NeuralProphet之六：多元时间序列预测-CFANZ编程社区

NeuralProphet之六：多元时间序列预测

时间序列论文: NeuralProphet: Explainable Forecasting at Scale

NeuralProphet之一：安装与使用

NeuralProphet之二：季节性(Seasonality)

NeuralProphet之三：回归(Regressors)

NeuralProphet之四：事件(Events)

NeuralProphet之五：多时序预测模型

NeuralProphet之六：多元时间序列预测

NeuralProphet之七：NeuralProphet + Optuna

NeuralProphet之八：NeuralProphet部署

NeuralProphet官方示例一：建筑物用电量预测(Building load forecasting)

NeuralProphet官方示例二：日照辐射强度预测(Forecasting hourly solar irradiance)

NeuralProphet之六：多元时间序列预测
NeuralProphet 通过滞后回归(Lagged Regressors)为时间序列预测目标加入其他协变量。

导入库：

import pandas as pd
import matplotlib.pyplot as plt
from neuralprophet import NeuralProphet, set_log_level
set_log_level("ERROR")

导入数据：

# data_location = "https://raw.githubusercontent.com/ourownstory/neuralprophet-data/main/datasets/"
data_location = 'datasets/'
df_ercot = pd.read_csv(data_location + "multivariate/load_ercot_regions.csv")
df_ercot_y = pd.read_csv(data_location + "energy/load_ercot.csv")
df_ercot['y'] = df_ercot_y['y']
print(df_ercot.head())

数据情况：

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 154854 entries, 0 to 154853
Data columns (total 10 columns):
 #   Column    Non-Null Count   Dtype  
---  ------    --------------   -----  
 0   ds        154854 non-null  object 
 1   COAST     154854 non-null  float64
 2   EAST      154854 non-null  float64
 3   FAR_WEST  154854 non-null  float64
 4   NORTH     154854 non-null  float64
 5   NORTH_C   154854 non-null  float64
 6   SOUTHERN  154854 non-null  float64
 7   SOUTH_C   154854 non-null  float64
 8   WEST      154854 non-null  float64
 9   y         154854 non-null  float64
dtypes: float64(9), object(1)
memory usage: 11.8+ MB

NeuralProphet通过add_lagged_regressor注册协变量

df = df_ercot
m = NeuralProphet(
    n_forecasts=24,
    n_lags=24,
    learning_rate=0.01,
)
m = m.add_lagged_regressor(names=regions)
m.highlight_nth_step_ahead_of_each_forecast(24)
metrics = m.fit(df, freq="H")

预测

forecast = m.predict(df)
# fig = m.plot(forecast)
fig1 = m.plot(forecast[-365*24:])
fig2 = m.plot(forecast[-7*24:])
# comp = m.plot_components(forecast[-7*24:])
param = m.plot_parameters()