Mining Gold and Silver Using Machine Learning: Predicting Cancelled Orders at the Online Store of Tavex Norway

Liisa Kallas
Widespread usage of Internet has given rise to huge amounts of online platforms, among which are e-commerce sites. Many traditionally brick-and-mortar stores, like precious metals dealers, have also opened e-commerce platforms to sell their products. Moving business online has benefits of reaching a wider audience and can be more convenient for clients. On the other hand, it has also brought along new business challenges like how to get more website users to purchase and the attempt to predict which website visitors are going to place an order during their visit. This has been a popular topic of many research papers. A less covered area in previous studies is the problem of online orders that had been already placed, but then cancelled. The cancellation problem that happens due to customers not paying for their order has been also noted in the e-commerce site of a precious metals dealer. In this work, a data analytics approach is taken to predict, whether an online order will be cancelled or not after its placement at Tavex Gull og Sølv AS. Three supervised machine algorithms, namely logistic regression, random forest and support vector machine, were used. The best achieved model had the ability to identify 68% of all unpaid orders. These results indicate that data analytics can be very promising in predicting these types of orders.
Graduation Thesis language
Graduation Thesis type
Master - Conversion Master in IT
Rajesh Sharma
Defence year