Pitney Bowes MapInfo Vertical Mapper - Manual (Page 172)

Predictive Analysis
170
Vertical Mapper 3.7
In the Column section, you can choose either of the following options:
Existing Column enables you to choose an existing column in the target table that will be
updated with data from the source table.
New Column enables you to specify the name of a new column that will be added to the target
table and will contain the data from the source table.
The Data Table list displays all the valid tables open in MapInfo Professional. The highlighted table
becomes the source data table.
The Column list displays the columns available in the selected table, and enables you to choose the
column containing the data.
The Cell Size box enables you to specify the size of the 32 bit grid that will temporarily hold the
source tables data before applying it to the target table.
The Advise button Enables you to run an analysis to indicate the number of problem polygons that
are found in the target table. These are polygons that are too small to have a single grid cell located
inside.
The Create a Problem Table option enables you to choose to generate a region table containing a
copy of all the problem polygons found in the target table.
Predictive Analysis
Predictive analysis is a method of “smart” classification. It examines the statistical characteristics of
multiple input grids found within a given “training” area and then locates other areas with similar
properties. For example, oil exploration is one application. Although bore holes give clear results,
they are expensive. Suppose experience reveals that the presence of oil in the soil is usually
strongly related to the presence of certain minerals or other elements. If we can establish the
correlation between the presence of these other elements and oil, then it becomes possible to
measure the concentration of the other elements to predict the presence of oil, which is a much less
expensive undertaking than drilling new bore holes.
Predictive analysis enables you to “teach” the program how to classify areas based on input criteria.
For example, if you identify several regions where oil is present, using these as models, you can
teach the program to identify other regions with similar characteristics.
Before running predictive analysis, you must first set up a “teaching table.” This is a MapInfo table
containing the region objects you are using as sample regions. It must have at least one integer
column, which holds the value of the “class” of region you are defining. In any teaching table, you
must define at least two different classes. For example, if you want to identify areas with probable
high crime rates based on demographic information, you must define both an area where crime is
known to be high, and an area where crime is known to be low. For more accurate results, you can
use more than one region for each class type in the teaching table.
1. Draw MapInfo region objects defining the areas you want to use as samples. You must define at
least two regions in different locations within the grid.
2. Save the region objects in a MapInfo table. Ensure the table has at least one column of type
integer. Use this column to define the “class” of the region. For example, you might assign the