features_test = test_features
features_0 = basic_features + mdums + cdums
features_1 = basic_features + mdums + cdums + structural_variables
features_2 = basic_features + mdums + cdums + structural_variables + political_variables
features_3 = basic_features + mdums + cdums + structural_variables + political_variables + survey_variables
features_4 = basic_features + mdums + cdums + structural_variables + political_variables + survey_variables + corona_variables
estimators = 200
model_baseline = api.Model(name="benchmark model",
col_outcome="ged_dummy_sb",
cols_features=features_benchmark,
steps=steps,
periods=periods,
outcome_type="real",
estimator=RandomForestRegressor(
n_jobs=-1,
criterion="mse",
n_estimators=estimators),
tags=["sb"])
model_0 = api.Model(name="basic model",
col_outcome="ged_dummy_sb",
cols_features=features_0,
steps=steps,
periods=periods,
outcome_type="real",
estimator=RandomForestRegressor(n_jobs=-1,
criterion="mse",
n_estimators=estimators),
features_test = test_features
features_0 = basic_features + mdums + cdums
features_1 = basic_features + mdums + cdums + structural_variables + corona_variables
features_2 = basic_features + mdums + cdums + structural_variables + political_variables
features_3 = basic_features + mdums + cdums + structural_variables + corona_variables + political_variables + survey_variables
#features_4 = basic_features + mdums + cdums + structural_variables + political_variables + survey_variables + corona_variables
estimators = 200
model_baseline = api.Model(
name = "benchmark model",
col_outcome= "ged_dummy_sb",
cols_features = features_benchmark,
steps = steps,
periods = periods,
outcome_type = "prob",
delta_outcome = True,
estimator=RandomForestClassifier(n_jobs=-1, n_estimators=estimators),
tags=["sb"]
)
model_d0 = api.Model(
name = "basic model",
col_outcome = "ged_dummy_sb",
cols_features = features_0,
steps = steps,
periods = periods,
outcome_type = "prob",
delta_outcome = True,
estimator = RandomForestClassifier(n_jobs=-1, n_estimators=estimators),
features_m2 = basic_features + structural_variables + corona_variables + political_variables
features_m3 = all_vars
elif task == 4:
features_m1 = basic_features + structural_variables + corona_variables
features_m2 = basic_features + structural_variables + corona_variables + political_variables
features_m3 = all_vars
#number of estimator
estimators = 200
#normal models
model_0 = api.Model(name="basic_model ",
col_outcome="ged_dummy_sb",
cols_features=features_m0,
steps=steps,
periods=periods,
outcome_type="real",
estimator=RandomForestRegressor(n_jobs=-1,
criterion="mse",
n_estimators=estimators),
tags=["sb"])
model_1 = api.Model(name="structural_model ",
col_outcome="ged_dummy_sb",
cols_features=features_m1,
steps=steps,
periods=periods,
outcome_type="real",
estimator=RandomForestRegressor(n_jobs=-1,
criterion="mse",
n_estimators=estimators),
tags=["sb"])
rf = RandomForestClassifier(n_jobs=-1, n_estimators=10_000)
# The currently latest model development run id
run_id = "d_2020_04_01"
periods: List[api.Period] = get_periods(run_id=run_id)
steps = [1, 3, 6, 9, 12, 18, 24, 30, 36, 38]
fullsample = api.Downsampling(share_positive=1.0, share_negative=1.0)
cm_sb_vdem_global = api.Model(
name="cm_sb_vdem_global",
col_outcome=cm["sb_vdem_global"]["col_outcome"],
cols_features=cm["sb_vdem_global"]["cols_features"],
steps=steps,
outcome_type="prob",
estimator=rf,
periods=periods,
downsampling=fullsample,
tags=["train_global"],
)
cm_sb_wdi_global = api.Model(
