%load_ext autoreload
%autoreload 2SOFTS
from fastcore.test import test_eq
from nbdev.showdoc import show_docSOFTS（系列核心融合时间序列）引入了新颖的星级聚合调度（STAD）模块。STAD模块采用集中策略，而不是通过分布式架构（如注意力机制）学习通道交互，在这种策略中，系列被聚合形成一个全球核心表示，同时保持线性复杂度。
参考文献 - 卢汉, 陈旭阳, 叶汉佳, 詹德川. “SOFTS: 高效的多元时间序列预测与系列核心融合”
import torch
import torch.nn as nn
import torch.nn.functional as F
from neuralforecast.losses.pytorch import MAE
from neuralforecast.common._base_multivariate import BaseMultivariate
from neuralforecast.common._modules import TransEncoder, TransEncoderLayer1. 辅助函数
1.1 嵌入
class DataEmbedding_inverted(nn.Module):
"""
数据嵌入
"""
def __init__(self, c_in, d_model, dropout=0.1):
super(DataEmbedding_inverted, self).__init__()
self.value_embedding = nn.Linear(c_in, d_model)
self.dropout = nn.Dropout(p=dropout)
def forward(self, x, x_mark):
x = x.permute(0, 2, 1)
# x: [批次变量时间]
if x_mark is None:
x = self.value_embedding(x)
else:
# 具备将协变量（如时间戳）作为标记的潜力
x = self.value_embedding(torch.cat([x, x_mark.permute(0, 2, 1)], 1))
# x: [批次变量 d_模型]
return self.dropout(x)1.2 STAD（星形聚合调度）
class STAD(nn.Module):
"""
星型聚合调度模块
"""
def __init__(self, d_series, d_core):
super(STAD, self).__init__()
self.gen1 = nn.Linear(d_series, d_series)
self.gen2 = nn.Linear(d_series, d_core)
self.gen3 = nn.Linear(d_series + d_core, d_series)
self.gen4 = nn.Linear(d_series, d_series)
def forward(self, input, *args, **kwargs):
batch_size, channels, d_series = input.shape
# 设置前馈网络
combined_mean = F.gelu(self.gen1(input))
combined_mean = self.gen2(combined_mean)
# 随机池化
if self.training:
ratio = F.softmax(combined_mean, dim=1)
ratio = ratio.permute(0, 2, 1)
ratio = ratio.reshape(-1, channels)
indices = torch.multinomial(ratio, 1)
indices = indices.view(batch_size, -1, 1).permute(0, 2, 1)
combined_mean = torch.gather(combined_mean, 1, indices)
combined_mean = combined_mean.repeat(1, channels, 1)
else:
weight = F.softmax(combined_mean, dim=1)
combined_mean = torch.sum(combined_mean * weight, dim=1, keepdim=True).repeat(1, channels, 1)
# 多层感知器融合
combined_mean_cat = torch.cat([input, combined_mean], -1)
combined_mean_cat = F.gelu(self.gen3(combined_mean_cat))
combined_mean_cat = self.gen4(combined_mean_cat)
output = combined_mean_cat
return output, None2. 模型
class SOFTS(BaseMultivariate):
""" SOFTS
**Parameters:**<br>
`h`: int, Forecast horizon. <br>
`input_size`: int, autorregresive inputs size, y=[1,2,3,4] input_size=2 -> y_[t-2:t]=[1,2].<br>
`n_series`: int, number of time-series.<br>
`futr_exog_list`: str list, future exogenous columns.<br>
`hist_exog_list`: str list, historic exogenous columns.<br>
`stat_exog_list`: str list, static exogenous columns.<br>
`hidden_size`: int, dimension of the model.<br>
`d_core`: int, dimension of core in STAD.<br>
`e_layers`: int, number of encoder layers.<br>
`d_ff`: int, dimension of fully-connected layer.<br>
`dropout`: float, dropout rate.<br>
`use_norm`: bool, whether to normalize or not.<br>
`loss`: PyTorch module, instantiated train loss class from [losses collection](https://nixtla.github.io/neuralforecast/losses.pytorch.html).<br>
`valid_loss`: PyTorch module=`loss`, instantiated valid loss class from [losses collection](https://nixtla.github.io/neuralforecast/losses.pytorch.html).<br>
