QMF Vision comes with geo data at the continent, country, state/province, and city levels. It does not contain street address level data. However, with properly geocoded address data QMF Vision and QMF for Workstation can display it. Here is some data for street address in Boston, MA USA.
It has been geocoded by Google via the calculated column called LOCATION defined as this:
Some applications may want the Latitude and Longitude as an ordered pair as in the Location column. Others may want two separate columns. Here is some SQL that will split the LOCATION column into LAT and LNG
SELECT SHOP_NAME, ADDRESS, CITY, STATE, --TWSHAWN.GEOBOSTON_T
ZIP, PLACE, LOCATION,
SUBSTR(LOCATION,1, POSSTR(LOCATION,',')-1) as LAT,
SUBSTR(LOCATION,POSSTR(LOCATION,',')+1) as LNG
Another way to do this is at the calculated column level.
Here is the formula for a LAT column:
They may also be split in QMF Vision using the SPLIT function, however if you do it in the SQL or Calculated column, it will be usable by other applications.
Part 2) Deploying the data in QMF Vision
I have connected to the QMF Catalog on Db2 for z/OS and pulled the data from the query
Notice that the LAT and LNG have a datatype of Double precision numbers
We need to change these data types.
Click where it says Double.
Scroll down to the choice of Latitude
Click on LNG
Click on Double again and scroll down to Longitude
Notice that the column headings no longer say 1,2,3
Now click the Create Dashboard button
Remove whatever column is in the X slot and put in LAT and LNG.
They must be in the order LAT, LNG with NO OTHER columns preceding them or succeeding them.
In the Toolbox click the icon for Charts and select the Map chart
At a high level, the map will consolidate addresses that are close together into a summary total. Here it is showing 58 for the Boston total.
Upon Zooming in, the summary number will be broken up into smaller clumps
Eventually the clumps will be single addresses
Note that if two addresses are very close then the current magnification level will NOT split the location indicator into individual markers. You may end up with some spots still marked as 2 or 3.
In QMF Vision, there is one field that may be added to the pin point and that shows up when the mouse is hovered over the pin. This field must be Numeric in nature.
In QMF for Workstation applications you may have a s many fields in the information popup or tooltip text as you wish, as well as click events. See the following video from the
IBMQMF - YouTube channel
Machine Learning for z/OS and QMF.
A not-well know feature of QMF is the ability for IzODA (IBM Open Data Analytics for z/OS) to connect to a QMF Catalog, have QMF run a query and return the data to the Machine Learning for z/OS component for analysis. In general, running an SQL query may be better than simply pulling an entire table from Db2 or an entire VSAM file. The QMF SQL query can convert NULL values to group averages or normalize columns or eliminate duplicate rows.
In the case of geocoding consider that without the granularity of Latitude and Longitude, the customers are lumped together at the City level or the Zip code level.
As an example, a K-Means Cluster or K-Nearest Neighbor analysis will likely yield much better insights if the data is groupable by Latitude and Longitude coordinates.