name="cm_sb_wdi_global",
col_outcome=cm["sb_wdi_global"]["col_outcome"],
cols_features=cm["sb_wdi_global"]["cols_features"],
steps=steps,
outcome_type="prob",
estimator=rf,
periods=periods,
downsampling=fullsample,
tags=["train_global"],
features_0 = basic_features + mdums + cdums
features_1 = benchmark_features
features_2 = basic_features + mdums + cdums + structural_variables + corona_variables + political_variables
#features_1 = political_variables_part
#features_2 = basic_features + mdums + cdums + structural_variables + political_variables
features_3 = basic_features + mdums + cdums + structural_variables + political_variables + survey_variables
#features_4 = basic_features + mdums + cdums + structural_variables
estimators = 200
model_0 = api.Model(name="t4_model_basic",
col_outcome="ged_dummy_sb",
cols_features=features_0,
steps=steps,
periods=periods_t4,
outcome_type="real",
estimator=RandomForestRegressor(n_jobs=-1,
criterion="mse",
n_estimators=estimators),
tags=["sb"])
model_1 = api.Model(name="t4_model_benchmark",
col_outcome="ged_dummy_sb",
cols_features=features_1,
steps=steps,
periods=periods_t4,
outcome_type="real",
estimator=RandomForestRegressor(n_jobs=-1,
criterion="mse",
n_estimators=estimators),
tags=["sb"])
from views.specs.periods import get_periods
log = logging.getLogger(__name__)
rf = RandomForestClassifier(n_jobs=-1, n_estimators=1_000)
# The currently latest model development run id
run_id = "d_2020_04_01"
periods: List[api.Period] = get_periods(run_id=run_id)
steps = [1, 3, 6, 9, 12, 18, 24, 30, 36, 38]
pgm_sb_allthemes = api.Model(
name="pgm_sb_allthemes",
col_outcome=pgm["sb_allthemes"]["col_outcome"],
cols_features=pgm["sb_allthemes"]["cols_features"],
steps=steps,
outcome_type="prob",
estimator=rf,
periods=periods,
tags=["train_africa"],
)
pgm_sb_pgd_natural = api.Model(
name="pgm_sb_pgd_natural",
col_outcome=pgm["sb_pgd_natural"]["col_outcome"],
cols_features=pgm["sb_pgd_natural"]["cols_features"],
steps=steps,
outcome_type="prob",
estimator=rf,
periods=periods,
tags=["train_africa"],
)
pgm_sb_pgd_social = api.Model(
# In[13]:
# Specify number of estimators in RF estimator
n_estimators = 200
# In[14]:
# Define the benchmark models.
benchmark_delta = api.Model(
name="benchmark_delta",
col_outcome="ln_ged_best_sb",
cols_features=cols_features,
steps=steps,
outcome_type="real",
periods=periods,
estimator=RandomForestRegressor(
n_estimators=n_estimators,
criterion="mse",
n_jobs=-1,
),
delta_outcome=True,
downsampling=downsampling,
)
models = [benchmark_delta]
# ## Model fit, prediction, and evaluation
# In[15]:
#get_().run_cell_magic('time', '', '# Train all models\nfor model in models:\n model.fit_estimators(df)')
features_m0_t3 = basic_features
features_m1_t3 = basic_features + structural_variables
features_m2_t3 = basic_features + structural_variables + political_variables
features_m3_t3 = basic_features + structural_variables + political_variables + survey_variables
##number of estimator
estimators = 200
##task 1, normal models
model_0_t1 = api.Model(
name="basic_model_t1",
col_outcome="ged_dummy_sb",
cols_features=features_m0_t1,
steps=steps,
periods=periods_t1,
outcome_type="real",
estimator=RandomForestRegressor(n_jobs=-1,
criterion="mse",
n_estimators=estimators),
tags=["sb"])
model_1_t1 = api.Model(
name="structural_model_t1",
col_outcome="ged_dummy_sb",
cols_features=features_m1_t1,
steps=steps,
periods=periods_t1,
outcome_type="real",
estimator=RandomForestRegressor(n_jobs=-1,
criterion="mse",