Importing all the libraries and the dataset of the tesla

In [1]:
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import numpy as np
In [2]:
df = pd.read_csv('TSLA.csv')
df.head()
Out[2]:
Date Open High Low Close Adj Close Volume
0 2019-12-26 85.582001 86.695999 85.269997 86.188004 86.188004 53169500
1 2019-12-27 87.000000 87.061996 85.222000 86.075996 86.075996 49728500
2 2019-12-30 85.758003 85.800003 81.851997 82.940002 82.940002 62932000
3 2019-12-31 81.000000 84.258003 80.416000 83.666000 83.666000 51428500
4 2020-01-02 84.900002 86.139999 84.342003 86.052002 86.052002 47660500

Changing the index to date column to visulaise the line chart

In [3]:
df['Date'] = pd.to_datetime(df['Date'])
df.set_index('Date',inplace=True)
In [4]:
df.head()
Out[4]:
Open High Low Close Adj Close Volume
Date
2019-12-26 85.582001 86.695999 85.269997 86.188004 86.188004 53169500
2019-12-27 87.000000 87.061996 85.222000 86.075996 86.075996 49728500
2019-12-30 85.758003 85.800003 81.851997 82.940002 82.940002 62932000
2019-12-31 81.000000 84.258003 80.416000 83.666000 83.666000 51428500
2020-01-02 84.900002 86.139999 84.342003 86.052002 86.052002 47660500

According to the exploratory data analysis, the data has a shape of (504, 6), meaning that the dataset contains 506 rows and 6 columns. Furthermore, all of the columns have a float data type, except for the column "volume," it has an integer data type. Furthermore, in the last two years, the highest stock value is 1229.92 and the lowest is 72.2.

In [5]:
def EDA(x):
  print(x.shape)
  print('*'*50)
  print(x.info())
  print('*'*50)
  print(x.describe())

EDA(df)

(504, 6)
**************************************************
<class 'pandas.core.frame.DataFrame'>
DatetimeIndex: 504 entries, 2019-12-26 to 2021-12-23
Data columns (total 6 columns):
 #   Column     Non-Null Count  Dtype  
---  ------     --------------  -----  
 0   Open       504 non-null    float64
 1   High       504 non-null    float64
 2   Low        504 non-null    float64
 3   Close      504 non-null    float64
 4   Adj Close  504 non-null    float64
 5   Volume     504 non-null    int64  
dtypes: float64(5), int64(1)
memory usage: 27.6 KB
None
**************************************************
              Open         High  ...    Adj Close        Volume
count   504.000000   504.000000  ...   504.000000  5.040000e+02
mean    524.973616   536.834630  ...   525.537733  5.174611e+07
std     293.116894   298.880181  ...   293.459521  3.802533e+07
min      74.940002    80.972000  ...    72.244003  9.800600e+06
25%     201.125004   203.827999  ...   201.600498  2.436525e+07
50%     601.644989   613.274994  ...   599.204986  3.904170e+07
75%     718.040024   731.750000  ...   718.569992  7.081100e+07
max    1234.410034  1243.489990  ...  1229.910034  3.046940e+08

[8 rows x 6 columns]

There are 0 null values and 0 duplicate values. Even though the values are duplicated, We cannot eliminate them since the stock value can be the same on a future day.

In [6]:
def missing_values(x):
    print(x.isnull().sum())
    print(x.duplicated().sum())

missing_values(df)

Open         0
High         0
Low          0
Close        0
Adj Close    0
Volume       0
dtype: int64
0

Visualizing the column "Close" since the column "close" would be the final price of the stock price for everyday. Additionally, the graph meantion that the stock value has a linear treand from the month January 2020 to febrauary 2021 in between it has small noise and dipped down in the month of march and had large noise had a linear trend till October 2021 and raised in November in the year 2021.

In [7]:
from matplotlib.pyplot import figure
plt.figure(figsize=(16,8))
plt.title('Close Price History')
plt.plot(df['Close'], color='red')
plt.xlabel('Date', fontsize=18)
plt.ylabel('Close Price USD', fontsize = 18)
plt.show()

The heatmap tells that every column has an autocorelation since every value is nearby 1.