`max_steps`: int=1000, maximum number of training steps.<br>
`learning_rate`: float=1e-3, Learning rate between (0, 1).<br>
`num_lr_decays`: int=-1, Number of learning rate decays, evenly distributed across max_steps.<br>
`early_stop_patience_steps`: int=-1, Number of validation iterations before early stopping.<br>
`val_check_steps`: int=100, Number of training steps between every validation loss check.<br>
`batch_size`: int=32, number of different series in each batch.<br>
`step_size`: int=1, step size between each window of temporal data.<br>
`scaler_type`: str='identity', type of scaler for temporal inputs normalization see [temporal scalers](https://nixtla.github.io/neuralforecast/common.scalers.html).<br>
`random_seed`: int=1, random_seed for pytorch initializer and numpy generators.<br>
`num_workers_loader`: int=os.cpu_count(), workers to be used by `TimeSeriesDataLoader`.<br>
`drop_last_loader`: bool=False, if True `TimeSeriesDataLoader` drops last non-full batch.<br>
`alias`: str, optional, Custom name of the model.<br>
`optimizer`: Subclass of 'torch.optim.Optimizer', optional, user specified optimizer instead of the default choice (Adam).<br>
`optimizer_kwargs`: dict, optional, list of parameters used by the user specified `optimizer`.<br>
`lr_scheduler`: Subclass of 'torch.optim.lr_scheduler.LRScheduler', optional, user specified lr_scheduler instead of the default choice (StepLR).<br>
`lr_scheduler_kwargs`: dict, optional, list of parameters used by the user specified `lr_scheduler`.<br>
`**trainer_kwargs`: int, keyword trainer arguments inherited from [PyTorch Lighning's trainer](https://pytorch-lightning.readthedocs.io/en/stable/api/pytorch_lightning.trainer.trainer.Trainer.html?highlight=trainer).<br>
**References**<br>
[Lu Han, Xu-Yang Chen, Han-Jia Ye, De-Chuan Zhan. "SOFTS: Efficient Multivariate Time Series Forecasting with Series-Core Fusion"](https://arxiv.org/pdf/2404.14197)
"""
# Class attributes
SAMPLING_TYPE = 'multivariate'
EXOGENOUS_FUTR = False
EXOGENOUS_HIST = False
EXOGENOUS_STAT = False
def __init__(self,
h,
input_size,
n_series,
futr_exog_list = None,
hist_exog_list = None,
stat_exog_list = None,
hidden_size: int = 512,
d_core: int = 512,
e_layers: int = 2,
d_ff: int = 2048,
dropout: float = 0.1,
use_norm: bool = True,
loss = MAE(),
valid_loss = None,
max_steps: int = 1000,
learning_rate: float = 1e-3,
num_lr_decays: int = -1,
early_stop_patience_steps: int =-1,
val_check_steps: int = 100,
batch_size: int = 32,
step_size: int = 1,
scaler_type: str = 'identity',
random_seed: int = 1,
num_workers_loader: int = 0,
drop_last_loader: bool = False,
optimizer = None,
optimizer_kwargs = None,
lr_scheduler = None,
lr_scheduler_kwargs = None,
**trainer_kwargs):
super(SOFTS, self).__init__(h=h,
input_size=input_size,
n_series=n_series,
stat_exog_list = None,
futr_exog_list = None,
hist_exog_list = None,
loss=loss,
valid_loss=valid_loss,
max_steps=max_steps,
learning_rate=learning_rate,
num_lr_decays=num_lr_decays,
early_stop_patience_steps=early_stop_patience_steps,
val_check_steps=val_check_steps,
batch_size=batch_size,
step_size=step_size,
scaler_type=scaler_type,
random_seed=random_seed,
