Predicting Closing Price BBRI using LSTM

Predicting Closing Price of BBRI's Stock

November 09, 2022 · 45 mins read

Predicting Closing Price of BBRI’s Stock

BBRI is the stock name of Bank Rakyat Indonesia. PT Bank Rakyat Indonesia (Persero) Tbk (People’s Bank of Indonesia, commonly known as BRI or Bank BRI) is one of the largest banks in Indonesia. It specialises in small scale and microfinance style borrowing from and lending to its approximately 30 million retail clients through its over 4,000 branches, units and rural service posts. It also has a comparatively small, but growing, corporate business. As of 2010 it is the second largest bank in Indonesia by asset.

This dataset can be found using Pandas DataReader and will pick from January 2008 to the end of May 2022. It will be more than 14 years. In this session, we will predict the closing price of BBRI with Long Short-Term Memory.

Long short-term memory (LSTM) is an artificial recurrent neural network (RNN) architecture used in the field of deep learning. Unlike standard feedforward neural networks, LSTM has feedback connections. It can not only process single data points (such as images), but also entire sequences of data (such as speech or video). For example, LSTM is applicable to tasks such as unsegmented, connected handwriting recognition, speech recognition and anomaly detection in network traffic or IDS’s (intrusion detection systems).

!pip install --upgrade pandas
!pip install --upgrade pandas-datareader

Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/
Requirement already satisfied: pandas in /usr/local/lib/python3.7/dist-packages (1.3.5)
Requirement already satisfied: pytz>=2017.3 in /usr/local/lib/python3.7/dist-packages (from pandas) (2022.5)
Requirement already satisfied: numpy>=1.17.3 in /usr/local/lib/python3.7/dist-packages (from pandas) (1.21.6)
Requirement already satisfied: python-dateutil>=2.7.3 in /usr/local/lib/python3.7/dist-packages (from pandas) (2.8.2)
Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.7/dist-packages (from python-dateutil>=2.7.3->pandas) (1.15.0)
Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/
Requirement already satisfied: pandas-datareader in /usr/local/lib/python3.7/dist-packages (0.9.0)
Collecting pandas-datareader
  Downloading pandas_datareader-0.10.0-py3-none-any.whl (109 kB)
     |████████████████████████████████| 109 kB 36.7 MB/s 
[?25hRequirement already satisfied: lxml in /usr/local/lib/python3.7/dist-packages (from pandas-datareader) (4.9.1)
Requirement already satisfied: pandas>=0.23 in /usr/local/lib/python3.7/dist-packages (from pandas-datareader) (1.3.5)
Requirement already satisfied: requests>=2.19.0 in /usr/local/lib/python3.7/dist-packages (from pandas-datareader) (2.23.0)
Requirement already satisfied: python-dateutil>=2.7.3 in /usr/local/lib/python3.7/dist-packages (from pandas>=0.23->pandas-datareader) (2.8.2)
Requirement already satisfied: pytz>=2017.3 in /usr/local/lib/python3.7/dist-packages (from pandas>=0.23->pandas-datareader) (2022.5)
Requirement already satisfied: numpy>=1.17.3 in /usr/local/lib/python3.7/dist-packages (from pandas>=0.23->pandas-datareader) (1.21.6)
Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.7/dist-packages (from python-dateutil>=2.7.3->pandas>=0.23->pandas-datareader) (1.15.0)
Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests>=2.19.0->pandas-datareader) (3.0.4)
Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests>=2.19.0->pandas-datareader) (2022.9.24)
Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests>=2.19.0->pandas-datareader) (2.10)
Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests>=2.19.0->pandas-datareader) (1.24.3)
Installing collected packages: pandas-datareader
  Attempting uninstall: pandas-datareader
    Found existing installation: pandas-datareader 0.9.0
    Uninstalling pandas-datareader-0.9.0:
      Successfully uninstalled pandas-datareader-0.9.0
Successfully installed pandas-datareader-0.10.0

# This Python 3 environment comes with many helpful analytics libraries installed
# It is defined by the kaggle/python Docker image: https://github.com/kaggle/docker-python
# For example, here's several helpful packages to load


import numpy as np # linear algebra
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)

# Input data files are available in the read-only "../input/" directory
# For example, running this (by clicking run or pressing Shift+Enter) will list all files under the input directory

import os
for dirname, _, filenames in os.walk('/kaggle/input'):
    for filename in filenames:
        print(os.path.join(dirname, filename))

