jshbrmn/irt-service
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

sneaky update: add multi-objective functions

Browse Source
This commit is contained in:
Joshua Burman
2023-11-17 19:02:42 -05:00
parent 4001200e9b
commit 82b6cd25ed
4 changed files with 54 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files

56
app/models/objective_function.py
View File

@ -1,13 +1,17 @@
from __future__ import annotations
from typing import TYPE_CHECKING, Any, Literal, Union
from pydantic import BaseModel
from typing import Dict, List, AnyStr
from pulp import lpSum
from pydantic import BaseModel, validator
from typing import Dict, List
from pulp import lpSum, LpMinimize, LpMaximize
from models.targets.tif_target import TifTarget
from models.targets.tcc_target import TccTarget
from models.problem import Problem
if TYPE_CHECKING:
from models.solver_run import SolverRun
class ObjectiveFunction(BaseModel):
# minimizing tif/tcc target value is only option currently
# as we add more we can build this out to be more dynamic
@ -15,17 +19,45 @@ class ObjectiveFunction(BaseModel):
tif_targets: List[TifTarget]
tcc_targets: List[TccTarget]
target_variance_percentage: int = 10
objective: AnyStr = "minimize"
objective: Literal[1,-1] = 1
functions: List[Literal['tcc', 'tif']] = ['tcc', 'tif']
weight: Dict = {'tif': 1, 'tcc': 1}
def for_problem(self, problem_handler: Problem) -> None:
problem_handler.problem += lpSum([
bundle.count * problem_handler.solver_bundles_var[bundle.id]
for bundle in problem_handler.bundles
] + [
problem_handler.solver_items_var[item.id]
for item in problem_handler.items
])
@validator("objective", pre=True)
def set_objective(cls, v) -> List[int]:
if v == 'minimize':
return 1
elif v == 'maximize':
return -1
else:
return None
def for_problem(self, problem_handler: Problem, solver_run: SolverRun) -> List[lpSum]:
functions = []
for function in self.functions:
if function == 'tcc':
functions.append(lpSum([
bundle.trf(solver_run.irt_model, solver_run.theta_cut_score)
* problem_handler.solver_bundles_var[bundle.id]
for bundle in problem_handler.bundles
] + [
item.irf(solver_run.irt_model, solver_run.theta_cut_score) *
problem_handler.solver_items_var[item.id]
for item in problem_handler.items
]))
elif function == 'tif':
problem_handler.problem += lpSum([
bundle.tif(solver_run.irt_model, solver_run.theta_cut_score)
* problem_handler.solver_bundles_var[bundle.id]
for bundle in problem_handler.bundles
] + [
item.iif(solver_run.irt_model, solver_run.theta_cut_score) *
problem_handler.solver_items_var[item.id]
for item in problem_handler.items
])
return functions
def increment_targets_drift(self,
limit: float or bool,