In [8]:
sns.heatmap(df.corr(), annot = True)
Out[8]:
<matplotlib.axes._subplots.AxesSubplot at 0x7f382759a150>
In [9]:
df1=df.reset_index()['Close']
df1.head()
Out[9]:
0    86.188004
1    86.075996
2    82.940002
3    83.666000
4    86.052002
Name: Close, dtype: float64
In [10]:
df1.hist()
Out[10]:
<matplotlib.axes._subplots.AxesSubplot at 0x7f38274f8f10>

LSTM is very much sensitive to the scale of the data. These particular dataset values are on a different scale. Hence, we are using minmaxscaler to scale the data in the range between 0 and 1

In [11]:
from sklearn.preprocessing import MinMaxScaler
scaler=MinMaxScaler(feature_range=(0,1))
df1=scaler.fit_transform(np.array(df1).reshape(-1,1))

df1.shape
Out[11]:
(504, 1)

In Time-Series analysis it is very important to split the data since the data is specified in the date range. If we split the data with cross validation or random seed method, the training set would not have the sequence of the data and it would be impossible for us to predict the future. Additionally, every data of time series is dependent on the previous data. Hence, we need to split it by length of the data.

In [12]:
##splitting dataset into train and test split
training_size = int(len(df1)*0.70)
test_size = len(df1)- training_size
train_data,test_data = df1[0:training_size,:],df1[training_size:len(df1),:1]
print(train_data.shape)
print(test_data.shape)

(352, 1)
(152, 1)

The above code gives me the perfect split of the initial 70% of the data as a training set and the below 30% of the data as a testing set. Now my shape of the training data is (352,1) and test data (152,1)

Data Pre-processing: In time-series, if we need to compute the next day data, we need to consider the n- number of previous days that need to be written as a time step. In our data set, I used time step = 10, which means I can find the pattern for the first 10 data and predict the 11th one. Secondly, the algorithm finds the patterns for the data 2 – 11 and predict the 12th one and go on.

In [13]:
# convert an array of values into a dataset matrix
def create_dataset(dataset, time_step=1):
	dataX, dataY = [], []
	for i in range(len(dataset)-time_step-1):
		a = dataset[i:(i+time_step), 0]   ###i=0, 0,1,2,3-----99   100 
		dataX.append(a)
		dataY.append(dataset[i + time_step, 0])
	return np.array(dataX), np.array(dataY)
In [14]:
# reshape into X=t,t+1,t+2,t+3 and Y=t+4
time_step = 10
X_train, y_train = create_dataset(train_data, time_step)
X_test, ytest = create_dataset(test_data, time_step)

Additionally, we are adding the data into X_train, y_train, X_test and y_test with the method of the time step. Thus, the first 10 data would go to the X_train and the 11th would go to the y_train. This strategy is similar for testing datasets too.

Reshaping: Before implementing LSTM, we need to reshape all the values to 3 dimensional and the reshape input be the samples, time steps and features. We need to reshape the data because we give the LSTM input as time steps and features. Post reshaping my shape of the data as become X_train (341, 10, 1), X_test (141, 10, 1)

In [15]:
print(X_train.shape), print(y_train.shape)
print('*'*50)
print(X_test.shape), print(ytest.shape)

(341, 10)
(341,)
**************************************************
(141, 10)
(141,)
Out[15]:
(None, None)
In [16]:
# reshape input to be [samples, time steps, features] which is required for LSTM
X_train =X_train.reshape(X_train.shape[0],X_train.shape[1] , 1)
X_test = X_test.reshape(X_test.shape[0],X_test.shape[1] , 1)

print(X_train.shape,X_test.shape)

(341, 10, 1) (141, 10, 1)

Model Building: The LSTM is the Long short term memory algorithm of recurrent neural network used in deep learning and it has feedback connections. LSTM consists of a cell, input gate and a forget gate. The three gates manage the flow of information into and out of the cell, and the cell remembers values across unlimited time intervals.

In [17]:
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense
from tensorflow.keras.layers import LSTM

To create a model on LSTM, we need to import Sequential, Dense and LSTM from the library tensorflow. We created the Sequential model with the LSTM neural network, with the 1st layer contains 50 hidden layers, and the 1st input shape should be your (X_train.shape[1],1) value which is (10,1) as per my model. Since it is a stacked LSTM model we used the 2nd and 3rd layers with 50 hidden layers. Finally, we have added the “Dense” layer which is the output. Finally, I have compiled the value with Mean squared error and optimizer I used is Adam.

