LightGBMCV

Time series cross validation with LightGBM.

LightGBMCV

LightGBMCV(freq, lags=None, lag_transforms=None, date_features=None, num_threads=1, target_transforms=None)

Create LightGBM CV object.

Parameters:

Name	Type	Description	Default
freq	str or int	Pandas offset alias, e.g. ‘D’, ‘W-THU’ or integer denoting the frequency of the series.	required
lags	list of int	Lags of the target to use as features. Defaults to None.	None
lag_transforms	dict of int to list of functions	Mapping of target lags to their transformations. Defaults to None.	None
date_features	list of str or callable	Features computed from the dates. Can be pandas date attributes or functions that will take the dates as input. Defaults to None.	None
num_threads	int	Number of threads to use when computing the features. Use -1 to use all available CPU cores. Defaults to 1.	1
target_transforms	list of transformers	Transformations that will be applied to the target before computing the features and restored after the forecasting step. Defaults to None.	None

LightGBMCV.fit

fit(df, n_windows, h, id_col='unique_id', time_col='ds', target_col='y', step_size=None, num_iterations=100, params=None, static_features=None, dropna=True, keep_last_n=None, eval_every=10, weights=None, metric='mape', verbose_eval=True, early_stopping_evals=2, early_stopping_pct=0.01, compute_cv_preds=False, before_predict_callback=None, after_predict_callback=None, input_size=None, weight_col=None)

Train boosters simultaneously and assess their performance on the complete forecasting window.

Parameters:

Name	Type	Description	Default
df	pandas DataFrame	Series data in long format.	required
n_windows	int	Number of windows to evaluate.	required
h	int	Forecast horizon.	required
id_col	str	Column that identifies each serie. Defaults to ‘unique_id’.	‘unique_id’
time_col	str	Column that identifies each timestep, its values can be timestamps or integers. Defaults to ‘ds’.	‘ds’
target_col	str	Column that contains the target. Defaults to ‘y’.	‘y’
step_size	int	Step size between each cross validation window. If None it will be equal to h. Defaults to None.	None
num_iterations	int	Maximum number of boosting iterations to run. Defaults to 100.	100
params	dict	Parameters to be passed to the LightGBM Boosters. Defaults to None.	None
static_features	list of str	Names of the features that are static and will be repeated when forecasting. Defaults to None.	None
dropna	bool	Drop rows with missing values produced by the transformations. Defaults to True.	True
keep_last_n	int	Keep only these many records from each serie for the forecasting step. Can save time and memory if your features allow it. Defaults to None.	None
eval_every	int	Number of boosting iterations to train before evaluating on the whole forecast window. Defaults to 10.	10
weights	sequence of float	Weights to multiply the metric of each window. If None, all windows have the same weight. Defaults to None.	None
metric	str or callable	Metric used to assess the performance of the models and perform early stopping. Defaults to ‘mape’.	‘mape’
verbose_eval	bool	Print the metrics of each evaluation.	True
early_stopping_evals	int	Maximum number of evaluations to run without improvement. Defaults to 2.	2
early_stopping_pct	float	Minimum percentage improvement in metric value in early_stopping_evals evaluations. Defaults to 0.01.	0.01
compute_cv_preds	bool	Compute predictions for each window after finding the best iteration. Defaults to False.	False
before_predict_callback	callable	Function to call on the features before computing the predictions. This function will take the input dataframe that will be passed to the model for predicting and should return a dataframe with the same structure. The series identifier is on the index. Defaults to None.	None
after_predict_callback	callable	Function to call on the predictions before updating the targets. This function will take a pandas Series with the predictions and should return another one with the same structure. The series identifier is on the index. Defaults to None.	None
input_size	int	Maximum training samples per serie in each window. If None, will use an expanding window. Defaults to None.	None

Returns:

Type	Description
list of tuple	List of (boosting rounds, metric value) tuples.

LightGBMCV.predict

predict(h, before_predict_callback=None, after_predict_callback=None, X_df=None)

Compute predictions with each of the trained boosters.

