Session stitching using sequence fingerprinting for web page visits

The nature of people's web navigation has significantly changed in recent years. advent smartphones and other handheld devices given rise to users consulting websites with more than one device, or using a shared device. As result, large volumes seemingly disjoint data are available, which when analysed together can support decision-making. task identifying sessions by linking such back specific person, however, is hard. idea session stitching aims overcome this machine learning inference identify similar identical users. Many efforts use various demographic device-based features train matching algorithms. However, often these variables not available for every dataset recorded differently, making streamlined setup difficult. Besides, they result vast feature spaces hard actionable interpretation. In paper, we present an alternative approach based on the fingerprinting pages visited single session. By behavioural patterns from sequences page visits, obtain that be used without requiring sensitive user-agent as IP, geo location, device details common approaches. Using sequential fingerprints does rely pre-defined features, but only requires recording our actionable. empirically tested real-life logs compared regular state-of-the-art embedding techniques. Results ecommerce context show still strong performance fewer facilitating decision-making inform subsequent related activities marketing customer analysis.