Can anyone please elaborate the difference between derived and collective data ?

Broadly speaking the difference is as stated in OSM's Legal FAQ:

3d. If I use your data together with someone else's data, do I have to apply your license to their data too?

If the two datasets are independent, no, you don't; this is a Collective Database.

If you adapt them to work together (for example, by taking footpaths from the OSM data, roads from the third-party data, and connecting them for routing), this is a Derivative Database and so you must (as per 3b).

However, if the two datasets are matched "trivially" by, for example, automated matching using a simple criterion such as name/locality, this is not "substantial" and remains a Collective Database.

There is no hard and fast rule: copyright law rarely works in black and white. However, if you have specific usage queries, then do post them here and people can suggest what would apply. (Note that no-one here is able to give formal legal advice or speak officially for the OSM Foundation.)

The most important thing to remember is that you cannot get legal advice from the OpenStreetMap community, so no matter what answers you get here (including this one), if you need to be sure you should employ a lawyer to look at your precise circumstances. A slight variation that seems insignificant to you could alter the legal position of your use.

However, the general rule is that if you can superimpose a separate dataset over OpenStreetMap data (often by compositing after rendering), and you do not change the OpenStreetMap data in any way, that is usually considered a produced work.

If, however, your use of the data requires you to combine OpenStreetMap data with another source (e.g. a set of geodata containing streets that aren't in OSM) then that is a derived work.

In either case you still need to credit OpenStreetMap as the source of your data.

Remember: this is not legal advice.