# You can write up to 5GB to the current directory (/kaggle/working/) that gets preserved as output when you create a version using "Save & Run All" 
# You can also write temporary files to /kaggle/temp/, but they won't be saved outside of the current session

1) Data Preparation

Using Pandas DataReader, we can get dataset about the stock that we want to use. Of course we’re using some libraries to make it happen.

import numpy as np
import pandas as pd
import pandas_datareader as web
import matplotlib.pyplot as plt
plt.style.use('fivethirtyeight')

# Catching data BBRI
bbri = web.DataReader('bbri.jk', 'yahoo', '2008-01-01', '2022-05-31')
bbri
High Low Open Close Volume Adj Close
Date
2008-01-02 740.0 730.0 735.0 735.0 44765000.0 499.284210
2008-01-03 735.0 710.0 735.0 720.0 102940000.0 489.094513
2008-01-04 755.0 715.0 720.0 750.0 143670000.0 509.473419
2008-01-07 750.0 730.0 735.0 745.0 63700000.0 506.077026
2008-01-08 755.0 745.0 750.0 750.0 84225000.0 509.473419
... ... ... ... ... ... ...
2022-05-24 4500.0 4320.0 4320.0 4460.0 165758600.0 4460.000000
2022-05-25 4450.0 4350.0 4430.0 4350.0 138961700.0 4350.000000
2022-05-27 4540.0 4440.0 4500.0 4540.0 209962100.0 4540.000000
2022-05-30 4540.0 4390.0 4540.0 4430.0 193724500.0 4430.000000
2022-05-31 4630.0 4450.0 4470.0 4630.0 379626500.0 4630.000000

3556 rows × 6 columns

# Let's visualize the Closing Price

rolling_mean20 = bbri['Close'].rolling(window=20).mean()
rolling_mean50 = bbri['Close'].rolling(window=50).mean()
rolling_mean200 = bbri['Close'].rolling(window=200).mean()

plt.figure(figsize = (16,8))
plt.plot(bbri['Close'], color = 'b', label = 'BBRI Close')
plt.plot(rolling_mean20, color = 'r', linewidth = 2.5, label = 'MA20')
plt.plot(rolling_mean50, color = 'y', linewidth = 2.5, label = 'MA50')
plt.plot(rolling_mean200, color = 'c',linewidth = 2.5, label = 'MA200')

plt.xlabel('Date', fontsize = 15)
plt.ylabel('Closing Price in Rupiah (Rp)', fontsize = 18)
plt.title('Closing Price of BBRI')
plt.legend(loc = 'lower right')

<matplotlib.legend.Legend at 0x7f430d2c2210>

png

Here the visualization of BBRI’s closing price since 2008. Based on this graph, it’s been a good money when you invest your money into this stock. Just saying. I’m showing you the Moving Average (MA) too withing 20, 50 and 200 days to make it more technical with these indicators.

2) Data Pre-processing

Now let’s straight to the pre-processing. We have to prepare up like splitting our dataset into Train and Test dataset. We do this by call some libraries from sklearn.

# Don't forget to using 'values' attribute before apply it into Neural Network!

X = bbri.drop('Close', axis = 1).values
y = bbri['Close'].values

from sklearn.model_selection import train_test_split

# Split the Data into Train and Test dataset with 20% test size.
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=101)

X_train.shape

(2844, 5)

So, it’s already splitted with 80:20. Notice our X_train size is 2844 (80% of the dataset).

# Scale it with MinMaxScaler
from sklearn.preprocessing import MinMaxScaler

scaler = MinMaxScaler()

X_train = scaler.fit_transform(X_train)
X_test = scaler.fit_transform(X_test)

After scaling our X dataset, we should reshape our dataset into proper one. Especially on X dataset, we should turn it into 3-dimensional data before apply it to LSTM NN model or you’ll get some advices from the notebook.

y_train = y_train.reshape(-1,1)
y_test = y_test.reshape(-1,1)

# Reshape it into 3-dimensional before input it into LSTM Model.
# Reshape the data to be 3-dimensional in the form [number of samples, number of time steps, and number of features].

X_train = np.reshape(X_train, (X_train.shape[0], X_train.shape[1], 1))
X_test = np.reshape(X_test, (X_test.shape[0], X_test.shape[1], 1))

3) Make a LSTM Model

Long Short Term Memory networks – usually just called “LSTMs” – are a special kind of RNN, capable of learning long-term dependencies. They were introduced by Hochreiter & Schmidhuber (1997), and were refined and popularized by many people in following work.1 They work tremendously well on a large variety of problems, and are now widely used.

