1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
|
# pylint: disable=too-many-arguments
"""Probabilistic Exponential Smoothing Model"""
__docformat__ = "numpy"
import logging
from typing import List, Optional, Union
import numpy
import pandas as pd
from nixtlats import TimeGPT
from openbb_terminal.core.session.current_user import get_current_user
from openbb_terminal.decorators import check_api_key, log_start_end
logger = logging.getLogger(__name__)
@log_start_end(log=logger)
@check_api_key(["API_KEY_NIXTLA"])
def get_timegpt_model(
data: Union[pd.Series, pd.DataFrame],
time_col: str = "ds",
target_col: str = "y",
forecast_horizon: int = 12,
levels: Optional[List[float]] = None,
freq: Union[str, None] = None,
finetune_steps: int = 0,
clean_ex_first: bool = True,
residuals: bool = False,
date_features: Optional[List[str]] = None,
) -> pd.DataFrame:
"""TimeGPT was trained on the largest collection of data in history -
over 100 billion rows of financial, weather, energy, and web data -
and democratizes the power of time-series analysis.
Parameters
----------
data : Union[pd.Series, pd.DataFrame]
Input data.
time_col: str:
Column that identifies each timestep, its values can be timestamps or integers. Defaults to "ds".
target_column: str:
Target column to forecast. Defaults to "y".
forecast_horizon: int
Number of days to forecast. Defaults to 12.
levels: List[float]
Confidence levels between 0 and 100 for prediction intervals.
freq: Optional[str, None]
Frequency of the data. By default, the freq will be inferred automatically.
finetune_steps: int
Number of steps used to finetune TimeGPT in the new data.
clean_ex_first: bool
Clean exogenous signal before making forecasts using TimeGPT.
residuals: bool
Whether to show residuals for the model. Defaults to False.
date_features: Optional[List[str]]
Specifies which date attributes have highest weight according to model.
Returns
-------
pd.DataFrame
Forecasted values.
"""
timegpt = TimeGPT(
token=get_current_user().credentials.API_KEY_NIXTLA,
)
if levels is None:
levels = [80, 95]
if isinstance(data[time_col].values[0], pd.Timestamp):
data[time_col] = data[time_col].dt.strftime("%Y-%m-%d")
elif isinstance(data[time_col].values[0], numpy.datetime64):
data[time_col] = pd.to_datetime(data[time_col]).dt.strftime("%Y-%m-%d")
date_features_param = True if "auto" in date_features else date_features # type: ignore
fcst_df = timegpt.forecast(
data,
time_col=time_col,
target_col=target_col,
h=forecast_horizon,
freq=freq,
level=levels,
finetune_steps=finetune_steps,
clean_ex_first=clean_ex_first,
add_history=residuals,
date_features=date_features_param if date_features else False,
)
return fcst_df, timegpt.weights_x
|
