2
$\begingroup$

I have a set of user sessions. Session consists of an ordered list of types of actions that user made (for example, bought a gun, played a mission, etc). I want to create/calculate session that have most possible similarity to all provided sessions (most common types of actions users make in order they make them)

Unfortunately, I know nothing about data science but I tried to google a way to do that. I've found this document: https://cran.r-project.org/web/packages/TraMineR/vignettes/TraMineR-state-sequence.pdf And it looks like 9.1 and 9.2 describe things similar to what I want. But I dont know this for certain and even if it's true I stil don't know how to use it for my scenario.

Improve this question
$\endgroup$
1
  • $\begingroup$ Please accept as the answer if you happy with it. $\endgroup$
    – Snympi
    Commented Jun 15, 2017 at 21:35

1 Answer 1

Reset to default
4
$\begingroup$

One way would be not to approach this as a calculation per session. Most data science solutions like to end up with a number, probability or classification. I suggest you structure your data differently so that you try to answer the question - what next action is likely given the last action.

In order to do this you would have to restructure your session data and use information from across all your sessions. For example, if you compare in how many sessions a player 'buys a gun', and if so record over all those sessions what their next action is, e.g. in 60% they 'play a mission' next. You will then have a probability of their next action based on the number of choices players made in all those sessions.

Once you have those probabilities, you will be able to answer the question, 'What comes next?'. This will in turn enable you to build that most average session that you are after by stepping through a session and building it by the most probable next step.

Improve this answer
$\endgroup$

Not the answer you're looking for? Browse other questions tagged or ask your own question.