Only way I have in mind is maintenance window. During scaling down, you can send request to DT to fire up maintenance window that will cover those hosts that will be disconnected. Than do scaling down, after few minutes disable maintenance window. This is really dirty workaround but I don't have other idea for now.
@Tarun A. Without being fully aware of your process for autoscaling, I would propose a few ideas. If these devices are defined and chosen prior to the autoscaling action, setup an automatic tag to tag them and have a maintenance window that is set off of that defined tag. That would suppress these alarms.
Within the Alert profile page, select the alert profile and exclude the hosts slated for autoscaling to be excluded from alerts.
Globally, change your settings to not alert on graceful shutdowns and have your hosts shutdown gracefully. This will then not alert you when the host is shutdown. You can also go into the settings page and disable the one agent on the host prior to the auto scaling.
There is now the option to post a 'marked for termination' event via the Events API.