LSTMs are explicitly designed to avoid the long-term dependency problem. Remembering information for long periods of time is practically their default behavior, not something they struggle to learn!

All recurrent neural networks have the form of a chain of repeating modules of neural network. In standard RNNs, this repeating module will have a very simple structure, such as a single tanh layer.

from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, Dropout, LSTM

We use 5 neurons in input layer because we have 5 features along training our model. We add 1 hidden layer with 5 neurons and also ReLU activation. Finally adding 1 layer for output. We choose Adam optimizer with mean squared error for calculating model’s loss.

# just to remember our dataset shape

print(X_train.shape)
print(y_train.shape)
print(X_test.shape)
print(y_test.shape)

(2844, 5, 1)
(2844, 1)
(712, 5, 1)
(712, 1)

model = Sequential()

# input layer
model.add(LSTM(5, return_sequences=True, input_shape=(X_train.shape[1],1)))
model.add(LSTM(5, return_sequences= False))

# hidden layer
model.add(Dense(5, activation='relu'))
model.add(Dense(5, activation='relu'))

# output layer
model.add(Dense(1))

# compiler
model.compile(optimizer='adam', loss = 'mse')

I love efficiency, so I add Callbacks feature to the model with EarlyStopping from TensorFlow to avoid overfitting and improve generalization.

from tensorflow.keras.callbacks import EarlyStopping

early_stop = EarlyStopping(monitor = 'val_loss', mode = 'min', verbose = 1, patience = 25)

I set epochs to 500 iterations, but at the same time I do believe this model won’t iterate more than 500 cause EarlyStopping will save us.

# Let's train our model!

model.fit(x = X_train,
          y = y_train,
          validation_data = (X_test, y_test),
          epochs = 500, 
          callbacks = [early_stop])

