This tutorial guides you through a sample decision cube analysis project.
An overview (The theory)
The decision cube tool was developed to help people analyze high dimensional data. Imagine that your firm is selling books, and it is your job to find trends and guidelines in the trade database to make a good prediction about:
- what to order next month, or
- what kind of books needs more promotion, because people do not buy them.
Or you just want to analyze the sales data to have a better view about the customers of your firm.
At first you should prepare a raw data table that contains sales data. A raw data table might look like this:
|
Month
|
Day
|
Unit
|
Genre
|
Pages
|
Items sold
|
Profit
|
|
Aug
|
4
|
New York 1
|
Romantic
|
50-100
|
40
|
400
|
|
Oct
|
14
|
Washington
|
Modern
|
30-50
|
3
|
45
|
|
Feb
|
7
|
Los Angeles
|
Art
|
50-100
|
13
|
52
|
|
Dec
|
28
|
New York 2
|
Sci-fi
|
100-200
|
24
|
180
|
|
Aug
|
5
|
Washington
|
Fantasy
|
200-300
|
21
|
140
|
|
Jul
|
19
|
New York 2
|
Classical
|
300-500
|
10
|
140
|
|
Jun
|
22
|
Washington
|
History
|
300-500
|
12
|
144
|
Of course a real data table should contain much more rows. The more data means the more precise analysis. After the raw data is prepared, it can be imported into DBMyne. DBMyne gives you an easy to use import wizard and an intuitive visualization interface for your data.
The following figure illustrates a 3d graph created by DBMyne.
Imagine, that you tell DBMyne to load Month to horizontal axis 1, and Unit to horizontal axis 2, and Items sold to the vertical axis, and DBMyne presents you the following 3d graph. Of course
- you can drag the graph and rotate it by moving the mouse,
- you can interactively change the columns on any axis.
A detailed view (The practice)
Step 1 – Create raw data for data mining
Export your data to be analyzed into a database table. The database type should be one that DBMyne supports, eg: MySQL, DBase, Paradox, FoxPro, Firebird, Interbase. The list of database types will be extended in the future.
Tip: If you are a beginner, I suggest you to export your data to a DBF file called “DBMyne raw data” to your “My Documents” folder (Excel save as can do it).
Step 2 – Create a new project
Start DBMyne. The first screen lets you choose, whether to open an existing project, or to create a new one. Specify the name of the new project (I suggest: DBMyne first steps) and press the create button.
Step 3a – Import raw data into DBMyne, create a connection
The first step of data mining is to import raw data into your DBMyne project. If you do not have data connection to your raw data defined yet, create a new one on the import / connection tab, by pressing the new button.
Specify the name and type of the new connection and press ok. If you have created a DBF file then choose DBase files for connection type, and “My DBase files” for connection name.
Specify the connection parameters on the next dialog. If you have created a DBF file, then specify the folder where you have placed the DBF file. I have suggested to use the “My Documents” folder.
Step 3b – Select the table
Go to the table tab, and select your DBF file. I have suggested to use the “DBMyne raw data” name for your DBF file.
Step 3c – Select the code page
Go to the character set tab, and select the character set your data is encoded in. If you are unsure what to do, don’t change the default setting. DBMyne tries to detect the character set you are using.
Step 3d – Choose columns for horizontal and vertical axis
Go to the dimensions tab and specify which column to use for the horizontal axis, by adding columns to the dimensions list. By adding columns to the facts list, you specify which columns to use for the vertical axis.
Step 4a – Refine dimensions (not required)
Go to the refine step on the top level wizard. In this step you can specify detailed information for the dimensions. Use the new, edit, remove, remove all, and default dimensions buttons to edit the dimensions. The most important task of this step is to give a nice name to the dimensions by using the edit button.
Step 4a – Refine facts (not required)
Choose the facts step on the second level wizard. You will see a list of the facts you have chosen. DBMyne has built in aggregate functions like: count, min, max, average and deviation. By default DBMyne offers to apply each aggregate function to all facts. In this step you should remove the unneeded aggregate-fact pairs, and give a nice name to the remaining items.
Step 5 – Visualize data
Press the visualize button on the top level wizard, and you will see the visualization tool. At first choose the two horizontal dimension from the first two list (choose two different). At second choose a fact to visualize. After you have specified the three axes, DBMyne will let you see the results. Try dragging the 3D graph with the mouse. The graph type buttons let you choose between different visualization methods. The physics buttons let you tell DBMyne how to follow your movements. Experiment with it!
Step 6 – Export data
The export function of DBMyne lets you export your analyzed data directly to Microsoft Excel (Excel must be installed). You can choose multiple dimensions and facts. DBMyne does sub grouping while exporting. By moving dimensions or facts up or down, you can control main or sub groups.
If you want to export only a subset of your data, you can define filters for the export.