In [18]:
model=Sequential()
model.add(LSTM(50,return_sequences=True,input_shape=(10,1)))
model.add(LSTM(50,return_sequences=True))
model.add(LSTM(50))
model.add(Dense(1))
model.compile(loss='mean_squared_error',optimizer='adam')
In [19]:
model.summary()

Model: "sequential"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 lstm (LSTM)                 (None, 10, 50)            10400     
                                                                 
 lstm_1 (LSTM)               (None, 10, 50)            20200     
                                                                 
 lstm_2 (LSTM)               (None, 50)                20200     
                                                                 
 dense (Dense)               (None, 1)                 51        
                                                                 
=================================================================
Total params: 50,851
Trainable params: 50,851
Non-trainable params: 0
_________________________________________________________________

The next step is to fit the model into the training data and considering my epochs =200. However, my 109th epoch gave me the minimal loss with the good forecasting LSTM model. The lower the rate of loss, the model works better.

In [20]:
model.fit(X_train,y_train,validation_data=(X_test,ytest),epochs=120,batch_size=64,verbose=1)

Epoch 1/120
6/6 [==============================] - 7s 293ms/step - loss: 0.0772 - val_loss: 0.0717
Epoch 2/120
6/6 [==============================] - 0s 33ms/step - loss: 0.0189 - val_loss: 0.0066
Epoch 3/120
6/6 [==============================] - 0s 32ms/step - loss: 0.0111 - val_loss: 0.0411
Epoch 4/120
6/6 [==============================] - 0s 32ms/step - loss: 0.0107 - val_loss: 0.0278
Epoch 5/120
6/6 [==============================] - 0s 30ms/step - loss: 0.0041 - val_loss: 0.0053
Epoch 6/120
6/6 [==============================] - 0s 31ms/step - loss: 0.0043 - val_loss: 0.0042
Epoch 7/120
6/6 [==============================] - 0s 36ms/step - loss: 0.0027 - val_loss: 0.0072
Epoch 8/120
6/6 [==============================] - 0s 33ms/step - loss: 0.0021 - val_loss: 0.0053
Epoch 9/120
6/6 [==============================] - 0s 34ms/step - loss: 0.0021 - val_loss: 0.0044
Epoch 10/120
6/6 [==============================] - 0s 34ms/step - loss: 0.0017 - val_loss: 0.0038
Epoch 11/120
6/6 [==============================] - 0s 34ms/step - loss: 0.0016 - val_loss: 0.0037
Epoch 12/120
6/6 [==============================] - 0s 29ms/step - loss: 0.0016 - val_loss: 0.0043
Epoch 13/120
6/6 [==============================] - 0s 29ms/step - loss: 0.0016 - val_loss: 0.0039
Epoch 14/120
6/6 [==============================] - 0s 31ms/step - loss: 0.0016 - val_loss: 0.0036
Epoch 15/120
6/6 [==============================] - 0s 30ms/step - loss: 0.0016 - val_loss: 0.0040
Epoch 16/120
6/6 [==============================] - 0s 33ms/step - loss: 0.0016 - val_loss: 0.0038
Epoch 17/120
6/6 [==============================] - 0s 32ms/step - loss: 0.0016 - val_loss: 0.0038
Epoch 18/120
6/6 [==============================] - 0s 37ms/step - loss: 0.0016 - val_loss: 0.0043
Epoch 19/120
6/6 [==============================] - 0s 32ms/step - loss: 0.0016 - val_loss: 0.0039
Epoch 20/120
6/6 [==============================] - 0s 30ms/step - loss: 0.0015 - val_loss: 0.0046
Epoch 21/120
6/6 [==============================] - 0s 32ms/step - loss: 0.0016 - val_loss: 0.0040
Epoch 22/120
6/6 [==============================] - 0s 31ms/step - loss: 0.0016 - val_loss: 0.0040
Epoch 23/120
6/6 [==============================] - 0s 35ms/step - loss: 0.0016 - val_loss: 0.0042
Epoch 24/120
6/6 [==============================] - 0s 35ms/step - loss: 0.0017 - val_loss: 0.0041
Epoch 25/120
6/6 [==============================] - 0s 33ms/step - loss: 0.0015 - val_loss: 0.0055
Epoch 26/120
6/6 [==============================] - 0s 32ms/step - loss: 0.0015 - val_loss: 0.0038
Epoch 27/120
6/6 [==============================] - 0s 29ms/step - loss: 0.0015 - val_loss: 0.0051
Epoch 28/120
6/6 [==============================] - 0s 35ms/step - loss: 0.0015 - val_loss: 0.0042
Epoch 29/120
6/6 [==============================] - 0s 33ms/step - loss: 0.0015 - val_loss: 0.0041
Epoch 30/120
6/6 [==============================] - 0s 32ms/step - loss: 0.0015 - val_loss: 0.0041
Epoch 31/120
6/6 [==============================] - 0s 29ms/step - loss: 0.0015 - val_loss: 0.0044