Parameters:

Name	Type	Description	Default
h	int	Forecast horizon.	required
before_predict_callback	callable	Function to call on the features before computing the predictions. This function will take the input dataframe that will be passed to the model for predicting and should return a dataframe with the same structure. The series identifier is on the index. Defaults to None.	None
after_predict_callback	callable	Function to call on the predictions before updating the targets. This function will take a pandas Series with the predictions and should return another one with the same structure. The series identifier is on the index. Defaults to None.	None
X_df	DataFrame	Dataframe with the future exogenous features. Should have the id column and the time column. Defaults to None.	None

Returns:

Type	Description
DataFrame	Predictions for each serie and timestep, with one column per window.

LightGBMCV.setup

setup(df, n_windows, h, id_col='unique_id', time_col='ds', target_col='y', step_size=None, params=None, static_features=None, dropna=True, keep_last_n=None, weights=None, metric='mape', input_size=None, weight_col=None)

Initialize internal data structures to iteratively train the boosters. Use this before calling partial_fit.

Parameters:

Name	Type	Description	Default
df	pandas DataFrame	Series data in long format.	required
n_windows	int	Number of windows to evaluate.	required
h	int	Forecast horizon.	required
id_col	str	Column that identifies each serie. Defaults to ‘unique_id’.	‘unique_id’
time_col	str	Column that identifies each timestep, its values can be timestamps or integers. Defaults to ‘ds’.	‘ds’
target_col	str	Column that contains the target. Defaults to ‘y’.	‘y’
step_size	int	Step size between each cross validation window. If None it will be equal to h. Defaults to None.	None
params	dict	Parameters to be passed to the LightGBM Boosters. Defaults to None.	None
static_features	list of str	Names of the features that are static and will be repeated when forecasting. Defaults to None.	None
dropna	bool	Drop rows with missing values produced by the transformations. Defaults to True.	True
keep_last_n	int	Keep only these many records from each serie for the forecasting step. Can save time and memory if your features allow it. Defaults to None.	None
weights	sequence of float	Weights to multiply the metric of each window. If None, all windows have the same weight. Defaults to None.	None
metric	str or callable	Metric used to assess the performance of the models and perform early stopping. Defaults to ‘mape’.	‘mape’
input_size	int	Maximum training samples per serie in each window. If None, will use an expanding window. Defaults to None.	None
weight_col	str	Column containing sample weights. Higher weights increase the influence of those samples during fitting and evaluation.	None

Returns:

Type	Description
LightGBMCV	CV object with internal data structures for partial_fit.

LightGBMCV.partial_fit

partial_fit(num_iterations, before_predict_callback=None, after_predict_callback=None, weight_col=None)

Train the boosters for some iterations.

Parameters:

Name	Type	Description	Default
num_iterations	int	Number of boosting iterations to run	required
before_predict_callback	callable	Function to call on the features before computing the predictions. This function will take the input dataframe that will be passed to the model for predicting and should return a dataframe with the same structure. The series identifier is on the index. Defaults to None.	None
after_predict_callback	callable	Function to call on the predictions before updating the targets. This function will take a pandas Series with the predictions and should return another one with the same structure. The series identifier is on the index. Defaults to None.	None

Returns:

Type	Description
float	Weighted metric after training for num_iterations.

Example

This shows an example with just 4 series of the M4 dataset. If you want to run it yourself on all of them, you can refer to this notebook.

import random

from datasetsforecast.m4 import M4, M4Info
from fastcore.test import test_eq, test_fail
from mlforecast.target_transforms import Differences
from nbdev import show_doc

from mlforecast.lag_transforms import SeasonalRollingMean

group = 'Hourly'
await M4.async_download('data', group=group)
df, *_ = M4.load(directory='data', group=group)
df['ds'] = df['ds'].astype('int')
ids = df['unique_id'].unique()
random.seed(0)
sample_ids = random.choices(ids, k=4)
sample_df = df[df['unique_id'].isin(sample_ids)]
sample_df

	unique_id	ds	y
86796	H196	1	11.8
86797	H196	2	11.4
86798	H196	3	11.1
86799	H196	4	10.8
86800	H196	5	10.6
…	…	…	…
325235	H413	1004	99.0
325236	H413	1005	88.0
325237	H413	1006	47.0
325238	H413	1007	41.0
325239	H413	1008	34.0

info = M4Info[group]
horizon = info.horizon
valid = sample_df.groupby('unique_id').tail(horizon)
train = sample_df.drop(valid.index)
train.shape, valid.shape

((3840, 3), (192, 3))

What LightGBMCV does is emulate LightGBM’s cv function where several Boosters are trained simultaneously on different partitions of the data, that is, one boosting iteration is performed on all of them at a time. This allows to have an estimate of the error by iteration, so if we combine this with early stopping we can find the best iteration to train a final model using all the data or even use these individual models’ predictions to compute an ensemble.

