【pytorch】用 GRU 做时间序列预测

数据

df.to_csv('traffic1.txt', header=None, index=None)
df

数据为每小时记录一次的交通流量数据，每周有几天出现早高峰。

dataset

import torch
import numpy as np;
from torch.autograd import Variable


def normal_std(x):
    return x.std() * np.sqrt((len(x) - 1.)/(len(x)))

class Data_utility(object):
    # train and valid is the ratio of training set and validation set. test = 1 - train - valid
    def __init__(self, file_name, train, valid, cuda, horizon, window, normalize = 2):
        self.cuda = cuda;
        self.P = window;
        self.h = horizon
        fin = open(file_name);
        self.rawdat = np.loadtxt(fin,delimiter=',')
        if(len(self.rawdat.shape)) == 1:
            self.rawdat = self.rawdat.reshape(len(self.rawdat), -1)
        self.dat = np.zeros(self.rawdat.shape)        
        self.n, self.m = self.dat.shape;
        self.normalize = 2
        self.scale = np.ones(self.m);
        self._normalized(normalize);
        self._split(int(train * self.n), int((train+valid) * self.n), self.n);
        
        self.scale = torch.from_numpy(self.scale).float();
        tmp = self.test[1] * self.scale.expand(self.test[1].size(0), self.m);
            
        if self.cuda:
            self.scale = self.scale.cuda();
        self.scale = Variable(self.scale);
        
        self.rse = normal_std(tmp);
        self.rae = torch.mean(torch.abs(tmp - torch.mean(tmp)));
    
    def _normalized(self, normalize):
        #normalized by the maximum value of entire matrix.
       
        if (normalize == 0):
            self.dat = self.rawdat
            
        if (normalize == 1):
            self.dat = self.rawdat / np.max(self.rawdat);
            
        #normlized by the maximum value of each row(sensor).
        if (normalize == 2):
            for i in range(self.m):
                self.scale[i] = np.max(np.abs(self.rawdat[:,i]));
                self.dat[:,i] = self.rawdat[:,i] / np.max(np.abs(self