Metadata Category Discrepancies

[timg456789]

How is the "category" in the AS metadata being determined? I'm seeing discrepancies where ISP's are showing as hosting. For example ASN 5650 has a category of hosting, but if you compare it against PeeringDB it's classified as an ISP with mostly inbound traffic https://www.peeringdb.com/api/net?asn=5650.

I'm using the AS metadata to supplement lead verification on a public website. I'm seeing a lot of traffic come through classified as hosting which I'm comparing against PeeringDB and through captchas for bot detection and I'm seeing a lot of things that look more like ISP traffic than from what I would expect would come from a hosting datacenter where bot traffic would constantly spam from.

[trick77]

https://lens.ipverse.net/AS5650 is classified as ISP.
Quoting from the as-metadata README:
"Categorization is based on multiple signals and reflects the AS's primary function. It won't always be correct — many networks defy clean categorization (e.g. an ISP that also runs a hosting business, or a university with its own transit infrastructure). Take it as a useful default, not gospel. An overlay mechanism for user-supplied corrections may be added in the future."

[timg456789]

After the recent update to the as-metadata project I'm seeing a lot of AS's that used to be categorized as hosting are now showing as ISP. I think an overlay for user-supplied data would be very useful. If it said the site who made the determination and had some way to verify site ownership it could help create an application level perspective identifying what type of traffic was coming from that AS. I would be willing to help supply information in a structured way from my website where I may end up having to create my own overlay as I build a list of what IP's are sending bot traffic.