In order to have a good estimate of the forecasting performance of our model we compute predictions for the whole test period and compute a metric on that. Since this step can slow down training, there’s an eval_every parameter that can be used to control this, that is, if eval_every=10 (the default) every 10 boosting iterations we’re going to compute forecasts for the complete window and report the error.

We also have early stopping parameters:

• early_stopping_evals: how many evaluations of the full window should we go without improving to stop training?
• early_stopping_pct: what’s the minimum percentage improvement we want in these early_stopping_evals in order to keep training?

This makes the LightGBMCV class a good tool to quickly test different configurations of the model. Consider the following example, where we’re going to try to find out which features can improve the performance of our model. We start just using lags.

static_fit_config = dict(
    n_windows=2,
    h=horizon,
    params={'verbose': -1},
    compute_cv_preds=True,
)
cv = LightGBMCV(
    freq=1,
    lags=[24 * (i+1) for i in range(7)],  # one week of lags
)

hist = cv.fit(train, **static_fit_config)

[LightGBM] [Info] Start training from score 51.745632
[10] mape: 0.590690
[20] mape: 0.251093
[30] mape: 0.143643
[40] mape: 0.109723
[50] mape: 0.102099
[60] mape: 0.099448
[70] mape: 0.098349
[80] mape: 0.098006
[90] mape: 0.098718
Early stopping at round 90
Using best iteration: 80

By setting compute_cv_preds we get the predictions from each model on their corresponding validation fold.

cv.cv_preds_

	unique_id	ds	y	Booster	window
0	H196	865	15.5	15.522924	0
1	H196	866	15.1	14.985832	0
2	H196	867	14.8	14.667901	0
3	H196	868	14.4	14.514592	0
4	H196	869	14.2	14.035793	0
…	…	…	…	…	…
187	H413	956	59.0	77.227905	1
188	H413	957	58.0	80.589641	1
189	H413	958	53.0	53.986834	1
190	H413	959	38.0	36.749786	1
191	H413	960	46.0	36.281225	1

The individual models we trained are saved, so calling predict returns the predictions from every model trained.

preds = cv.predict(horizon)
preds

	unique_id	ds	Booster0	Booster1
0	H196	961	15.670252	15.848888
1	H196	962	15.522924	15.697399
2	H196	963	14.985832	15.166213
3	H196	964	14.985832	14.723238
4	H196	965	14.562152	14.451092
…	…	…	…	…
187	H413	1004	70.695242	65.917620
188	H413	1005	66.216580	62.615788
189	H413	1006	63.896573	67.848598
190	H413	1007	46.922797	50.981950
191	H413	1008	45.006541	42.752819

We can average these predictions and evaluate them.

def evaluate_on_valid(preds):
    preds = preds.copy()
    preds['final_prediction'] = preds.drop(columns=['unique_id', 'ds']).mean(1)
    merged = preds.merge(valid, on=['unique_id', 'ds'])
    merged['abs_err'] = abs(merged['final_prediction'] - merged['y']) / merged['y']
    return merged.groupby('unique_id')['abs_err'].mean().mean()

eval1 = evaluate_on_valid(preds)
eval1

0.11036194712311806

Now, since these series are hourly, maybe we can try to remove the daily seasonality by taking the 168th (24 * 7) difference, that is, substract the value at the same hour from one week ago, thus our target will be $z_t = y_{t} - y_{t-168}$. The features will be computed from this target and when we predict they will be automatically re-applied.

cv2 = LightGBMCV(
    freq=1,
    target_transforms=[Differences([24 * 7])],
    lags=[24 * (i+1) for i in range(7)],
)
hist2 = cv2.fit(train, **static_fit_config)

