Database Reference
In-Depth Information
To work through the tools in the following sections, open the
DMAddins_SampleData.xlsx
sample workbook. Find the Table Analysis Tools Sample worksheet and select the Excel table. You're
ready to use the Table Analysis Tools. This table contains demographic data for individuals and a
column indicating whether those individuals purchased a bike (see Figure 14-3).
Figure 14-3:
This dataset contains demographic data for a bike company.
Analyze Key Influencers
Generally speaking, a dataset is defined by its “structure” (cells, rows, columns) as well as its relation-
ships with other datasets. But there is another way to think of a defined dataset, and that is the pat-
terns that can be discovered in the data itself, irrespective of the data's structure. A simple example of
a data pattern can be shown in the following two number sequences:
Sequence 1: {0, 1, 2, 3, 4}
Sequence 2: {2687, 2696, 2589, 694, 466}
In these two sequences, a pattern can easily be discerned in which the values of Sequence 2 seem to
be influenced by the values in Sequence 1 (the higher the value from Sequence 1, the lower the value
in Sequence 2). Our brains are relatively good at detecting simple patterns, but more complex
patterns and/or larger datasets require CPU power. The Analyze Key Influencers Table Analysis tool
does exactly this.
When you use the Analyze Key Influencers tool, a lot of complex work is happening behind the scenes:
1.
A temporary mining structure is created on the Analysis Services instance to hold the data in
the table that is being analyzed.
2.
Using the Microsoft Naïve Bayes algorithm, a mining model is generated that allows each
column to be scored in terms of predicting the target column.
3.
The results of this scoring process are inserted into a new report, which is added to a new
worksheet automatically. Optionally, other reports can be added that further analyze any
combination of target column with one input column.
Search WWH ::
Custom Search