BorderThink.py

import numpy
import pandas
import argparse
import sys

from mesa.batchrunner import FixedBatchRunner
from BorderModel import BorderModel

# Define possibilities
parser = argparse.ArgumentParser(description='BorderThink automates the different parameters for the BorderModel simulation')
parser.add_argument('theory', type=str, help="Which theory do you want to test?\
					contact - target - ethnocentrism - scaled_ethnocentrism - media")
parser.add_argument('stage', type=int, help="Which stage do you want to simulate? (there are multiple follow-up models) 1, 2 ..")
parser.add_argument('iterations', type=int, help='How many times should each variable parameter be tested?')
parser.add_argument('max_steps', type=int, help='What is the step ceiling for this model?')

args = parser.parse_args()

fixed_params = {
	"width": 100,
	"height": 240,
	"return_chance": 0.05,
	"home_chance": 0.005,
	"decay_limit": 140,
	"sound_mean_interval": 0.1,
	"border_heights": [ 124, 104 ],
	"init_big_inventory": True
}

if args.theory == "contact":
	fixed_params = { **fixed_params,
					 "domestic_travel_chance_nl": 0.005,
					 "domestic_travel_chance_be": 0.005,
					 "ethnocentrism_nl": 0,
					 "ethnocentrism_be": 0,
					 "scaled_ethnocentrism": False,
					 "media_receptiveness": False }

	if args.stage == 1: # range: 0 -> 0.0100, step = 0.0010
		parameters_list = [ { "abroad_travel_chance_nl": probability,
						   "abroad_travel_chance_be": probability } \
						   for probability in numpy.arange(0.0000, 0.0101, 0.0001) ]
	elif args.stage == 2: # range: 1/10 -> 1/10^7
		parameters_list = [ { "abroad_travel_chance_nl": 1 / pow(10, power),
							   "abroad_travel_chance_be": 1 / pow(10, power) } \
							   for power in [0] + numpy.arange(1, 8, 1) ]
elif args.theory == "target":
	fixed_params = { **fixed_params,
					 "abroad_travel_chance_be": 0.001,
					 "abroad_travel_chance_nl": 0.001,
					 "domestic_travel_chance_be": 0.005,
					 "ethnocentrism_nl": 0,
					 "ethnocentrism_be": 0,
					 "scaled_ethnocentrism": False,
					 "media_receptiveness": False }


	if args.stage == 1: # range: 0 -> 0.100, step = 0.001
		parameters_list = [ { "domestic_travel_chance_nl": probability } \
							   for probability in numpy.arange(0.000, 0.101, 0.001) ]
	elif args.stage == 2: # range: 0.100 -> 1.00, step = 0.050
		parameters_list = [ { "domestic_travel_chance_nl": probability } \
							   for probability in numpy.arange(0.100, 1.050, 0.050) ]
elif args.theory == "target2":
	fixed_params = { **fixed_params,
					 "abroad_travel_chance_be": 0.001,
					 "abroad_travel_chance_nl": 0.001,
					 "domestic_travel_chance_be": 0.005,
					 "domestic_travel_chance_nl": 0.005,
					 "ethnocentrism_nl": 0,
					 "ethnocentrism_be": 0,
					 "scaled_ethnocentrism": False,
					 "media_receptiveness": False }

	parameters_list = [ { "target_accel_count": acceleration_count } \
							   for acceleration_count in range(1, 21, 1) ]

	print("Stage is unsupported for this theory and will be ignored")
elif args.theory == "ethnocentrism":
	fixed_params = { **fixed_params,
					 "domestic_travel_chance_be": 0.005,
					 "domestic_travel_chance_nl": 0.005,
					 "scaled_ethnocentrism": False,
					 "media_receptiveness": False }