num_workers_loader=num_workers_loader,
drop_last_loader=drop_last_loader,
optimizer=optimizer,
optimizer_kwargs=optimizer_kwargs,
lr_scheduler=lr_scheduler,
lr_scheduler_kwargs=lr_scheduler_kwargs,
**trainer_kwargs)
self.h = h
self.enc_in = n_series
self.dec_in = n_series
self.c_out = n_series
self.use_norm = use_norm
# Architecture
self.enc_embedding = DataEmbedding_inverted(input_size,
hidden_size,
dropout)
self.encoder = TransEncoder(
[
TransEncoderLayer(
STAD(hidden_size, d_core),
hidden_size,
d_ff,
dropout=dropout,
activation=F.gelu
) for l in range(e_layers)
]
)
self.projection = nn.Linear(hidden_size, self.h, bias=True)
def forecast(self, x_enc):
# Normalization from Non-stationary Transformer
if self.use_norm:
means = x_enc.mean(1, keepdim=True).detach()
x_enc = x_enc - means
stdev = torch.sqrt(torch.var(x_enc, dim=1, keepdim=True, unbiased=False) + 1e-5)
x_enc /= stdev
_, _, N = x_enc.shape
enc_out = self.enc_embedding(x_enc, None)
enc_out, attns = self.encoder(enc_out, attn_mask=None)
dec_out = self.projection(enc_out).permute(0, 2, 1)[:, :, :N]
# De-Normalization from Non-stationary Transformer
if self.use_norm:
dec_out = dec_out * (stdev[:, 0, :].unsqueeze(1).repeat(1, self.h, 1))
dec_out = dec_out + (means[:, 0, :].unsqueeze(1).repeat(1, self.h, 1))
return dec_out
def forward(self, windows_batch):
insample_y = windows_batch['insample_y']
y_pred = self.forecast(insample_y)
y_pred = y_pred[:, -self.h:, :]
y_pred = self.loss.domain_map(y_pred)
# 在 n_series == 1 的情况下，domain_map 可能已经压缩了最后一个维度。
if y_pred.ndim == 2:
return y_pred.unsqueeze(-1)
else:
return y_predshow_doc(SOFTS)show_doc(SOFTS.fit, name='SOFTS.fit')show_doc(SOFTS.predict, name='SOFTS.predict')3. 使用示例
import pandas as pd
import matplotlib.pyplot as plt
from neuralforecast import NeuralForecast
from neuralforecast.models import SOFTS
from neuralforecast.utils import AirPassengersPanel, AirPassengersStatic
from neuralforecast.losses.pytorch import MSE
Y_train_df = AirPassengersPanel[AirPassengersPanel.ds<AirPassengersPanel['ds'].values[-12]].reset_index(drop=True) # 132次列车
Y_test_df = AirPassengersPanel[AirPassengersPanel.ds>=AirPassengersPanel['ds'].values[-12]].reset_index(drop=True) # 12项测试
model = SOFTS(h=12,
input_size=24,
n_series=2,
hidden_size=256,
d_core=256,
e_layers=2,
d_ff=64,
dropout=0.1,
use_norm=True,
loss=MSE(),
valid_loss=MAE(),
early_stop_patience_steps=3,
batch_size=32)
fcst = NeuralForecast(models=[model], freq='M')
fcst.fit(df=Y_train_df, static_df=AirPassengersStatic, val_size=12)
forecasts = fcst.predict(futr_df=Y_test_df)
# 情节预测
fig, ax = plt.subplots(1, 1, figsize = (20, 7))
Y_hat_df = forecasts.reset_index(drop=False).drop(columns=['unique_id','ds'])
plot_df = pd.concat([Y_test_df, Y_hat_df], axis=1)
plot_df = pd.concat([Y_train_df, plot_df])
plot_df = plot_df[plot_df.unique_id=='Airline1'].drop('unique_id', axis=1)
plt.plot(plot_df['ds'], plot_df['y'], c='black', label='True')
plt.plot(plot_df['ds'], plot_df['SOFTS'], c='blue', label='Forecast')
ax.set_title('AirPassengers Forecast', fontsize=22)
ax.set_ylabel('Monthly Passengers', fontsize=20)
ax.set_xlabel('Year', fontsize=20)
ax.legend(prop={'size': 15})
ax.grid()Give us a ⭐ on Github