# df: DataFrame = dl.load_df_corn_pkl_smpl_25_20171018().sample(2000)
df: DataFrame = dl.load_df_corn_pkl_smpl_25_20171018()
logging.debug("data shape: %s", df.shape)
y = df.pop('Dry_Yield')
X = df
###
# transform
###
X, label_cols = preprocessing.shape_gis_pps(X)
transform_pipe = make_one_hot_pipeline(X, label_cols, [
preprocessing.FillNaTransformer(),
preprocessing.NumericTransformer(),
StandardScaler()
])
###
# Run Model
###
kcv = KFold(n_splits=10, shuffle=True, random_state=972)
kf_runs = itertools.islice(kcv.split(X), 3)
scores = []
for i, (train_split_idx, test_split_idx) in enumerate(kf_runs):
log.info("Running kfold: %s", i)
X_train_split, y_train_split = X.iloc[train_split_idx], y.iloc[
# df: DataFrame = dl.load_df_corn_pkl_smpl_25_20171018().sample(2000)
# dl.dump_sample(df, 'df-corn-20171018-scratch-debug.pickle')
df: DataFrame = dl.load_pickled('df-corn-20171018-scratch-debug.pickle')
# df: DataFrame = dl.load_df_corn_pkl_smpl_25_20171018()
log.debug("data shape: %s", df.shape)
y = df.pop('Dry_Yield')
X = df
###
# transform pipeline setup
###
X, label_cols = dl.shape_gis_pps(X)
transform_pipe, one_hot_label_enc = make_one_hot_pipeline(
X, label_cols,
[preprocessing.FillNaTransformer(),
preprocessing.NumericTransformer()])
# , StandardScaler()
###
# Run Model
###
def save_plot(name: str):
plt.tight_layout(pad=1.25)
plt.savefig(create_result_file_path(name))
kcv = KFold(n_splits=7, shuffle=True, random_state=988)
kf_runs = itertools.islice(kcv.split(X), 2)