Epoch 32/120
6/6 [==============================] - 0s 30ms/step - loss: 0.0015 - val_loss: 0.0046
Epoch 33/120
6/6 [==============================] - 0s 32ms/step - loss: 0.0015 - val_loss: 0.0045
Epoch 34/120
6/6 [==============================] - 0s 33ms/step - loss: 0.0015 - val_loss: 0.0042
Epoch 35/120
6/6 [==============================] - 0s 31ms/step - loss: 0.0015 - val_loss: 0.0048
Epoch 36/120
6/6 [==============================] - 0s 32ms/step - loss: 0.0015 - val_loss: 0.0045
Epoch 37/120
6/6 [==============================] - 0s 28ms/step - loss: 0.0015 - val_loss: 0.0048
Epoch 38/120
6/6 [==============================] - 0s 29ms/step - loss: 0.0015 - val_loss: 0.0044
Epoch 39/120
6/6 [==============================] - 0s 34ms/step - loss: 0.0015 - val_loss: 0.0050
Epoch 40/120
6/6 [==============================] - 0s 29ms/step - loss: 0.0014 - val_loss: 0.0045
Epoch 41/120
6/6 [==============================] - 0s 30ms/step - loss: 0.0015 - val_loss: 0.0050
Epoch 42/120
6/6 [==============================] - 0s 31ms/step - loss: 0.0015 - val_loss: 0.0047
Epoch 43/120
6/6 [==============================] - 0s 29ms/step - loss: 0.0015 - val_loss: 0.0049
Epoch 44/120
6/6 [==============================] - 0s 29ms/step - loss: 0.0014 - val_loss: 0.0046
Epoch 45/120
6/6 [==============================] - 0s 32ms/step - loss: 0.0015 - val_loss: 0.0053
Epoch 46/120
6/6 [==============================] - 0s 31ms/step - loss: 0.0015 - val_loss: 0.0046
Epoch 47/120
6/6 [==============================] - 0s 31ms/step - loss: 0.0014 - val_loss: 0.0050
Epoch 48/120
6/6 [==============================] - 0s 31ms/step - loss: 0.0014 - val_loss: 0.0046
Epoch 49/120
6/6 [==============================] - 0s 32ms/step - loss: 0.0014 - val_loss: 0.0052
Epoch 50/120
6/6 [==============================] - 0s 35ms/step - loss: 0.0014 - val_loss: 0.0045
Epoch 51/120
6/6 [==============================] - 0s 32ms/step - loss: 0.0014 - val_loss: 0.0047
Epoch 52/120
6/6 [==============================] - 0s 30ms/step - loss: 0.0014 - val_loss: 0.0052
Epoch 53/120
6/6 [==============================] - 0s 30ms/step - loss: 0.0015 - val_loss: 0.0043
Epoch 54/120
6/6 [==============================] - 0s 30ms/step - loss: 0.0015 - val_loss: 0.0057
Epoch 55/120
6/6 [==============================] - 0s 33ms/step - loss: 0.0014 - val_loss: 0.0042
Epoch 56/120
6/6 [==============================] - 0s 29ms/step - loss: 0.0014 - val_loss: 0.0048
Epoch 57/120
6/6 [==============================] - 0s 31ms/step - loss: 0.0014 - val_loss: 0.0050
Epoch 58/120
6/6 [==============================] - 0s 30ms/step - loss: 0.0014 - val_loss: 0.0064
Epoch 59/120
6/6 [==============================] - 0s 34ms/step - loss: 0.0015 - val_loss: 0.0039
Epoch 60/120
6/6 [==============================] - 0s 31ms/step - loss: 0.0015 - val_loss: 0.0078
Epoch 61/120
6/6 [==============================] - 0s 34ms/step - loss: 0.0017 - val_loss: 0.0037
Epoch 62/120
6/6 [==============================] - 0s 33ms/step - loss: 0.0015 - val_loss: 0.0066
Epoch 63/120
6/6 [==============================] - 0s 33ms/step - loss: 0.0014 - val_loss: 0.0037
Epoch 64/120
6/6 [==============================] - 0s 33ms/step - loss: 0.0014 - val_loss: 0.0058
Epoch 65/120
6/6 [==============================] - 0s 33ms/step - loss: 0.0014 - val_loss: 0.0042
Epoch 66/120
6/6 [==============================] - 0s 33ms/step - loss: 0.0013 - val_loss: 0.0061
Epoch 67/120
6/6 [==============================] - 0s 33ms/step - loss: 0.0014 - val_loss: 0.0039
Epoch 68/120
6/6 [==============================] - 0s 33ms/step - loss: 0.0014 - val_loss: 0.0067
Epoch 69/120
6/6 [==============================] - 0s 30ms/step - loss: 0.0016 - val_loss: 0.0036
Epoch 70/120
6/6 [==============================] - 0s 36ms/step - loss: 0.0015 - val_loss: 0.0092
Epoch 71/120
6/6 [==============================] - 0s 37ms/step - loss: 0.0016 - val_loss: 0.0034
Epoch 72/120
6/6 [==============================] - 0s 35ms/step - loss: 0.0016 - val_loss: 0.0074
Epoch 73/120
6/6 [==============================] - 0s 32ms/step - loss: 0.0014 - val_loss: 0.0036
Epoch 74/120
6/6 [==============================] - 0s 33ms/step - loss: 0.0015 - val_loss: 0.0057
Epoch 75/120
6/6 [==============================] - 0s 35ms/step - loss: 0.0014 - val_loss: 0.0042
Epoch 76/120
6/6 [==============================] - 0s 36ms/step - loss: 0.0013 - val_loss: 0.0057
