Skip to content

ValentinViennot/Much-Match

 
 

Repository files navigation

MuchMatch: Swipe it right!

Abstract

The near-unlimited choice of items presented in online stores leads users on journeys of untargeted and unsatisfying scrolling through bottomless pages on the search for products they really like.

With Much-Match, we introduce a totally new, intuitive and fun way of browsing shopping catalogues, leveraging a familiar technology: swiping. The backend algorithm of Much-Match continuously analyses what users swipe left and right and how they swipe it, creating a user profile that allows for meaningful and targeted suggestions.
Swipes are characterized measuring features like speed, reaction time or hesitation which provide an insight into people’s personal profile. Our preliminary analysis has shown that users indeed expose different general swiping behaviours. This enables two key mechanisms in our backend design: First, users can be matched to a user group and a behavioural pattern in a meaningful way. Second, swipes can be characterised related to how a user normally swipes, allowing to detect how much a user likes or dislikes a product.

Much-Match is a new way of navigating users straight to their purchase through swiping analysis. We give meaningful and targeted suggestions where user information seemed to be unavailable. This narrows down the vast choice in the online shopping world, leading to buying decisions that are quicker and more convenient.

Demo App

Based on determined features during our data collection and analysis, and for demonstration purposes during our final presentation, we developed an app to showcase our experiments with the previously described algorithm.

Although we were lacking a huge data collection and a proper data study, we were already able to give insightful data in our demonstration.

The demo app is a dashboard displaying computed features from collected data during a swipe gesture. This data is compared to a small set of swiping test data we were able to collect in two weeks. We tried to put a name on those features based on users feedback and from our analysis. We also wanted to show if a swipe was usual or not usual.
A phone can be connected through a websocket connection for a better experience.

Swiping analysis

Swiping analysis features

Detect unsusual behaviors

Impulsiveness detection

Interaction based on gesture analysis

Hesitation and commercial popup

Authors

  • Fristedt Andersson Anna Vera Linnea
  • Rüther Johannes Leonhard
  • Sørensen Mikkel Frisaek
  • Viennot Valentin

About

Dev resources for CS-489 class @ epfl - Demo using swiping data analysis to predict customer behavior

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 99.1%
  • Vue 0.5%
  • TypeScript 0.4%
  • JavaScript 0.0%
  • Shell 0.0%
  • HTML 0.0%