We arr trying to develop an algorithm to calculate the average usage of a feature on a website. e.g. usage of search button on google page.

Approach 1- Average usage = count of feature use/ count of time users landed on the page This approach has issue as a single user's data can manipulate the overall average i.e. a outlier can occur. e.g. 10 users had used the user 20 times out of 100 but a user used 100 time out of 100. So, the average will not be giving the correct picture.

Approach 2- Fiding Mode of the usage percentage and adjusting the usage which are greater than the mode. Than finding out the percentage of all the data, in this scenario we achieve the problem due to outliers.

The problem with this approach is that data is too static i.e. not dependent on time. Like if a feature usage increased recently it should have more weightage while calculating feature usage percentage.

So, my question is what algorithm or if any tool available for calculating the feature usage percentage considering time as a function.

  • 2
    $\begingroup$ check reddit´s hot post algorithm, it relies on decayed waights over time $\endgroup$ – juvian Aug 12 at 16:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.