There are multiple places the data can come from. Local NWS offices, several universities provide models (Texas A&M is the one I'm most familiar with, pretty sure there are frequently used ones out of Florida and Ohio, too). The companies that provide the graphic computers also have internal models (WSI and WeatherCentral, WSI bought WeatherCentral 2-ish years ago). There are some national models (GFS, European, a handful of others).
There's a big divide even inside the industry as to how much personal analysis is necessary. Some folks repeat what the weather service says practically verbatim, even when it's pretty clear they are not correct. The science involved is no longer top priority in many people's opinion. Then there are some very well-versed mets who can look at the data and come up with an amazingly accurate forecast for a coming storm before it's on (if you'll pardon the pun) anyone else's radar. Some stations want a scientist, some want a pretty face regardless of credentials.
Edit:
The radar thing doesn't have a clear answer. Stations can own their own radar, with varying power, features and ability. The effectiveness of that also depends on having the equipment/computers at the station to display that information in a useful way. On the other hand, stations can simply use the radar data from the NWS, something I believe that's covered in the fees they pay to WSI (or whatever equipment provider they use). But you lose the ability to control the radar and all you have is the data they feed you.