Pack: Sudoku solving with RapidMiner (Who Wants to be a Data Miner 2012?)

Description:

Process from the http://rcomm2012.org live data mining challenge. The task was to partially solve a Sudoku puzzle by solving three subtasks. Processes 1-3 in this pack are the solutions to the actual tasks, process 0 loads the required data into your repository and process 4 is the bonus that solves the entire puzzle.

This process loads the two required data sets from rapid-i.com and stores them in your repository. The first data set contains the numbers 1 to 9 and the second contains the predefined values in a Sudoku. See the comment of process 1 for more details on the format.

Description:

Process from the http://rcomm2012.org live data mining challenge. The task was to partially solve a Sudoku puzzle by solving three subtasks. Processes 1-3 in this pack are the actual solutions to the tasks, process 0 loads the required data into your repository and process 4 is the bonus that solves the entire puzzle.

This process loads the numbers 1..9 and uses two Cartesian Product operators to compute the set of all combinations x, y, v. Finally we compute the index of the 3x3 sub-field as z=ceil(x/3) + ceil(y/3) * 3 - 3. We end up with a table representing all values we could possibly insert into a Sudoku puzzle, i.e. any combination of column x, row y, and value v. The sub-field index z is implicitly defined by x and y but will help us later.

The result is saved in the table "All Possibilities" and will be used by process 2.

Note that the data set "All Possibilities" has the same format as the input data set "Sudoku Predefined" which contains the numbers that are predefined in the Sudoku puzzle.

Description:

Process from the http://rcomm2012.org live data mining challenge. The task was to partially solve a Sudoku puzzle by solving three subtasks. Processes 1-3 in this pack are the actual solutions to the tasks, process 0 loads the required data into your repository and process 4 is the bonus that solves the entire puzzle.

IMPORTANT: Save this process as "RCOMM 2012 Sudoku Challenge: 2 - Drop impossible values" since it is included from process 3.

This process eliminates impossible rows from the table "All Possibilities" that can be excluded by looking at one row of the input data set "Sudoku Predefined". The table "All Possibilities" is defined by process 1, see the comments therein for details. The file "Sudoku Predefined" contains one predefined number in the Sudoku puzzle for row. The row x, y, z, v means that the value v occurs at column x, row y. z is the implicitly defined index of the 3x3 sub-field containing the cell x, y. The row considered by this process is defined by the macro %{example} which is set to "1" in the process context. The process starts by extracting the values of x, y, z, and v for the selected example into macros. Then, it computes a new column that determines whether a row in "All Possibilities" is still possible. That is the case if it matches on v and either x, y, or z. These examples are then removed. This process is repeatedly called by process 3 which assigns different values to the macro %{examples}.

Description:

Process from the http://rcomm2012.org live data mining challenge. The task was to partially solve a Sudoku puzzle by solving three subtasks. Processes 1-3 in this pack are the actual solutions to the tasks, process 0 loads the required data into your repository and process 4 is the bonus that solves the entire puzzle.

IMPORTANT: Save this process as "RCOMM 2012 Sudoku Challenge: 3 - Drop impossible values (loop)" since it is included from process 4. This process uses process 2 repeatedly to eliminate values that can be excluded by looking at all entries in "Sudoku Predefined". Therefore it iterates over "Sudoku Predefined" each time selecting one to be processed by process 2. A combination of Remember and Recall is used to feed the data set containing the remaining possibilities - starting with "All Possibilities" into the subprocess. Once only possible combinations remain, we are interested in those cells where there is only one possibiliy left. This is achieved by an Aggregate operator. Finally we append the newly generated knowledge (those cell values that are certain) to those which were predefined.

Description:

Process from the http://rcomm2012.org live data mining challenge. The task was to partially solve a Sudoku puzzle by solving three subtasks. Processes 1-3 in this pack are the actual solutions to the tasks, process 0 loads the required data into your repository and process 4 is the bonus that solves the entire puzzle.

Finally, this process solves the entire Sudoku. It uses process 3 to add inferred cell values to the set of predefined values. It does so repeatedly until 81 values are defined, i.e. all cells are solved. Finally, it uses a Pivot operator to transform the vertical representation into a normal 2D-representation of a Sudoku.