GluonTS 快速预测时间序列

目录

头文件

1、定义数据集

2、构建简单前馈网络（MLP）进行预测

3、训练完成之后对测试数据预测进行评估

4、对结果进行画图展示

5、训练代码

---

 

头文件

import mxnet as mx
from mxnet import gluon
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import json
from gluonts.dataset.common import ListDataset

1、定义数据集

定义一个样本数量为10，每个样本有100个时间序列的数据集，数据集预测的长度是后24个数值

N = 10  # number of time series
T = 100  # number of timesteps
prediction_length = 24
freq = "1H"
custom_dataset = np.random.normal(size=(N, T))
start = pd.Timestamp("01-01-2019", freq=freq)  # can be different for each time series

注释：start表示这个时间序列开始的时间，freq表示时间按照这个时间长度进行不断累加（相当于等差数列）

2、构建简单前馈网络（MLP）进行预测

这里构建输入为（100-24），两层隐藏层（40， 40），输出层为24

def creat_SFFEstimator():
    from gluonts.model.simple_feedforward import SimpleFeedForwardEstimator
    from gluonts.trainer import Trainer

    estimator = SimpleFeedForwardEstimator(
        num_hidden_dimensions=[40, 40],
        prediction_length=prediction_length,
        context_length=100,
        freq=freq,
        trainer=Trainer(ctx="cpu",
                        epochs=5,
                        learning_rate=1e-3,
                        num_batches_per_epoch=1
                        ),
    )
    # 使用构建的Estimator（可自定义），在训练集上训练
    predictor = estimator.train(train_ds)
    return predictor

3、训练完成之后对测试数据预测进行评估

def evaluation(predictor):
    from gluonts.evaluation.backtest import make_evaluation_predictions
    from gluonts.evaluation import Evaluator

    forecast_it, ts_it = make_evaluation_predictions(
        dataset=test_ds,  # test dataset
        predictor=predictor,  # predictor
        num_samples=10,  # number of sample paths we want for evaluation
    )
    forecasts = list(forecast_it)
    tss = list(ts_it)

    ts_entry = tss[0]
    forecast_entry = forecasts[0]

    evaluator = Evaluator(quantiles=[0.1, 0.5, 0.9])
    agg_metrics, item_metrics = evaluator(iter(tss), iter(forecasts), num_series=len(test_ds))
    print (agg_metrics)

    return forecast_entry, ts_entry

4、对结果进行画图展示

def plot_prob_forecasts(forecast_entry, ts_entry):
    plot_length = 48
    prediction_intervals = (50.0, 90.0)
    legend = ["observations", "median prediction"] + [f"{k}% prediction interval" for k in prediction_intervals][::-1]

    fig, ax = plt.subplots(1, 1, figsize=(10, 7))
    ts_entry[-plot_length:].plot(ax=ax)  # plot the time series
    forecast_entry.plot(prediction_intervals=prediction_intervals, color='g')
    plt.grid(which="both")
    plt.legend(legend, loc="upper left")
    plt.show()

5、训练代码

def unconstom_train():
    # 创建
    predictor = creat_SFFEstimator()
    # 评估、预测
    forecast_entry, ts_entry = evaluation(predictor)
    # 展示
    plot_prob_forecasts(forecast_entry, ts_entry)

unconstom_train()

实验结果