University of Twente Student Theses

Login

Predicting purchasing intent of website visitors with deep feature learning

Hoek, J.M. van (2020) Predicting purchasing intent of website visitors with deep feature learning.

This is the latest version of this item.

[img] PDF
711kB
Abstract:In E-commerce, there is a growing interest in understanding website visitors. Being able to accurately predict when a visitor has an intention to make a purchase or not can be very valuable for e-commerce businesses. Several studies have aimed to make these predictions based on machine learning algorithms. Different datasets were used to test these methods, often consisting of data about a particular website or behavior of visitors. However, most of these methods only use raw data to predict the buying behavior of visitors. It is not yet clear if that is the most effective way to make these predictions. The objective of this paper is to test this issue by using deep feature learning to predict the purchasing intent of website visitors, based on website data, clickstream data and session data. Results of four different autoencoders in combination with a SVM/WSVM classifier is compared to a SVM/WSVM that only uses raw data. Of all the autoencoders, the deep autoencoder shows the best results with an average accuracy of 0,61 and an average TPR of 0,71. However, traditional methods used on this dataset like a WSVM with raw data and decision trees significantly outperform the autoencoders. Nevertheless, the use of autoencoders in the prediction of buying behavior can potentially become a lot more effective when improvements are made to the configuration of the autoencoders.
Item Type:Essay (Bachelor)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science
Programme:Business & IT BSc (56066)
Link to this item:https://purl.utwente.nl/essays/82211
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page