Epoch 77/120
6/6 [==============================] - 0s 33ms/step - loss: 0.0013 - val_loss: 0.0037
Epoch 78/120
6/6 [==============================] - 0s 32ms/step - loss: 0.0014 - val_loss: 0.0063
Epoch 79/120
6/6 [==============================] - 0s 30ms/step - loss: 0.0014 - val_loss: 0.0037
Epoch 80/120
6/6 [==============================] - 0s 33ms/step - loss: 0.0013 - val_loss: 0.0059
Epoch 81/120
6/6 [==============================] - 0s 32ms/step - loss: 0.0014 - val_loss: 0.0039
Epoch 82/120
6/6 [==============================] - 0s 30ms/step - loss: 0.0013 - val_loss: 0.0053
Epoch 83/120
6/6 [==============================] - 0s 32ms/step - loss: 0.0013 - val_loss: 0.0042
Epoch 84/120
6/6 [==============================] - 0s 29ms/step - loss: 0.0013 - val_loss: 0.0047
Epoch 85/120
6/6 [==============================] - 0s 31ms/step - loss: 0.0013 - val_loss: 0.0045
Epoch 86/120
6/6 [==============================] - 0s 33ms/step - loss: 0.0013 - val_loss: 0.0050
Epoch 87/120
6/6 [==============================] - 0s 34ms/step - loss: 0.0013 - val_loss: 0.0041
Epoch 88/120
6/6 [==============================] - 0s 30ms/step - loss: 0.0013 - val_loss: 0.0047
Epoch 89/120
6/6 [==============================] - 0s 31ms/step - loss: 0.0013 - val_loss: 0.0049
Epoch 90/120
6/6 [==============================] - 0s 35ms/step - loss: 0.0013 - val_loss: 0.0035
Epoch 91/120
6/6 [==============================] - 0s 32ms/step - loss: 0.0013 - val_loss: 0.0050
Epoch 92/120
6/6 [==============================] - 0s 32ms/step - loss: 0.0013 - val_loss: 0.0053
Epoch 93/120
6/6 [==============================] - 0s 32ms/step - loss: 0.0015 - val_loss: 0.0031
Epoch 94/120
6/6 [==============================] - 0s 30ms/step - loss: 0.0014 - val_loss: 0.0057
Epoch 95/120
6/6 [==============================] - 0s 33ms/step - loss: 0.0013 - val_loss: 0.0052
Epoch 96/120
6/6 [==============================] - 0s 30ms/step - loss: 0.0013 - val_loss: 0.0031
Epoch 97/120
6/6 [==============================] - 0s 30ms/step - loss: 0.0013 - val_loss: 0.0068
Epoch 98/120
6/6 [==============================] - 0s 30ms/step - loss: 0.0013 - val_loss: 0.0033
Epoch 99/120
6/6 [==============================] - 0s 32ms/step - loss: 0.0013 - val_loss: 0.0048
Epoch 100/120
6/6 [==============================] - 0s 33ms/step - loss: 0.0012 - val_loss: 0.0040
Epoch 101/120
6/6 [==============================] - 0s 30ms/step - loss: 0.0012 - val_loss: 0.0045
Epoch 102/120
6/6 [==============================] - 0s 34ms/step - loss: 0.0012 - val_loss: 0.0039
Epoch 103/120
6/6 [==============================] - 0s 34ms/step - loss: 0.0012 - val_loss: 0.0065
Epoch 104/120
6/6 [==============================] - 0s 34ms/step - loss: 0.0013 - val_loss: 0.0036
Epoch 105/120
6/6 [==============================] - 0s 32ms/step - loss: 0.0013 - val_loss: 0.0028
Epoch 106/120
6/6 [==============================] - 0s 30ms/step - loss: 0.0013 - val_loss: 0.0061
Epoch 107/120
6/6 [==============================] - 0s 33ms/step - loss: 0.0013 - val_loss: 0.0034
Epoch 108/120
6/6 [==============================] - 0s 33ms/step - loss: 0.0011 - val_loss: 0.0048
Epoch 109/120
6/6 [==============================] - 0s 29ms/step - loss: 0.0012 - val_loss: 0.0031
Epoch 110/120
6/6 [==============================] - 0s 30ms/step - loss: 0.0012 - val_loss: 0.0028
Epoch 111/120
6/6 [==============================] - 0s 32ms/step - loss: 0.0013 - val_loss: 0.0061
Epoch 112/120
6/6 [==============================] - 0s 30ms/step - loss: 0.0011 - val_loss: 0.0030
Epoch 113/120
6/6 [==============================] - 0s 38ms/step - loss: 0.0011 - val_loss: 0.0037
Epoch 114/120
6/6 [==============================] - 0s 33ms/step - loss: 0.0012 - val_loss: 0.0055
Epoch 115/120
6/6 [==============================] - 0s 29ms/step - loss: 0.0011 - val_loss: 0.0035
Epoch 116/120
6/6 [==============================] - 0s 31ms/step - loss: 0.0011 - val_loss: 0.0031
Epoch 117/120
6/6 [==============================] - 0s 30ms/step - loss: 0.0011 - val_loss: 0.0045
Epoch 118/120
6/6 [==============================] - 0s 35ms/step - loss: 0.0011 - val_loss: 0.0051
Epoch 119/120
6/6 [==============================] - 0s 33ms/step - loss: 0.0011 - val_loss: 0.0030
Epoch 120/120
6/6 [==============================] - 0s 29ms/step - loss: 0.0011 - val_loss: 0.0034
Out[20]:
<keras.callbacks.History at 0x7f37b1b228d0>

