I am currently playing around with pytorch models for time series prediction. I have managed to successfully run a model to predict test data. I was wondering how can I use it to predict beyond test data? I will attach my code below. I essentially want the model to continue running for say 1000 more points after the test data.

import torch
import torch.nn as nn
import torchvision.transforms as transforms
import torchvision.datasets as dsets
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import sys
from torch.autograd import Variable
from sklearn.preprocessing import MinMaxScaler

# Load the data set
data_test = np.loadtxt('data.txt')
training_set = np.reshape(data,(len(data),1))

def sliding_windows(data, seq_length):
    x = []
    y = []

    for i in range(len(data)-seq_length-1):
        _x = data[i:(i+seq_length)]
        _y = data[i+seq_length]

    return np.array(x),np.array(y)

sc = MinMaxScaler()
training_data = sc.fit_transform(training_set)

batch_size = 4
x, y = sliding_windows(training_data, batch_size)
train_size = int(len(y) * 0.5)
test_size = len(y) - train_size

data_X = Variable(torch.Tensor(np.array(x)))
data_Y = Variable(torch.Tensor(np.array(y)))

training_set_X = Variable(torch.Tensor(np.array(x[0:train_size])))
training_set_Y = Variable(torch.Tensor(np.array(y[0:train_size])))

testing_set_X = Variable(torch.Tensor(np.array(x[train_size:len(x)])))
testing_set_Y = Variable(torch.Tensor(np.array(y[train_size:len(y)])))

class LSTMModel(nn.Module):
    def __init__(self, input_dim, hidden_dim, layer_dim, output_dim):
        super(LSTMModel, self).__init__()
        # Hidden dimensions
        self.hidden_dim = hidden_dim

        # Number of hidden layers
        self.layer_dim = layer_dim

        # Inlcude dropout if you want (decided not to since it did not seem to affect results very much)
        #self.dropout = nn.Dropout(p=0.5)

        # Builds the LSTM
        self.lstm = nn.LSTM(input_dim, hidden_dim, layer_dim, batch_first=True)

        # Readout layer
        # Linear = Applies a linear transformation to the incoming data: y = xW^T + b
        self.fc = nn.Linear(hidden_dim, output_dim)

    def forward(self, x):
        # Initialize hidden state
        h0 = torch.zeros(self.layer_dim, x.size(0), self.hidden_dim).requires_grad_()

        # Initialize cell state
        c0 = torch.zeros(self.layer_dim, x.size(0), self.hidden_dim).requires_grad_()

        # We need to detach as we are doing truncated backpropagation through time (BPTT)
        # If we don't, we'll backprop all the way to the start even after going through another batch
        out, (hn, cn) = self.lstm(x, (h0.detach(), c0.detach()))

        # Index hidden state of last time step
        # out[:, -1, :] --> Last time step hidden state
        out = self.fc(out[:, -1, :]) 
        return out

input_dim = 1
hidden_dim = 50
layer_dim = 1
output_dim = 1
num_epochs = 100
lr = 0.1

model = LSTMModel(input_dim, hidden_dim, layer_dim, output_dim)

criterion =  torch.nn.MSELoss() 

optimizer = torch.optim.Adam(model.parameters(), lr=lr)  

decayRate = 1
lr_scheduler = torch.optim.lr_scheduler.ExponentialLR(optimizer = optimizer, gamma = decayRate)

for epoch in range(num_epochs):
    outputs = model(training_set_X)
    loss = criterion(outputs, training_set_Y)

train_predict = model(data_X)

data_predict = train_predict.data.numpy()
dataY_plot = data_Y.data.numpy()

data_predict = sc.inverse_transform(data_predict)
dataY_plot = sc.inverse_transform(dataY_plot)

dataY_plot = np.array(dataY_plot)
data_predict = np.array(data_predict)

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.