categorical_cols = ["equipo"]
lag_list = [1]
rolling_list = [1, 3, 6, 12, 24]
#df_ewm = interact_categorical_numerical(
# df, lag_col, numerical_cols,
# categorical_cols, lag_list,
# rolling_list, agg_funct="sum",
# rolling_function = "ewm", freq=None,
# group_name=None, store_name=False)
df_rolling = interact_categorical_numerical(df,
lag_col,
numerical_cols,
categorical_cols,
lag_list,
rolling_list,
agg_funct="sum",
rolling_function="rolling",
freq=None,
group_name=None,
store_name=False)
#df_expansion = interact_categorical_numerical(
# df, lag_col, numerical_cols,
# categorical_cols, lag_list,
# rolling_list, agg_funct="sum",
# rolling_function = "expanding", freq=None,
# group_name=None, store_name=False)
id_columns = ['equipo', "partido_equipo_num"]
#df= df.merge(df_ewm, on = id_columns )
#df= df.merge(df_expansion, on = id_columns )
new_numerical_cols.append(name)
numerical_cols = numerical_cols + new_numerical_cols
print(numerical_cols)
#Processing lag and window functions.
lag_col = "date"
categorical_cols = ["fullVisitorId"]
lag_list = [0, 1, 3, 7]
rolling_list = [1, 3, 7, 14]
df_ewm = interact_categorical_numerical(df_timeseries,
lag_col,
numerical_cols,
categorical_cols,
lag_list,
rolling_list,
agg_funct="sum",
rolling_function="ewm",
freq=None,
group_name=None,
store_name=False)
df_ewm = clean_dataset(df_ewm)
df_ewm = df_ewm.replace(np.nan, 0)
print("TEST: ", len(df))
df = df.merge(df_ewm, on=["fullVisitorId", "date"], how="inner")
print("TEST: ", len(df))
del df_ewm
df.to_csv("../input/df_rolling_past.csv")
#Processing lag and window functions.