Sprint 2 – Diverge - AgileBusinessAnalysis/01_TEAM GitHub Wiki

Analyse recommendation concept possibility

For the recommendation concept on the basis of the captured data, two possible solutions were investigated. On the one hand, user-based collaborative filtering, which bases its recommendation on user similarities. And on the other hand, item-based collaborative filtering, which bases its recommendation on the similarities of already made decisions.

User-based collaborative filtering

User-based collaborative filtering focuses on user attributes and recommends tipps on the basis of user attribute similarities.

The example shows that the upper and bottom user do have attributes with high similarities. Due to this fact and the fact that the upper user does like tipp B, this tipp will be recommended to the bottom user. Chances are high, that due to same attributes, they also have a similar taste when it comes to tipps.

Item-based collaborative filtering

Item-based collaborative filtering does not focus on whether users do match each other, it does focus only on conducted item decisions.

As indicated with the blue circle, if users do make the same tipp decision(s) (both users do accept tipp D), the concept assumes that in this example, the user on the bottom also would like tipp B and would dislike tipp A, similar to the other user who liked tipp D before. Normally, there would be much more data and therefore much more decisions behind such a created insight (recommend tipp B to bottom user), but due to simplification reasons, this example used a limited amount of data visualised.