I am making my first foray into ML type work using GCPs AutoML offering. I work for a water systems technology company and would like to use AutoML to identify potential customers. We have a dataset of events which includes dates between 1950-2020, coordinates, a few pieces of numeric data regarding the event, and a name to identify two events having taken place in the same 'location' with different coordinates or dates.
This information should be enough to infer whether there is a good advertising opportunity for our products/services within a geographic area. The current understanding of the data set is that a cluster of events in a similar timeframe & geography should indicate a potential need for our products, I think there could be additional data introduced down the line, eg. average income, previous clients/web visits, etc.
The dataset does NOT have any field that quantifies whether the event is good or bad indicator for customers. This is where I do not know how to continue.
We are prepared to do some test advertising within geographic areas to determine if our manual analysis is accurate. Then use the results of those tests to set a value on the 'training data' we used in our advertising.
I suppose this is more a workflow/data science question than an GCP AutoML question specifically.
EXAMPLE DATA: (A big 'D' and small 'F'/'L', along with the geographic & time clusters, indicates potential customers in that geographic area.) The dataset has 50,000+ entries currently.
NAME COORDINATE DATE D F L No.
FFAY xx.aa yy.aa 1950/02/15 50 5 15 1
FFAY xx.ab yy.ab 1950/02/15 75 2 19 2
FFAY xx.ac yy.ba 1950/02/16 60 9 22 3
FFAY xx.ad yy.cb 1950/02/17 50 5 25 4
IATT yx.aa xy.aa 1958/11/04 90 2 17 1
IATT yx.aa xy.aa 1958/11/08 75 3 13 2
IATT yx.aa xy.aa 1975/06/22 55 5 32 1
IATT yx.aa xy.aa 1975/06/22 67 12 45 2
.....
.....
.....
FFAY xx.af yy.af 1999/05/27 20 3 7 1
FFAY xx.ag yy.ag 1999/05/30 22 7 9 2
