Skip to Main content Skip to Navigation

Analyse des comportements des clients sur un site marchand en ligne

Abstract : In recent years, electronic (e)-commerce has revolutionized retail by changing the way customers shop. The user’s choices are no longer limited by the availability of products in stores in his or her area. Today, they can search and purchase products from any international store, anywhere, anytime, using e-commerce sites. However, e-merchants have some difficulty understanding the behavior of their customers: Why does a customer choose one product over another? Why do they prefer one brand over another? Why does he visit the same product more often? Why is the purchase made during the lunch break? What factors influence their choice and purchases? These behaviors are very complex because they are influenced by (1) factors internal to the e-commerce site such as prices, product availability, delivery times, product ratings, user opinions, etc., (2) and external factors, such as international events (e.g. Black Friday), holidays, the customer’s budget, etc., (3). To learn more about consumer behavior, a more in-depth analysis is required. A fine analysis of the customer’s behavior consists of understanding his or her path on the web site. These behaviors help e-merchants to understand and discover customer needs. As a result, e-merchants can recommend relevant products to their customers, anticipate future purchases and ensure product availability. Their main tasks are more focused on identifying and understanding the abandonment and purchase process, from the customer’s first stimulus to their decision. To achieve this objective, user actions could be tracked. These actions are known as ”clickstreams”. To analyze them and extract useful information from them, several operating tools have been developed. They aim to analyze the user’s path and discover their hidden behaviors. Some well-known data mining and machine learning techniques are pattern mining, trend discovery, customer grouping and classification, collaborative filtering in recommendation systems, and purchase prediction. In this thesis, we present two new models of user behaviors. The first model, called ”frequent behavior”, represents behaviors that are similar between several user sessions. To model them, we propose a new representation of frequent and temporal patterns, and to extract them we propose the pattern detection algorithm called SEPM (Sequential Event Pattern Mining) based on this new representation. The second pattern, called ”interesting behavior”, represents behaviors that are similar in the evolution of their characteristics over time. We model them by multivariate time series and we propose the GCBag framework (Generative time series Clustering with Bagging) to detect them. In addition to these two models, we present the complete analysis process of these models starting from the preparation of the data, through the application of the algorithms and up to the analysis of the resulting behaviors
Document type :
Complete list of metadata
Contributor : Abes Star :  Contact
Submitted on : Tuesday, October 26, 2021 - 4:12:11 PM
Last modification on : Wednesday, November 3, 2021 - 10:02:00 AM


Version validated by the jury (STAR)


  • HAL Id : tel-03404380, version 1


Mohamad Kanaan. Analyse des comportements des clients sur un site marchand en ligne. Modélisation et simulation. Université de Lyon, 2020. Français. ⟨NNT : 2020LYSE1227⟩. ⟨tel-03404380⟩



Record views


Files downloads