It looks like most of that was done in camera by snap zooming here and there. Personally, I think there are much more effective ways to tell the story. You can kind of simulate the snap zoom in AE with Scale as long as the edge of the frame doesn't show. If you see the edge of the frame then you need to set up a track matte for the footage first then animate scale to simulate the zoom. The other part of the look for your example is hand held camera. You can also simulate that in AE but it's difficult to really get it to look right. I find the most organic way to simulate a hand held camera is to take out my phone, find something that I can photograph that is about the same distance from the camera as the subject in the shot I want to apply hand held effects to, then shoot a little footage with my phone, import that footage, track the subject, then apply the tracking to a solid and use the solid as the parent of the footage I want to apply the hand held effect to. Here also it's important to first set up a track matte for the footage and scale it up a bit so the edges of the frame don't give away the simulated hand held effect.
On a couple of the shots where there is a snap zoom the camera is also moving toward or away from the subject. This presents a situation that is very difficult to simulate in AE because of the perspective change. Perspective is controlled by camera position. Discounting lens distortion, if you change camera position you change perspective, if you change focal length you change the framing of the shot only. When you move the camera toward the subject while zooming the relationship between the background and foreground changes at the same time the framing changes so you get some very odd effects. For example, put a camera on a dolly, frame up your actor, then dolly in and zoom out at the same time to keep the actor the same size in the frame and you'll get a very odd looking "am I going crazy" type effect that they typically use in scenes where the actor is loosing his mind or there is some kind of time warp happening.
Long and short of this discussion, almost everything you see in the camera moves in your sample was done in camera by hand holding and snap zooming. Some of it can be simulated in AE with nothing more than a track matte, scale and position animation.
This effect looks more abstract to me than just simulating "real" physical camera moves and zooms. The dancer's movements are supposed to seem to sort of impact reality.
If the footage is larger than the required final output (as this may have been if it was 4K GoPro footage, for instance) you could get that effect simply using scale and position animations. The trick would be creating the right kind of animation curves (no simple easy eases there) and syncing everything up to the music and motion. I see a lot of manual labor in that music video.