Epoch 1/500
89/89 [==============================] - 7s 21ms/step - loss: 6728794.5000 - val_loss: 7068232.0000
Epoch 2/500
89/89 [==============================] - 1s 8ms/step - loss: 6723610.5000 - val_loss: 7057610.0000
Epoch 3/500
89/89 [==============================] - 1s 8ms/step - loss: 6707948.0000 - val_loss: 7036904.5000
Epoch 4/500
89/89 [==============================] - 1s 9ms/step - loss: 6682745.5000 - val_loss: 7005203.5000
Epoch 5/500
89/89 [==============================] - 1s 8ms/step - loss: 6644921.0000 - val_loss: 6959078.0000
Epoch 6/500
89/89 [==============================] - 1s 8ms/step - loss: 6591286.5000 - val_loss: 6895181.5000
Epoch 7/500
89/89 [==============================] - 1s 8ms/step - loss: 6519325.5000 - val_loss: 6811321.0000
Epoch 8/500
89/89 [==============================] - 1s 9ms/step - loss: 6426884.5000 - val_loss: 6705898.5000
Epoch 9/500
89/89 [==============================] - 1s 8ms/step - loss: 6312536.0000 - val_loss: 6577160.5000
Epoch 10/500
89/89 [==============================] - 1s 8ms/step - loss: 6175521.0000 - val_loss: 6425698.5000
Epoch 11/500
89/89 [==============================] - 1s 8ms/step - loss: 6015493.5000 - val_loss: 6250165.0000
Epoch 12/500
89/89 [==============================] - 1s 8ms/step - loss: 5833048.0000 - val_loss: 6052452.5000
Epoch 13/500
89/89 [==============================] - 1s 8ms/step - loss: 5629385.5000 - val_loss: 5832922.5000
Epoch 14/500
89/89 [==============================] - 1s 8ms/step - loss: 5405945.0000 - val_loss: 5595558.0000
Epoch 15/500
89/89 [==============================] - 1s 8ms/step - loss: 5165781.5000 - val_loss: 5340338.0000
Epoch 16/500
89/89 [==============================] - 1s 8ms/step - loss: 4910532.0000 - val_loss: 5073329.5000
Epoch 17/500
89/89 [==============================] - 1s 9ms/step - loss: 4644864.0000 - val_loss: 4794452.5000
Epoch 18/500
89/89 [==============================] - 1s 8ms/step - loss: 4371365.0000 - val_loss: 4511833.0000
Epoch 19/500
89/89 [==============================] - 1s 8ms/step - loss: 4094801.7500 - val_loss: 4226146.0000
Epoch 20/500
89/89 [==============================] - 1s 9ms/step - loss: 3819316.5000 - val_loss: 3943429.7500
Epoch 21/500
89/89 [==============================] - 1s 8ms/step - loss: 3548088.2500 - val_loss: 3669539.5000
Epoch 22/500
89/89 [==============================] - 1s 9ms/step - loss: 3286434.0000 - val_loss: 3403550.5000
Epoch 23/500
89/89 [==============================] - 1s 10ms/step - loss: 3037673.7500 - val_loss: 3152802.2500
Epoch 24/500
89/89 [==============================] - 1s 14ms/step - loss: 2804934.2500 - val_loss: 2922020.7500
Epoch 25/500
89/89 [==============================] - 1s 15ms/step - loss: 2591608.5000 - val_loss: 2708664.2500
Epoch 26/500
89/89 [==============================] - 1s 14ms/step - loss: 2399314.0000 - val_loss: 2518329.2500
Epoch 27/500
89/89 [==============================] - 2s 18ms/step - loss: 2230064.0000 - val_loss: 2351074.2500
Epoch 28/500
89/89 [==============================] - 2s 17ms/step - loss: 2084102.5000 - val_loss: 2208367.7500
Epoch 29/500
89/89 [==============================] - 1s 15ms/step - loss: 1961157.5000 - val_loss: 2089056.8750
Epoch 30/500
89/89 [==============================] - 2s 17ms/step - loss: 1860061.6250 - val_loss: 1990759.0000
Epoch 31/500
89/89 [==============================] - 1s 14ms/step - loss: 1779148.5000 - val_loss: 1912582.8750
Epoch 32/500
89/89 [==============================] - 1s 15ms/step - loss: 1716673.1250 - val_loss: 1850092.3750
Epoch 33/500
89/89 [==============================] - 1s 16ms/step - loss: 1669327.7500 - val_loss: 1804724.8750
Epoch 34/500
89/89 [==============================] - 1s 15ms/step - loss: 1635036.3750 - val_loss: 1770084.8750
Epoch 35/500
89/89 [==============================] - 1s 12ms/step - loss: 1610981.7500 - val_loss: 1746241.6250
Epoch 36/500
89/89 [==============================] - 1s 14ms/step - loss: 1594289.1250 - val_loss: 1729051.3750
Epoch 37/500
89/89 [==============================] - 1s 16ms/step - loss: 1274190.0000 - val_loss: 1078870.7500
Epoch 38/500
89/89 [==============================] - 1s 14ms/step - loss: 876666.5000 - val_loss: 917421.1875
Epoch 39/500
89/89 [==============================] - 1s 12ms/step - loss: 738829.5000 - val_loss: 777625.6875
Epoch 40/500
