def test_netchop_improvement(key):
res = Poisson(
ddf[key].values,
add_constant(ddf.method_simultaneous)
).fit()
print(res.summary())
return res
"station_diur_temp_rng_c", "precipitation_amt_mm",
"reanalysis_dew_point_temp_k", "reanalysis_air_temp_k",
"reanalysis_relative_humidity_percent",
"reanalysis_specific_humidity_g_per_kg", "reanalysis_precip_amt_kg_per_m2",
"reanalysis_max_air_temp_k", "reanalysis_min_air_temp_k",
"reanalysis_avg_temp_k", "reanalysis_tdtr_k", "ndvi_se", "ndvi_sw",
"ndvi_ne", "ndvi_nw"
]
n_features = len(features_list)
df_train_features = df_train_features.fillna(df_train_features.mean())
df_test_features = df_test_features.fillna(df_test_features.mean())
X_train = df_train_features[features_list].values
X_test = df_test_features[features_list].values
y_train = df_train_labels["total_cases"].values
# Model:
poisson_mod = Poisson(endog=y_train, exog=X_train).fit(maxiter=61)
print(poisson_mod.summary())
predictions = poisson_mod.predict(X_test)
predictions_rounded = np.rint(predictions).astype(np.int64)
print(predictions_rounded)
write_result(predictions_rounded,
"/poisson.csv",
sample_source=sample_submission_path,
write_source=predictions_path)
