarda bulunulmuştur. Fiyat – performans modellemesi 4 ayrı beyaz eşya grubunda yapılmıştır. Bunlar sırasıyla kurutma makinası, çamaşır makinası, bulaşık makinası ve buzdolabıdır. Belirtilen beyaz eşyaların bazı teknik verileri veri kazıma işlemine izin verilen ilgili sitelerden çekilip veritabanında tutulmuştur ve bu veriler kapsamında her bir beyaz eşya grubu özelinde müşteri için önemli olacak özellikler modellemeye dahil edilmiştir. Python programlama dilinde de bu modelleme görsel hale getirilmiş ve arayüz oluşturulmuştur. Oluşturulan arayüzde, her bir beyaz eşyada müşteri için en önemli teknik özellikler müşterinin seçimine bırakılmıştır. Seçilen teknik özellikler bağlamında, elde edilen güncel veriler baz alınarak, müşteriye istemiş olduğu beyaz eşya özelliklerine göre ortalama bir fiyat gösterilmektedir. Ayrıca müşteri arayüzde hangi markayı seçtiyse o markayla ilgili veri tabanındaki ürünler müşteri karşısına çıkmaktadır ve müşteri istediği ürüne tıklayarak ilgili ürünün sitesine yönlendirilmektedir. Müşteri hangi ürünün sitesine yönlendirilmek istiyorsa tıkladığı butondaki ürünün genişlik, yükseklik, derinlik ölçüsü bilgileri cm cinsinden ve ürünün sahip olduğu kapasite kg cinsinden gösterilmektedir. Belirtilen optimizasyon çalışmaları birçok makine öğrenmesi algoritmasıyla gerçekleştirilmiştir. Tahmin algoritmaları ve sınıflandırma algoritmaları, modelleme için ayrı ayrı değerlendirilmiştir. Algoritmaların içeriğinde yer alan bazı parametreler değiştirilerek aynı algoritmanın aynı veri seti üzerinde modelleme performanslarındaki değişimi de gözlemlenmiştir. Kısaca, her algoritmanın tezde belirtilen 4 beyaz eşya grubu için modelleme ve optimizasyon yeteneği gözlemlenmiştir.
When considering today's technological studies, the importance of artificial intelligence and machine learning is undeniable. In this thesis study, how artificial intelligence and machine learning algorithms can be used specifically for white goods was examined. By obtaining current technical data from relevant sites and price information from sites where the prices of products whose updated technical data was taken were obtained, the relationships between the technical features of white goods and their prices have been optimized, modeled, and some inferences have been made. Price – performance modeling was performed in 4 separate white goods groups. These are drying machine, washing machine, dishwasher and refrigerator, respectively. Some technical data of the specified white goods were obtained from relevant sites where data scraping is allowed and kept in the database and in the context of this data features that will be important to the customer for each white goods group have been included in the modeling. This modeling has been made visual in the Python programming language and interface has been created. In the created interface, the most important features for each white goods were preferred by customers. In the context of the selected technical features, based on the current data obtained, an average price is shown to the customer according to the white goods features they wants. In addition, whichever brand has been selected by customer in the interface, the products in the database related to that brand are displayed to the customer and by clicking on the product they want, the customer is directed to the website of the relevant product. After the customers clicks on the button for the product which they wants to be directed, the product's width, length and depth informations are displayed as cm and the capacity of the product is displayed as kg. The specified optimization studies were carried out using many machine learning algorithms. Prediction algorithms and classification algorithms were evaluated separately for modelling. By changing some parameters in the content of the algorithms, the change in modeling performance of the same algorithm on the same data set was also observed. Briefly, the modeling and optimization ability of each algorithm for the 4 white goods groups mentioned in the thesis has been observed.
When considering today's technological studies, the importance of artificial intelligence and machine learning is undeniable. In this thesis study, how artificial intelligence and machine learning algorithms can be used specifically for white goods was examined. By obtaining current technical data from relevant sites and price information from sites where the prices of products whose updated technical data was taken were obtained, the relationships between the technical features of white goods and their prices have been optimized, modeled, and some inferences have been made. Price – performance modeling was performed in 4 separate white goods groups. These are drying machine, washing machine, dishwasher and refrigerator, respectively. Some technical data of the specified white goods were obtained from relevant sites where data scraping is allowed and kept in the database and in the context of this data features that will be important to the customer for each white goods group have been included in the modeling. This modeling has been made visual in the Python programming language and interface has been created. In the created interface, the most important features for each white goods were preferred by customers. In the context of the selected technical features, based on the current data obtained, an average price is shown to the customer according to the white goods features they wants. In addition, whichever brand has been selected by customer in the interface, the products in the database related to that brand are displayed to the customer and by clicking on the product they want, the customer is directed to the website of the relevant product. After the customers clicks on the button for the product which they wants to be directed, the product's width, length and depth informations are displayed as cm and the capacity of the product is displayed as kg. The specified optimization studies were carried out using many machine learning algorithms. Prediction algorithms and classification algorithms were evaluated separately for modelling. By changing some parameters in the content of the algorithms, the change in modeling performance of the same algorithm on the same data set was also observed. Briefly, the modeling and optimization ability of each algorithm for the 4 white goods groups mentioned in the thesis has been observed.