Post finding the loss, we need to predict the model with X_train and X_test data and check how well the model has worked for this dataset.

In [21]:
# prediction and check performance metrics
train_predict=model.predict(X_train)
test_predict=model.predict(X_test)
In [22]:
##Transformback to original form
train_predict=scaler.inverse_transform(train_predict)
test_predict=scaler.inverse_transform(test_predict)
In [23]:
import math
from sklearn.metrics import mean_squared_error
math.sqrt(mean_squared_error(y_train,train_predict))
Out[23]:
480.43697504178385
In [24]:
### Test Data RMSE
math.sqrt(mean_squared_error(ytest,test_predict))
Out[24]:
795.6736758628309
In [25]:
### Plotting 
# shift train predictions for plotting
look_back=10
trainPredictPlot = np.empty_like(df1)
trainPredictPlot[:, :] = np.nan
trainPredictPlot[look_back:len(train_predict)+look_back, :] = train_predict
# shift test predictions for plotting
testPredictPlot = np.empty_like(df1)
testPredictPlot[:, :] = np.nan
testPredictPlot[len(train_predict)+(look_back*2)+1:len(df1)-1, :] = test_predict
# plot baseline and predictions
plt.plot(scaler.inverse_transform(df1))
plt.plot(trainPredictPlot)
plt.plot(testPredictPlot)
plt.show()
In [26]:
len(test_data)
Out[26]:
152
In [27]:
x_input = test_data[142:].reshape(1,-1)
In [28]:
x_input.shape
Out[28]:
(1, 10)
In [29]:
temp_input = list(x_input)
temp_input = temp_input[0].tolist()
temp_input
Out[29]:
[0.816112765426733,
 0.7723868076422811,
 0.7655627644480828,
 0.7806620932112329,
 0.7382750785748848,
 0.7431556087525902,
 0.7149695826222269,
 0.748303917366994,
 0.8090640710870093,
 0.8592771752495172]
In [30]:
from numpy import array