89/89 [==============================] - 2s 17ms/step - loss: 620443.3750 - val_loss: 656790.5625
Epoch 41/500
89/89 [==============================] - 1s 17ms/step - loss: 518916.4062 - val_loss: 551334.6250
Epoch 42/500
89/89 [==============================] - 2s 21ms/step - loss: 432518.0312 - val_loss: 461674.6562
Epoch 43/500
89/89 [==============================] - 2s 18ms/step - loss: 359889.3125 - val_loss: 385974.8438
Epoch 44/500
89/89 [==============================] - 2s 19ms/step - loss: 299184.0938 - val_loss: 321966.5938
Epoch 45/500
89/89 [==============================] - 2s 20ms/step - loss: 248667.4844 - val_loss: 268694.4375
Epoch 46/500
89/89 [==============================] - 2s 20ms/step - loss: 205980.6719 - val_loss: 223295.9062
Epoch 47/500
89/89 [==============================] - 1s 11ms/step - loss: 170021.3281 - val_loss: 185113.6094
Epoch 48/500
89/89 [==============================] - 1s 8ms/step - loss: 139950.8906 - val_loss: 152330.4375
Epoch 49/500
89/89 [==============================] - 1s 8ms/step - loss: 114830.4688 - val_loss: 125235.7500
Epoch 50/500
89/89 [==============================] - 1s 8ms/step - loss: 94134.4844 - val_loss: 102681.2031
Epoch 51/500
89/89 [==============================] - 1s 9ms/step - loss: 77076.0703 - val_loss: 84842.9688
Epoch 52/500
89/89 [==============================] - 1s 9ms/step - loss: 63257.7656 - val_loss: 69759.1719
Epoch 53/500
89/89 [==============================] - 1s 8ms/step - loss: 51887.4766 - val_loss: 57550.2148
Epoch 54/500
89/89 [==============================] - 1s 13ms/step - loss: 42710.1641 - val_loss: 47487.1289
Epoch 55/500
89/89 [==============================] - 1s 16ms/step - loss: 35261.4727 - val_loss: 39156.7930
Epoch 56/500
89/89 [==============================] - 1s 11ms/step - loss: 29141.7969 - val_loss: 32645.3984
Epoch 57/500
89/89 [==============================] - 1s 8ms/step - loss: 24204.2168 - val_loss: 27203.0586
Epoch 58/500
89/89 [==============================] - 1s 8ms/step - loss: 20151.5801 - val_loss: 22898.7109
Epoch 59/500
89/89 [==============================] - 1s 8ms/step - loss: 16899.8926 - val_loss: 19403.0078
Epoch 60/500
89/89 [==============================] - 1s 8ms/step - loss: 14241.0938 - val_loss: 16547.8203
Epoch 61/500
89/89 [==============================] - 1s 9ms/step - loss: 12103.2207 - val_loss: 14157.0586
Epoch 62/500
89/89 [==============================] - 1s 8ms/step - loss: 10390.1396 - val_loss: 12558.1826
Epoch 63/500
89/89 [==============================] - 1s 8ms/step - loss: 8988.3750 - val_loss: 10850.4727
Epoch 64/500
89/89 [==============================] - 1s 8ms/step - loss: 7777.8579 - val_loss: 9270.9590
Epoch 65/500
89/89 [==============================] - 1s 9ms/step - loss: 6764.6582 - val_loss: 8420.1631
Epoch 66/500
89/89 [==============================] - 1s 8ms/step - loss: 5966.5669 - val_loss: 7300.9307
Epoch 67/500
89/89 [==============================] - 1s 8ms/step - loss: 5290.8247 - val_loss: 6495.5781
Epoch 68/500
89/89 [==============================] - 1s 9ms/step - loss: 4759.8613 - val_loss: 5921.5303
Epoch 69/500
89/89 [==============================] - 1s 8ms/step - loss: 4207.7979 - val_loss: 5239.5376
Epoch 70/500
89/89 [==============================] - 1s 8ms/step - loss: 3770.0527 - val_loss: 4750.5854
Epoch 71/500
89/89 [==============================] - 1s 8ms/step - loss: 3375.0735 - val_loss: 4312.5293
Epoch 72/500
89/89 [==============================] - 1s 9ms/step - loss: 3059.5713 - val_loss: 3904.9648
Epoch 73/500
89/89 [==============================] - 1s 9ms/step - loss: 2850.0164 - val_loss: 3737.8792
Epoch 74/500
89/89 [==============================] - 1s 8ms/step - loss: 2564.4143 - val_loss: 3438.4224
Epoch 75/500
89/89 [==============================] - 1s 9ms/step - loss: 2357.3450 - val_loss: 3358.5623
Epoch 76/500
89/89 [==============================] - 1s 9ms/step - loss: 2232.6135 - val_loss: 2793.4851
Epoch 77/500
89/89 [==============================] - 1s 9ms/step - loss: 1997.9294 - val_loss: 2636.9138
Epoch 78/500
89/89 [==============================] - 1s 9ms/step - loss: 1915.8571 - val_loss: 2517.6533
Epoch 79/500
89/89 [==============================] - 1s 9ms/step - loss: 1729.1658 - val_loss: 2317.3748
Epoch 80/500
89/89 [==============================] - 1s 9ms/step - loss: 1631.3309 - val_loss: 2103.8298