	if args.stage == 1:
		parameters_list = [ { "abroad_travel_chance_be": 0.001,
						 	  "abroad_travel_chance_nl": 0.001,
							  "ethnocentrism_nl": 0.85,
							  "ethnocentrism_be": probability } \
							   for probability in numpy.arange(0, 1.01, 0.01) ]
	elif args.stage == 2:
		parameters_list = [ { "abroad_travel_chance_be": 0.0001,
						 	  "abroad_travel_chance_nl": 0.0001,
							  "ethnocentrism_nl": 0.85,
							  "ethnocentrism_be": probability } \
							   for probability in numpy.arange(0, 1.01, 0.01) ]
	elif args.stage == 3:
		parameters_list = [ { "abroad_travel_chance_be": 0.0001,
						 	  "abroad_travel_chance_nl": 0.0001,
							  "ethnocentrism_nl": 0.85,
							  "ethnocentrism_be": probability } \
							   for probability in numpy.arange(0.90, 1.01, 0.01) ]
elif args.theory == "scaled_ethnocentrism":
	fixed_params = { **fixed_params,
					 "abroad_travel_chance_be": 0.001,
					 "abroad_travel_chance_nl": 0.001,
					 "domestic_travel_chance_be": 0.005,
					 "domestic_travel_chance_nl": 0.005,
					 "media_receptiveness": False,
					 "ethnocentrism_nl": 0,
					 "ethnocentrism_be": 0, }
	parameters_list = [ { "scaled_ethnocentrism": True } ]

	print("Stage is unsupported for this theory and will be ignored")
elif args.theory == "media":
	fixed_params = { **fixed_params,
					 "domestic_travel_chance_be": 0.005,
					 "domestic_travel_chance_nl": 0.005,
					 "scaled_ethnocentrism": False,
					 "ethnocentrism_nl": 0,
					 "ethnocentrism_be": 0, }

	if args.stage == 1: # guaranteed convergence
		parameters_list = [ { "abroad_travel_chance_be": 0.001,
					 		  "abroad_travel_chance_nl": 0.001,
							  "media_receptiveness": probability } \
							   for probability in numpy.arange(0, 1.05, 0.05) ]
	elif args.stage == 2: # guaranteed divergence
		parameters_list = [ { "abroad_travel_chance_be": 1 / pow(10, 6),
					 		  "abroad_travel_chance_nl": 1 / pow(10, 6),
							  "media_receptiveness": probability } \
							   for probability in numpy.arange(0, 1.05, 0.05) ]

# This theory is meant to test whether language change is accelerated when we increase
# the domestic travel chance for the Netherlands
elif args.theory == "nl_artefact_test":
	fixed_params = { **fixed_params,
					 "abroad_travel_chance_be": 0,
					 "abroad_travel_chance_nl": 0,
					 "domestic_travel_chance_be": 0,
					 "scaled_ethnocentrism": False,
					 "ethnocentrism_nl": 0,
					 "ethnocentrism_be": 0,
					 "media_receptiveness": False }

	parameters_list = [ { "domestic_travel_chance_nl": probability } \
						   for probability in numpy.arange(0.000, 0.051, 0.001) ]

# This theory is meant to test whether language change is accelerated when we increase
# the target acceleration count
elif args.theory == "nl_artefact_test2":
	fixed_params = { **fixed_params,
					 "abroad_travel_chance_be": 0,
					 "abroad_travel_chance_nl": 0,
					 "domestic_travel_chance_be": 0,
					 "domestic_travel_chance_nl": 0.005,
					 "scaled_ethnocentrism": False,
					 "ethnocentrism_nl": 0,
					 "ethnocentrism_be": 0,
					 "media_receptiveness": False }

	parameters_list = [ { "target_accel_count": acceleration_count } \
							   for acceleration_count in range(1, 21, 1) ]

	print("Stage is unsupported for this theory and will be ignored")
else:
	print("Argument not recognised")
	sys.exit(0)

print("Launching simulations for the '{}' theory".format(args.theory))
print("Fixed parameters: {}".format(len(fixed_params)))
print("Variable parameters: {}".format(len(parameters_list[0])))
print("Iterations for each parameter combination: {}".format(args.iterations))
print("Max steps: {}".format(args.max_steps))

print("Launching simulations NOW")

batch_run = FixedBatchRunner(
    BorderModel,
    parameters_list,
    fixed_params,
    iterations=args.iterations,
    max_steps=args.max_steps,
    model_reporters={"data": lambda model: model.datacollector }
)

batch_run.run_all()

print("Simulations finished. Generating report...")

pandas_runs = []

run_data = batch_run.get_model_vars_dataframe()
for run in run_data.iloc:
	panda = run.data.get_model_vars_dataframe()
	panda.index.name = "step"

	for column in run_data.columns.values:
		if column in ["data", "border_heights"]:
			continue

		panda[column.lower()] = run[column]

	pandas_runs.append(panda)

mother_panda = pandas.concat(pandas_runs)
mother_panda.to_csv("{}_stage{}.csv".format(args.theory, args.stage), sep=";")

print("Succesfully written report. Exiting...")