lst_output=[]
n_steps=10
i=0
while(i<30):
    
    if(len(temp_input)>10):
        #print(temp_input)
        x_input=np.array(temp_input[1:])
        print("{} day input {}".format(i,x_input))
        x_input=x_input.reshape(1,-1)
        x_input = x_input.reshape((1, n_steps, 1))
        #print(x_input)
        yhat = model.predict(x_input, verbose=0)
        print("{} day output {}".format(i,yhat))
        temp_input.extend(yhat[0].tolist())
        temp_input=temp_input[1:]
        #print(temp_input)
        lst_output.extend(yhat.tolist())
        i=i+1
    else:
        x_input = x_input.reshape((1, n_steps,1))
        yhat = model.predict(x_input, verbose=0)
        print(yhat[0])
        temp_input.extend(yhat[0].tolist())
        print(len(temp_input))
        lst_output.extend(yhat.tolist())
        i=i+1
    

print(lst_output)

[0.71900856]
11
1 day input [0.77238681 0.76556276 0.78066209 0.73827508 0.74315561 0.71496958
 0.74830392 0.80906407 0.85927718 0.71900856]
1 day output [[0.72514564]]
2 day input [0.76556276 0.78066209 0.73827508 0.74315561 0.71496958 0.74830392
 0.80906407 0.85927718 0.71900856 0.72514564]
2 day output [[0.72663534]]
3 day input [0.78066209 0.73827508 0.74315561 0.71496958 0.74830392 0.80906407
 0.85927718 0.71900856 0.72514564 0.72663534]
3 day output [[0.7243009]]
4 day input [0.73827508 0.74315561 0.71496958 0.74830392 0.80906407 0.85927718
 0.71900856 0.72514564 0.72663534 0.72430092]
4 day output [[0.71983224]]
5 day input [0.74315561 0.71496958 0.74830392 0.80906407 0.85927718 0.71900856
 0.72514564 0.72663534 0.72430092 0.71983224]
5 day output [[0.71244115]]
6 day input [0.71496958 0.74830392 0.80906407 0.85927718 0.71900856 0.72514564
 0.72663534 0.72430092 0.71983224 0.71244115]
6 day output [[0.7043093]]
7 day input [0.74830392 0.80906407 0.85927718 0.71900856 0.72514564 0.72663534
 0.72430092 0.71983224 0.71244115 0.70430928]
7 day output [[0.6949978]]
8 day input [0.80906407 0.85927718 0.71900856 0.72514564 0.72663534 0.72430092
 0.71983224 0.71244115 0.70430928 0.69499779]
8 day output [[0.68691635]]
9 day input [0.85927718 0.71900856 0.72514564 0.72663534 0.72430092 0.71983224
 0.71244115 0.70430928 0.69499779 0.68691635]
9 day output [[0.6816989]]
10 day input [0.71900856 0.72514564 0.72663534 0.72430092 0.71983224 0.71244115
 0.70430928 0.69499779 0.68691635 0.68169892]
10 day output [[0.67947286]]
11 day input [0.72514564 0.72663534 0.72430092 0.71983224 0.71244115 0.70430928
 0.69499779 0.68691635 0.68169892 0.67947286]
11 day output [[0.67415065]]
12 day input [0.72663534 0.72430092 0.71983224 0.71244115 0.70430928 0.69499779
 0.68691635 0.68169892 0.67947286 0.67415065]
12 day output [[0.6690942]]
13 day input [0.72430092 0.71983224 0.71244115 0.70430928 0.69499779 0.68691635
 0.68169892 0.67947286 0.67415065 0.6690942 ]
13 day output [[0.6644602]]
14 day input [0.71983224 0.71244115 0.70430928 0.69499779 0.68691635 0.68169892
 0.67947286 0.67415065 0.6690942  0.66446018]
14 day output [[0.6602583]]
15 day input [0.71244115 0.70430928 0.69499779 0.68691635 0.68169892 0.67947286
 0.67415065 0.6690942  0.66446018 0.66025829]
