fboehlandt
Active Member
- Joined
- Sep 9, 2008
- Messages
- 334
Hi everyone,
I am trying to solve a multiple objective optimization problem using the Minimax variable and Solver Engine. I have two objectives and the two separate optimizations to find the target values are linear in nature. However, the combined optimization using % deviations from the target values violates the linearity requirements. I'm not sure why that is and whether it can be avoided. I can solve via non-linear optimization but it is rather slow. Some additional information:
- all variables are binary
- there are two sets of variables linked by a constraint
- the objectives are as follows: maximize the cumulative loading scores whilst minimizing the number of groups used. Variables can be sorted into groups according to their highest or second highest loading (binary). Groups with at least one member are active (binary)
- Some variables are always categorized according to their maximum loading. Thus, the binary decision variables do not extend over the full lengths of the variable range
Since this is difficult to visualize I have uploaded the spreadsheet onto my website: http://hedge-fund-analysis.net/pages/excel-models.php (testfile.xlsm). Note that yellow shaded cells are variable cells. I'm looking for a way to retain linearity. Any help is greatly appreciated.
I am trying to solve a multiple objective optimization problem using the Minimax variable and Solver Engine. I have two objectives and the two separate optimizations to find the target values are linear in nature. However, the combined optimization using % deviations from the target values violates the linearity requirements. I'm not sure why that is and whether it can be avoided. I can solve via non-linear optimization but it is rather slow. Some additional information:
- all variables are binary
- there are two sets of variables linked by a constraint
- the objectives are as follows: maximize the cumulative loading scores whilst minimizing the number of groups used. Variables can be sorted into groups according to their highest or second highest loading (binary). Groups with at least one member are active (binary)
- Some variables are always categorized according to their maximum loading. Thus, the binary decision variables do not extend over the full lengths of the variable range
Since this is difficult to visualize I have uploaded the spreadsheet onto my website: http://hedge-fund-analysis.net/pages/excel-models.php (testfile.xlsm). Note that yellow shaded cells are variable cells. I'm looking for a way to retain linearity. Any help is greatly appreciated.
