from urbansim.models import SegmentedMNLDiscreteChoiceModel
seg_col = 'income_3_tenure'
sample_size = 100
prob_mode = 'single_chooser'
choice_mode = 'aggregate'
choice_col = 'zone_id'
default_model_expr = 'ln_avg_unit_price_zone + median_age_of_head + median_yearbuilt_post_1990 ' \
'+ percent_renter_hh_in_zone + townhome + jobs_within_45min'
remove_alts = False
#alts_filter = ['residential_units > 0']
hlcm = SegmentedMNLDiscreteChoiceModel(segmentation_col=seg_col, sample_size=sample_size,
probability_mode=prob_mode, choice_mode=choice_mode,
choice_column=choice_col, default_model_expr=default_model_expr)
alts = orca.get_table('zones').to_frame()
alts = alts.fillna(0)
### define models
# model1 = 'ln_dist_rail + ln_avg_unit_price_zone + median_age_of_head + median_yearbuilt_post_1990' \
# ' + percent_hh_with_child_x_hh_with_child + percent_renter_hh_in_zone' \
# ' + townhome + multifamily + jobs_within_45min + parks_3mi + golf_courses_3mi + schools_3mi + ' \
# 'fast_food_3mi + restauraunt_3mi + supermarket_3mi + cafes_3mi'
#
# model2 = 'ln_dist_rail + income5xlt_x_avg_unit_price_zone + median_age_of_head + median_yearbuilt_post_1990' \
# ' + median_yearbuilt_pre_1950 + percent_hh_with_child_x_hh_with_child + percent_renter_hh_in_zone' \
choice_mode = "aggregate"
choice_col = "zone_id"
default_model = (
"ln_avg_nonres_unit_price_zone+non_residential_sqft_zone+ln_residential_unit_density_zone"
"+ln_avg_unit_price_zone+median_age_of_head+mean_income+jobs_within_45min+ln_jobs_within_30min"
)
alts_filter = ["non_residential_sqft > 0"]
remove_alts = False
elcm = SegmentedMNLDiscreteChoiceModel(
segmentation_col=seg_col,
sample_size=sample_size,
probability_mode=prob_mode,
choice_mode=choice_mode,
choice_column=choice_col,
default_model_expr=default_model,
alts_fit_filters=None,
)
# alts = orca.get_table('alts_elcm').to_frame()
alts = orca.get_table("zones").to_frame()
choosers = establishments_for_estimation
##define models
model11 = (
"ln_jobs_within_30min+ln_avg_nonres_unit_price_zone+median_year_built+ln_residential_unit_density_zone"
"+ln_pop_within_20min+nonres_far+office+retail_or_restaurant+industrial_building"
seg_col = 'sector_id_retail_agg'
prob_mode = 'single_chooser'
choice_mode = 'aggregate'
choice_col = 'zone_id'
default_model = 'ln_avg_nonres_unit_price_zone+non_residential_sqft_zone+ln_residential_unit_density_zone' \
'+ln_avg_unit_price_zone+median_age_of_head+mean_income+jobs_within_45min+ln_jobs_within_30min'
alts_filter = ['non_residential_sqft > 0']
remove_alts = False
elcm = SegmentedMNLDiscreteChoiceModel(segmentation_col=seg_col,
sample_size=sample_size,
probability_mode=prob_mode,
choice_mode=choice_mode,
choice_column=choice_col,
default_model_expr=default_model,
alts_fit_filters=None)
#alts = orca.get_table('alts_elcm').to_frame()
alts = orca.get_table('buildings').to_frame().fillna(0)
choosers = establishments_for_estimation.fillna(0)
results = elcm.fit(choosers, alts, current_choice='building_id')
elcm.to_yaml('c:/urbansim2/urbansim/urbansim_drcog/config/elcm_yaml.yaml')
##define models
# model11 = 'ln_jobs_within_30min+ln_avg_nonres_unit_price_zone+median_year_built+ln_residential_unit_density_zone' \
# '+ln_pop_within_20min+nonres_far+office+retail_or_restaurant+industrial_building' \
# '+employees_x_ln_non_residential_sqft_zone+ln_dist_rail+rail_within_mile'