87 lines
3.3 KiB
Python
87 lines
3.3 KiB
Python
import numpy as np
|
|
import pandas as pd
|
|
from env import Env
|
|
import datetime
|
|
|
|
"""
|
|
The experiment.py is used to read the input value in the file xv.csv according to the Orthogonal Array (OA) table
|
|
in the file oa25.txt.
|
|
The file oa25 means that the number of total experiments is 25, as it has at most 6 inputs, each one has 5 levels.
|
|
Therefore, oa25.txt has 6 columns, and the values are labelled as 0, 1, 2, 3 and 4.
|
|
After reading these two files, the code below runs the model one-by-one (which is very ineffective), and
|
|
uses lists to record the outputs, and save the results in the sub-folder result.
|
|
"""
|
|
|
|
# This idx_start is used when the bulk running is unexpectedly stopped, and we need to rerun the model.
|
|
idx_start = 1 # first index is 1, not 0 ! If we read oa25.txt, then the max value is 25
|
|
|
|
# number of runs for each experiment
|
|
n_sample = 50
|
|
|
|
df_xv = pd.read_csv("xv.csv", header=None, index_col=None).transpose()
|
|
# the names of four input indicators
|
|
lst_xv_keys = [
|
|
'alpha',
|
|
'percent_search',
|
|
'is_RH_ratio',
|
|
'is_FH_ratio'
|
|
]
|
|
df_xv.columns = lst_xv_keys
|
|
|
|
# read the OA table
|
|
df_oa = pd.read_fwf("oa25.txt", header=None, widths=[1]*6)
|
|
n_row, n_col = df_oa.shape
|
|
# these para below keep unchanged
|
|
model_para = {
|
|
"n_worker": 1000,
|
|
"n_firm": 100
|
|
}
|
|
|
|
# defined six outputs
|
|
lst_op_key = ['out_w_avg_salary', 'out_w_gini_salary', 'out_f_avg_profit', 'out_f_avg_yield',
|
|
'out_f_gini_profit', 'out_w_percent_hired']
|
|
lst_2d_op_avg = []
|
|
lst_2d_xv = []
|
|
for idx_row in range(idx_start-1, n_row, 1):
|
|
print(f"Running the {idx_row + 1}-th experiment at {datetime.datetime.now()}.")
|
|
lst_value = []
|
|
for idx_col, k in enumerate(lst_xv_keys):
|
|
# read the corresponding value by mapping OA and xv.csv
|
|
oa_idx_level = int(df_oa.iat[idx_row, idx_col])
|
|
lst_value.append(df_xv.iloc[oa_idx_level][k])
|
|
lst_2d_xv.append(lst_value)
|
|
# merge the two dictionaries into one; send it to the AgentPy model
|
|
dct_merge = {**model_para, **dict(zip(lst_xv_keys, lst_value))}
|
|
# pprint.pprint(dct_merge)
|
|
lst_2d_op = []
|
|
|
|
for i in range(n_sample):
|
|
print(f"-- {i + 1}-th sample at {datetime.datetime.now()}.")
|
|
the_model = Env(dct_merge)
|
|
|
|
while 1:
|
|
the_model.step()
|
|
if not the_model.running:
|
|
break
|
|
|
|
# record the six outputs
|
|
lst_2d_op.append([the_model.out_w_avg_salary, the_model.out_w_gini_salary,
|
|
the_model.out_f_avg_profit, the_model.out_f_avg_yield,
|
|
the_model.out_f_gini_profit, the_model.out_w_percent_hired])
|
|
|
|
arr_op = np.array(lst_2d_op)
|
|
lst_2d_op_avg.append(arr_op.mean(axis=0).tolist())
|
|
|
|
# these codes below should be outside of loop. but temply inside
|
|
arr_op_avg = np.array(lst_2d_op_avg)
|
|
|
|
df_final = pd.concat([pd.DataFrame(lst_2d_xv, columns=lst_xv_keys).reset_index(drop=True),
|
|
df_oa.iloc[:, len(lst_xv_keys):].reset_index(drop=True)], axis=1)
|
|
|
|
df_final = pd.concat([df_final.reset_index(drop=True),
|
|
pd.DataFrame(lst_2d_op_avg, columns=lst_op_key).reset_index(drop=True)], axis=1)
|
|
|
|
# save to excel files
|
|
df_final.to_excel(f'result/experiment_result_{idx_row + 1}.xlsx')
|
|
df_final.to_csv(f'result/experiment_result_{idx_row + 1}.csv')
|