[LightGBM] [Info] Start training from score 0.519010
[10] mape: 0.089024
[20] mape: 0.090683
[30] mape: 0.092316
Early stopping at round 30
Using best iteration: 10

assert hist2[-1][1] < hist[-1][1]

Nice! We achieve a better score in less iterations. Let’s see if this improvement translates to the validation set as well.

preds2 = cv2.predict(horizon)
eval2 = evaluate_on_valid(preds2)
eval2

0.08956665504570135

assert eval2 < eval1

Great! Maybe we can try some lag transforms now. We’ll try the seasonal rolling mean that averages the values “every season”, that is, if we set season_length=24 and window_size=7 then we’ll average the value at the same hour for every day of the week.

cv3 = LightGBMCV(
    freq=1,
    target_transforms=[Differences([24 * 7])],
    lags=[24 * (i+1) for i in range(7)],
    lag_transforms={
        48: [SeasonalRollingMean(season_length=24, window_size=7)],
    },
)
hist3 = cv3.fit(train, **static_fit_config)

[LightGBM] [Info] Start training from score 0.273641
[10] mape: 0.086724
[20] mape: 0.088466
[30] mape: 0.090536
Early stopping at round 30
Using best iteration: 10

Seems like this is helping as well!

assert hist3[-1][1] < hist2[-1][1]

Does this reflect on the validation set?

preds3 = cv3.predict(horizon)
eval3 = evaluate_on_valid(preds3)
eval3

0.08961279023129345

Nice! mlforecast also supports date features, but in this case our time column is made from integers so there aren’t many possibilites here. As you can see this allows you to iterate faster and get better estimates of the forecasting performance you can expect from your model.

If you’re doing hyperparameter tuning it’s useful to be able to run a couple of iterations, assess the performance, and determine if this particular configuration isn’t promising and should be discarded. For example, optuna has pruners that you can call with your current score and it decides if the trial should be discarded. We’ll now show how to do that.

Since the CV requires a bit of setup, like the LightGBM datasets and the internal features, we have this setup method.

cv4 = LightGBMCV(
    freq=1,
    lags=[24 * (i+1) for i in range(7)],
)
cv4.setup(
    train,
    n_windows=2,
    h=horizon,
    params={'verbose': -1},
)

LightGBMCV(freq=1, lag_features=['lag24', 'lag48', 'lag72', 'lag96', 'lag120', 'lag144', 'lag168'], date_features=[], num_threads=1, bst_threads=8)

Once we have this we can call partial_fit to only train for some iterations and return the score of the forecast window.

score = cv4.partial_fit(10)
score

[LightGBM] [Info] Start training from score 51.745632

0.5906900462828166

This is equal to the first evaluation from our first example.

assert hist[0][1] == score

We can now use this score to decide if this configuration is promising. If we want to we can train some more iterations.

score2 = cv4.partial_fit(20)

This is now equal to our third metric from the first example, since this time we trained for 20 iterations.

assert hist[2][1] == score2

Using a custom metric

The built-in metrics are MAPE and RMSE, which are computed by serie and then averaged across all series. If you want to do something different or use a different metric entirely, you can define your own metric like the following:

def weighted_mape(
    y_true: pd.Series,
    y_pred: pd.Series,
    ids: pd.Series,
    dates: pd.Series,
):
    """Weighs the MAPE by the magnitude of the series values"""
    abs_pct_err = abs(y_true - y_pred) / abs(y_true)
    mape_by_serie = abs_pct_err.groupby(ids).mean()
    totals_per_serie = y_pred.groupby(ids).sum()
    series_weights = totals_per_serie / totals_per_serie.sum()
    return (mape_by_serie * series_weights).sum()

_ = LightGBMCV(
    freq=1,
    lags=[24 * (i+1) for i in range(7)],
).fit(
    train,
    n_windows=2,
    h=horizon,
    params={'verbose': -1},
    metric=weighted_mape,
)

[LightGBM] [Info] Start training from score 51.745632
[10] weighted_mape: 0.480353
[20] weighted_mape: 0.218670
[30] weighted_mape: 0.161706
[40] weighted_mape: 0.149992
[50] weighted_mape: 0.149024
[60] weighted_mape: 0.148496
Early stopping at round 60
Using best iteration: 60