Import As Import As

1.
import numpy as np
2. import matplotlib.pyplot as plt
Seed
1. np.random.seed(0)
Plot
1. t = np.arange(0,1500)
2. x = np.sin(0.015*t) + np.random.uniform(low=-1, high=1, size=(1500,))
3. x_actual = np.sin(0.015*t)
4. plt.plot(x)
5. plt.plot(x_actual)
Import
1. from sklearn.preprocessing import MinMaxScaler
Normalize
1. normalizer = MinMaxScaler(feature_range=(0, 1))
2. x = (np.reshape(x, (-1, 1)))
3. x = normalizer.fit_transform(x)
1. print(x)
[[0.53046776] [0.61911792] [0.56562058] ... [0.19383459] [0.1547965

] [0.24593019]]
Create Dataset
1. train = x[0:1000]
2. test = x[1000:]
1. print(train.shape)
(1000, 1)
1. def createDataset(data, step):
2. X, Y =[], []
3. for i in range(len(data)-step):
4. X.append(data[i:i+step])
5. Y.append(data[i+step])
6. return np.array(X), np.array(Y)
7.
8. step = 10
9. trainX,trainY = createDataset(train,step)
10. testX,testY = createDataset(test,step)
1. print(trainX[0])
[[0.53046776] [0.61911792] [0.56562058] [0.5399316 ] [0.48194208]

[0.59906553] [0.49667143] [0.73204294] [0.77249547] [0.48046345]]
1. print(trainY[0])
[0.69241117]
1. print(trainX.shape)
(990, 10, 1)
Import
1. from tensorflow.keras.models import Sequential
2. from tensorflow.keras.layers import Dense, SimpleRNN
Model Creation
1. model = Sequential()
2. model.add(SimpleRNN(units=1, activation="tanh"))
3. model.add(Dense(1))
4. model.compile(loss='mean_squared_error', optimizer='rmsprop')
1. history = model.fit(trainX,trainY, epochs=500, batch_size=16, verbose=2)
Train on 990 samples Epoch 1/500 990/990 - 0s - loss: 0.1776 Epoch 2/500
990/990 - 0s - loss: 0.1114 Epoch 3/500 990/990 - 0s - loss: 0.0705 Epoch 4/500
990/990 - 0s - loss: 0.0287 Epoch 99/500 990/990 - 0s - loss: 0.0287 Epoch
100/500 990/990 - 0s - loss: 0.0286 Epoch 101/500 990/990 - 0s - loss: 0.0284
Epoch 102/500 990/990 - 0s - loss: 0.0284 Epoch 103/500 990/990 - 0s - loss:
0.0283 Epoch 104/500 990/990 - 0s - loss: 0.0283 Epoch 105/500 990/990 - 0s -
loss: 0.0281 Epoch 106/500 990/990 - 0s - loss: 0.0281 Epoch 107/500 990/990 -
0s - loss: 0.0280 Epoch 108/500 990/990 - 0s - loss: 0.0279 Epoch 109/500
0.0241 Epoch 500/500 990/990 - 0s - loss: 0.0241
1. import matplotlib.pyplot as plt
2.
3. loss = history.history['loss']
4.
5. plt.plot(loss)
Prediction
1. trainPredict = normalizer.inverse_transform(model.predict(trainX))
2. testPredict= normalizer.inverse_transform(model.predict(testX))
3. predicted= np.concatenate((trainPredict,testPredict))
4. x = normalizer.inverse_transform(x)
Final Plot
1. plt.plot(x)
2. plt.plot(predicted)
3. # plt.axvline(len(trainX), c="r")

Import As Import As

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Import As Import As

Uploaded by

Copyright:

Available Formats

1.

[[0.53046776] [0.61911792] [0.56562058] ... [0.19383459] [0.1547965

[[0.53046776] [0.61911792] [0.56562058] [0.5399316 ] [0.48194208]

1. history = model.fit(trainX,trainY, epochs=500, batch_size=16, verbose=2)

You might also like