test_MNLDiscreteChoiceModel.py

# This is the script for MNLDiscreteChoiceModel estimation and simulation demo - Guangyu
# run it under semcog_urbansim folder

# In[ ]:
from urbansim.models.dcm import MNLDiscreteChoiceModel

import os
import numpy as np
import pandas as pd

import orca
from urbansim.utils import misc

import dataset
import variables
import models
import utils

orca.run(["build_networks"])
orca.run(["neighborhood_vars"])

# ## HLCM Estimation Function  old one
#%%
# def estimate_hlcm(yaml_config):
#     model = MNLDiscreteChoiceModel.from_yaml(str_or_buffer=misc.config(yaml_config))

#     hh = orca.get_table("households").to_frame(model.columns_used()).fillna(0)
#     b = orca.get_table("buildings").to_frame(model.columns_used()).fillna(0)

#     print(model.fit(hh, b, hh[model.choice_column]))

#     return model, model.fit_parameters

# ## HLCM Configs

# In[8]:
# hlcm_configs = [
#     "hlcm/large_area_income_quartile/hlcm400099.yaml",
#     "hlcm/large_area_income_quartile/hlcm200115.yaml",
#     "hlcm/large_area_income_quartile/hlcm100125.yaml",
# ......
# ]
# m1, m1para = estimate_hlcm(hlcm_configs[0])
# m1.to_yaml("./test.yaml")
# ## Estimate a particular HLCM submodel

# %%

###
# test estimation MNLDiscreteChoiceModel
######## estimation demo ###########
# %%
m2 = MNLDiscreteChoiceModel.from_yaml(str_or_buffer=misc.config("test_indiv.yaml"))
m2.probability_mode

# %%
# compile data for estimation
hh = orca.get_table("households").to_frame(m2.columns_used()).fillna(0)
b = orca.get_table("buildings").to_frame(m2.columns_used()).fillna(0)

# %%
m2.fit(hh, b, hh[m2.choice_column])

#%%
m2.to_yaml("test_indiv.yaml")  # save configs


# %%
######## simulation demo ###########
# select HH and buildings
hhs = hh.loc[hh.large_area_id == 161]
hhs = hhs.sample(n=5)
hhs.building_id = -1

bs = b.loc[b.large_area_id == 161]
bs = bs.sample(n=5)

# %%
# test simulation
chs = m2.predict(hhs, bs, debug=True)

# %%
chs


# %%
m2prob = m2.probabilities(hhs, bs)
m2prob