15 day output [[0.6564493]]
16 day input [0.70430928 0.69499779 0.68691635 0.68169892 0.67947286 0.67415065
 0.6690942  0.66446018 0.66025829 0.65644932]
16 day output [[0.6529236]]
17 day input [0.69499779 0.68691635 0.68169892 0.67947286 0.67415065 0.6690942
 0.66446018 0.66025829 0.65644932 0.65292358]
17 day output [[0.6496055]]
18 day input [0.68691635 0.68169892 0.67947286 0.67415065 0.6690942  0.66446018
 0.66025829 0.65644932 0.65292358 0.64960551]
18 day output [[0.64640594]]
19 day input [0.68169892 0.67947286 0.67415065 0.6690942  0.66446018 0.66025829
 0.65644932 0.65292358 0.64960551 0.64640594]
19 day output [[0.64330137]]
20 day input [0.67947286 0.67415065 0.6690942  0.66446018 0.66025829 0.65644932
 0.65292358 0.64960551 0.64640594 0.64330137]
20 day output [[0.6403373]]
21 day input [0.67415065 0.6690942  0.66446018 0.66025829 0.65644932 0.65292358
 0.64960551 0.64640594 0.64330137 0.64033729]
21 day output [[0.6375935]]
22 day input [0.6690942  0.66446018 0.66025829 0.65644932 0.65292358 0.64960551
 0.64640594 0.64330137 0.64033729 0.63759351]
22 day output [[0.6349983]]
23 day input [0.66446018 0.66025829 0.65644932 0.65292358 0.64960551 0.64640594
 0.64330137 0.64033729 0.63759351 0.63499832]
23 day output [[0.6325304]]
24 day input [0.66025829 0.65644932 0.65292358 0.64960551 0.64640594 0.64330137
 0.64033729 0.63759351 0.63499832 0.63253039]
24 day output [[0.6301773]]
25 day input [0.65644932 0.65292358 0.64960551 0.64640594 0.64330137 0.64033729
 0.63759351 0.63499832 0.63253039 0.63017732]
25 day output [[0.6279327]]
26 day input [0.65292358 0.64960551 0.64640594 0.64330137 0.64033729 0.63759351
 0.63499832 0.63253039 0.63017732 0.62793273]
26 day output [[0.6257938]]
27 day input [0.64960551 0.64640594 0.64330137 0.64033729 0.63759351 0.63499832
 0.63253039 0.63017732 0.62793273 0.62579381]
27 day output [[0.6237567]]
28 day input [0.64640594 0.64330137 0.64033729 0.63759351 0.63499832 0.63253039
 0.63017732 0.62793273 0.62579381 0.62375671]
28 day output [[0.62181747]]
29 day input [0.64330137 0.64033729 0.63759351 0.63499832 0.63253039 0.63017732
 0.62793273 0.62579381 0.62375671 0.62181747]
29 day output [[0.6199696]]
[[0.7190085649490356], [0.7251456379890442], [0.7266353368759155], [0.7243009209632874], [0.7198322415351868], [0.7124411463737488], [0.7043092846870422], [0.6949977874755859], [0.6869163513183594], [0.6816989183425903], [0.6794728636741638], [0.6741506457328796], [0.6690942049026489], [0.6644601821899414], [0.6602582931518555], [0.6564493179321289], [0.652923583984375], [0.6496055126190186], [0.6464059352874756], [0.6433013677597046], [0.6403372883796692], [0.6375935077667236], [0.6349983215332031], [0.6325303912162781], [0.6301773190498352], [0.6279327273368835], [0.6257938146591187], [0.6237567067146301], [0.6218174695968628], [0.6199696063995361]]
In [31]:
day_new=np.arange(1,11)
day_pred=np.arange(11,41)
In [32]:
len(df1)
Out[32]:
504
In [33]:
plt.plot(day_new,scaler.inverse_transform(df1[494:]))
plt.plot(day_pred,scaler.inverse_transform(lst_output))
Out[33]:
[<matplotlib.lines.Line2D at 0x7f37ae6c4b90>]
In [33]: