Originally posted by: Ghostwriter_muc

Hi everyone,

 

I need a execute block that gives us the different W[A] (time-Windows) as Output.

Probably this is quite easy, but I have no clue how to implement this as I'am about to getting started with Cplex.

 

.Mod

 

.Data

 

There is a number of Patients, each Patient has a defined set of activities. I defined the subsets for each activity, precedessor, successor and Minimum time lag (clinicalPathwaysPerPatient). 

Each Patient has his defined date of Arrival at the Hospital (alpha in the Data.file).  Each activity has an earliest and latest starting time (E and L in the Data.file). 

The earliest starting time additionally depends on the Arrival date, the sequence of the predeccesors (Activity 2 can not start before 1 is finished) and the Minimum time lag given as int dmin. 

 

Example:

Patient 1:

 

Activity 1: E=1, Arrival of Patient 1 is 1, there is no int dist given, therefore 0 (This is a problem for me, because i don't know how to implement this)--> earliest Starting time 1

Activity 2: E=1, Arrival of Patient 1 is 1, int dmin =0 --> earliest Starting time 1

Activity 3: E=1, Arrival of Patient 1 is 1, int dmin =3 --> earliest Starting time 4

 

calculating latest starting time backwards:

Activity 3: L=6, int dmin= 0 --> Latest starting time = 6

Activity 2: L=6, int dmin= 3 --> Latest starting time = 3

Activity 1: L=6, there is no int dist given, therefore 0 (This is a problem for me, because i don't know how to implement this) --> Latest starting time = 3

 

At the end i want to have time windows as followed: So I can sum t from 1 to 3 or from 4 to 6 and not from 1 to 6.

Activity 1: W1=[1,2,3]

Activity 2: W2=[1,2,3]

Activity 3: W3=[4,5,6]

 

I would be grateful for you're help and any tips! Do not hesitate to ask if you need further information.

 

Many thanks in advance!!

 

Kind regards

#DecisionOptimization
#OPLusingCPLEXOptimizer

Hi,

instead of attaching screenshots, can you attach .mod and .dat files so that other users may try ?

regards

#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: Ghostwriter_muc

I attached the .mod and .data files.

#DecisionOptimization
#OPLusingCPLEXOptimizer

Hi,

you could start with

int earliest[p in P][a in A];
 
 {int} activitiesWithoutOffset={pre | <pre,succ,d> in clinicalPathways};
 
 
 execute
 {
 
 
 for(var p in P) for(var a in A)
 {
     var temp=1000;
     if ( activitiesWithoutOffset.contains(a)) temp=E[a];
     for(var pathway in clinicalPathwaysPerPatient[p] )
     {
       if (pathway.succ==a) temp=earliest[p][pathway.pre]+pathway.dmin;
     }
     earliest[p][a]=temp;
}
writeln(earliest);
}

int startWindow[a in A]=min(p in P) earliest[p][a];

execute
{
writeln(startWindow);
}

regards

#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: Ghostwriter_muc

WOW! Thank you very very much. This helps a lot.

 

Is it possible to separate the startWindow into [1 1 4] for patient 1 and [1 1 4] for patient 2?

 

 

#DecisionOptimization

Hi,

if you write

{clinicalPathway} clinicalPathways = ...;

{clinicalPathway} clinicalPathwaysPerPatient[p in P] = {c | c in clinicalPathways : (firstOne[p] <= c.pre <= PHI[p]) || (firstOne[p] <= c.succ <= PHI[p])};

 // subset of activities
 execute {
     writeln(clinicalPathwaysPerPatient);
 }
 
 int earliest[p in P][a in A];
 
 {int} activitiesWithoutOffset[p in P]={pre | <pre,succ,d> in clinicalPathwaysPerPatient[p]};
 
 int noActivity=1000;
 
 execute
 {
 
 
 for(var p in P) for(var a in A)
 {
     var temp=noActivity;
     if ( activitiesWithoutOffset[p].contains(a)) temp=E[a];
     for(var pathway in clinicalPathwaysPerPatient[p] )
     {
       if (pathway.succ==a) temp=earliest[p][pathway.pre]+pathway.dmin;
     }
     earliest[p][a]=temp;
}
writeln(earliest);
}

int startWindow[a in A]=min(p in P) earliest[p][a];

execute
{
writeln(startWindow);
}
 

then you see

 

[[1 1 4 1000 1000 1000]
         [1000 1000 1000 1 1 4]]
 [1 1 4 1 1 4]

Do you have a specific hospital in mind or do you target a generic use case ?

regards

 

#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: Ghostwriter_muc

The implementation is very good and I'am deeply grateful for your time and hard work.

 

I forgot to tell you that alpha[P] is a very important condition. Alpha is the admission date for a patient. The result for patient 1 is right, but for patient 2 it is not. For patient 2 the admission period is period 2.  The earliest starting time can not begin before the admission date. The right result would be [ 1 1 4 2 2 5 ]. In which way can I implement this condition?

 

Regarding the format of the result, the time windows should look like the following example. A combination of the earliest (e.g = 1) and latest (e.g = 3) starting time, including the period 2. Because i need all 3 periods for my constraint.

 

. Because i have to sum 

 

Example

For patient 1: 

W[1] = [ 1 2 3]

W[2] = [ 1 2 3]

W[3] = [ 4 5 6]

 

For patient 2:

W[4] = [ 2 3]

W[5] = [ 2 3]

W[6] = [ 5 6]

 

It should like this W= [ 1 2 3 ], [ 1 2 3 ], [ 4 5 6 ],[  2 3  ], [ 2 3 ], [ 5 6 ],

 

I tried to solve this problem every day but i can't. It is really difficult for a beginner. Can you please help??!

Kind regards

#DecisionOptimization

Originally posted by: Ghostwriter_muc

I've tried to calculate the latest time windows, but i don't know how to calculate it. I've to start with the last activity. 

 

{clinicalPathways} clinicalPathwaysPerPatient = {< 1, 2, 0 > < 2, 3, 3 >} {< 4, 5, 0 > < 5, 6, 3 >}

L(i in A) - pathway.dmin --> ( 6 - 3) = 3 and then (3 - 0)

Is it possible to start from the back to the front?

 

The solution should be [ 3 3 6 3 3 6 ].

 

I solved the problem with the alpha. The solution is now [ 1 1 4 2 2 5 ]. I replaced E[i] with alpha[p]

 

Is it possible to combine the results from startWindow and FinsihWindow? 

--> W[ i in A ] = [ 1 2 3 ], [ 1 2 3 ], [ 4 5 6 ],[  2 3  ], [ 2 3 ], [ 5 6 ] 

 

kind regards

#DecisionOptimization
#OPLusingCPLEXOptimizer

Hi,

let me give you a small example:

setof(int) A = asSet(1..6);
 
 int startWindow[A]=[1,1,4,1,1,4];
int endWindow[A]=[ 1, 1, 4, 2, 2,5 ];

{int} W[a in A]=asSet(startWindow[a]..endWindow[a]);
execute
{
writeln(W);
}

which gives

 [{1} {1} {4} {1 2} {1 2} {4 5}]

regards

 

#DecisionOptimization
#OPLusingCPLEXOptimizer