If by class based you mean CBWFQ, then it's primary advantage is treating different traffic differently. Shapers throttle traffic to a rate specified by you with often little control over how the delayed traffic is queued. Is such traffic queued FIFO, or FQ, or WFQ? Is only tail drops implemented or can you do RED, or WRED. Can any traffic be prioritized to jump before all the other traffic? These are many of the features you often can select with CBWFQ that are unlikely to be available with FRTS, or other single function shapers. Oh, and CBWFQ often supports shaping and/or rate limiting too.
E.g., from a 7500:
policy-map shapeAvg1500cir256
class class-default
shape average 1450000 11600 11600
shape adaptive 256000
shape fecn-adapt
queue-limit 256
fair-queue
interface Serial6/0/3
bandwidth xxxx
no ip address
ip access-group 199 in
ip flow egress
encapsulation frame-relay
no ip mroute-cache
load-interval 30
serial restart-delay 0
no fair-queue
frame-relay lmi-type cisco
frame-relay broadcast-queue 128 8000 64
!
interface Serial6/0/3.100 point-to-point
ip address x.x.x.x
ip access-group 111 in
ip flow egress
no arp frame-relay
frame-relay interface-dlci 100
service-policy output shapeAvg1500cir256
!
interface Serial6/0/3.400 point-to-point
ip address x.x.x.x
ip access-group 111 in
ip flow egress
no arp frame-relay
frame-relay interface-dlci 400
service-policy output shapeAvg1500cir256
