Cognitive Website


We are creating cognitive website design which will determine most suitable layout for the user according to user behavior on the first page. We are making use of neural network API which will learn the correlation and act accordingly as the users behave on the site.


We are using nod.js and JavaScript to create learning patterns for the machine by giving it certain input and output parameters. These parameters help to create the correlation between user activity on the website and his/her response for the consecutive pages. These adaptive websites provide flexibility according to user’s personality. This intuitive ability of the website helps retain user for longer duration and which are primarily beneficial in banking, retail and e-commerce sectors.


  1. We created multiple layouts for the training module of the neural network API.
  2. We implemented Mind Mapping API in node.js which learns from pattern recognition.
  3. We used JavaScript to collect input and output parameters.
  4. The input parameters are – i) Loading time for the website ii) Mouse travel time based on user’s first activity iii) Mouse travel length based on user’s first activity iv) Typing speed of the user count of letters per minute v) Typing speed of the user by words per minute
  5. Output parameters – i) Optional color selection ii) Available tab options iii) Layout design (less info. / more info.)
  6. Feedback collected – i) Click on the result ii) Scrolling down the page iii) Time spent on website
  7. Learning iterations for training are 50.
  8. Once the neural network is trained final layout is suggested by the API based on the input parameters explained in point 4.

How to use

  1. First default layout is presented to the user
  2. We observe input parameters based on simulation script (we designed) which iterates the website for 50 different scenarios
  3. These iterations train the network. The iterations include mix of random numbers and logically chosen numbers to maintain order.
Share this project: