LSTMrawmaterial.py

# ANN for av count of tata mills using LSTM
import math
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import matplotlib as mpl
import seaborn as sns
import keras
from keras import backend as K
from keras.models import Sequential
from keras.layers import Dense, LSTM
from keras import losses
from keras.utils import plot_model
from sklearn.preprocessing import MinMaxScaler
from sklearn.metrics import mean_squared_error

np.random.seed(7)
dataset = pd.read_csv('tata-raw_material.csv', parse_dates = True, index_col = 0)
data = dataset.values
data = dataset.astype('float32')

# normalize the dataset
scaler = MinMaxScaler(feature_range=(0, 1))
dataset = scaler.fit_transform(data)
from sklearn.model_selection import train_test_split
train, test = train_test_split(dataset, test_size = 0.2)
def prepare_data(data, lags = 1):
    ''' Create lagged data from an input time series '''
    X, y = [], []
    for row in range(len(data) - lags - 1):
        a = data[row:(row + lags), 0]
        X.append(a)
        y.append(data[row + lags, 0])
    return np.array(X), np.array(y)
lags = 6
X_train, y_train = prepare_data(train, lags)
X_test, y_test = prepare_data(test, lags)

# reshape input to be [samples, time steps, features]
X_train = np.reshape(X_train, (X_train.shape[0], 1, X_train.shape[1]))
X_test = np.reshape(X_test, (X_test.shape[0], 1, X_test.shape[1]))

# create and fit the LSTM network
mdl = Sequential()
mdl.add(Dense(6, input_shape=(1, lags), activation='relu'))
mdl.add(LSTM(4, activation='relu'))
mdl.add(Dense(1, activation='relu'))
mdl.compile(loss='mean_squared_error', optimizer='adam')
mdl.fit(X_train, y_train, epochs=300, batch_size=1, verbose=2)

# make predictions
train_predict = mdl.predict(X_train)
test_predict = mdl.predict(X_test)

# invert transformation
train_predict = scaler.inverse_transform(train_predict)
y_train = scaler.inverse_transform([y_train])
test_predict = scaler.inverse_transform(test_predict)
y_test = scaler.inverse_transform([y_test])

# calculate root mean squared error
train_score = math.sqrt(mean_squared_error(y_train[0], train_predict[:,0]))
print('Train Score: {:.2f} RMSE'.format(train_score))
test_score = math.sqrt(mean_squared_error(y_test[0], test_predict[:,0]))
print('Test Score: {:.2f} RMSE'.format(test_score))

# shift train predictions for plotting
train_predict_plot = np.empty_like(data)
train_predict_plot[:, :] = np.nan
train_predict_plot[lags:len(train_predict)+lags, :] = train_predict
 
# shift test predictions for plotting
test_predict_plot = np.empty_like(data)
test_predict_plot[:, :] = np.nan
test_predict_plot[len(train_predict) + (lags * 2)+1:len(data)-1, :] = test_predict
 
# plot observation and predictions
plt.plot(data.values, label='Observed', color='#006699')
plt.plot(train_predict_plot, label='Prediction for Train Set', color='#006699', alpha=0.5)
plt.plot(test_predict_plot, label='Prediction for Test Set', color='#ff0066')
plt.legend(loc='upper left')
plt.title('LSTM Recurrent Neural Net')
plt.show()

# plot prediction with observation
mse = ((y_test.reshape(-1, 1) - test_predict.reshape(-1, 1)) ** 2).mean()
plt.title('Prediction quality: {:.2f} MSE ({:.2f} RMSE)'.format(mse, math.sqrt(mse)))
plt.plot(y_test.reshape(-1, 1), label='Observed', color='#006699')
plt.plot(test_predict.reshape(-1, 1), label='Prediction', color='#ff0066')
plt.legend(loc='upper left');
plt.show()