Epoch 81/500
89/89 [==============================] - 1s 9ms/step - loss: 1550.3922 - val_loss: 2060.3123
Epoch 82/500
89/89 [==============================] - 1s 9ms/step - loss: 1463.8959 - val_loss: 2043.3530
Epoch 83/500
89/89 [==============================] - 1s 9ms/step - loss: 1433.6577 - val_loss: 1867.8933
Epoch 84/500
89/89 [==============================] - 1s 9ms/step - loss: 1379.9380 - val_loss: 1746.9908
Epoch 85/500
89/89 [==============================] - 1s 9ms/step - loss: 1307.4775 - val_loss: 1657.3741
Epoch 86/500
89/89 [==============================] - 1s 8ms/step - loss: 1227.0078 - val_loss: 1609.0542
Epoch 87/500
89/89 [==============================] - 1s 9ms/step - loss: 1209.1255 - val_loss: 1649.5646
Epoch 88/500
89/89 [==============================] - 1s 8ms/step - loss: 1163.1046 - val_loss: 1585.8846
Epoch 89/500
89/89 [==============================] - 1s 8ms/step - loss: 1122.1663 - val_loss: 1381.3331
Epoch 90/500
89/89 [==============================] - 1s 8ms/step - loss: 1103.4403 - val_loss: 1333.7379
Epoch 91/500
89/89 [==============================] - 1s 9ms/step - loss: 1084.5853 - val_loss: 1350.7050
Epoch 92/500
89/89 [==============================] - 1s 9ms/step - loss: 1051.3956 - val_loss: 1296.9717
Epoch 93/500
89/89 [==============================] - 1s 9ms/step - loss: 1053.0925 - val_loss: 1309.9183
Epoch 94/500
89/89 [==============================] - 1s 9ms/step - loss: 1048.1725 - val_loss: 1329.9364
Epoch 95/500
89/89 [==============================] - 1s 10ms/step - loss: 995.8461 - val_loss: 1263.0800
Epoch 96/500
89/89 [==============================] - 1s 9ms/step - loss: 990.8187 - val_loss: 1250.9036
Epoch 97/500
89/89 [==============================] - 1s 9ms/step - loss: 958.6639 - val_loss: 1410.3779
Epoch 98/500
89/89 [==============================] - 1s 9ms/step - loss: 997.1367 - val_loss: 1562.8750
Epoch 99/500
89/89 [==============================] - 1s 9ms/step - loss: 1018.0589 - val_loss: 1231.3599
Epoch 100/500
89/89 [==============================] - 1s 9ms/step - loss: 937.6019 - val_loss: 1076.1615
Epoch 101/500
89/89 [==============================] - 1s 8ms/step - loss: 963.8284 - val_loss: 1120.6255
Epoch 102/500
89/89 [==============================] - 1s 9ms/step - loss: 946.0831 - val_loss: 1294.1167
Epoch 103/500
89/89 [==============================] - 1s 8ms/step - loss: 907.2535 - val_loss: 1006.6993
Epoch 104/500
89/89 [==============================] - 1s 9ms/step - loss: 907.2841 - val_loss: 1216.4642
Epoch 105/500
89/89 [==============================] - 1s 9ms/step - loss: 916.0087 - val_loss: 1297.0554
Epoch 106/500
89/89 [==============================] - 1s 9ms/step - loss: 921.1118 - val_loss: 1310.7106
Epoch 107/500
89/89 [==============================] - 1s 10ms/step - loss: 923.8624 - val_loss: 1510.4812
Epoch 108/500
89/89 [==============================] - 1s 9ms/step - loss: 954.5092 - val_loss: 1170.1067
Epoch 109/500
89/89 [==============================] - 1s 9ms/step - loss: 917.3290 - val_loss: 1012.5935
Epoch 110/500
89/89 [==============================] - 1s 9ms/step - loss: 893.7215 - val_loss: 1273.6051
Epoch 111/500
89/89 [==============================] - 1s 9ms/step - loss: 895.7330 - val_loss: 1229.0637
Epoch 112/500
89/89 [==============================] - 1s 9ms/step - loss: 908.8897 - val_loss: 1031.2739
Epoch 113/500
89/89 [==============================] - 1s 9ms/step - loss: 960.9525 - val_loss: 1069.6318
Epoch 114/500
89/89 [==============================] - 1s 9ms/step - loss: 872.4641 - val_loss: 1066.6165
Epoch 115/500
89/89 [==============================] - 1s 9ms/step - loss: 881.0060 - val_loss: 1067.6473
Epoch 116/500
89/89 [==============================] - 1s 9ms/step - loss: 878.1358 - val_loss: 1353.8790
Epoch 117/500
89/89 [==============================] - 1s 9ms/step - loss: 895.1171 - val_loss: 1000.1158
Epoch 118/500
89/89 [==============================] - 1s 9ms/step - loss: 915.8824 - val_loss: 970.6990
Epoch 119/500
89/89 [==============================] - 1s 9ms/step - loss: 904.9184 - val_loss: 1093.6401
Epoch 120/500
89/89 [==============================] - 1s 9ms/step - loss: 902.4607 - val_loss: 1033.6764
Epoch 121/500
89/89 [==============================] - 1s 9ms/step - loss: 889.4005 - val_loss: 1611.1193
Epoch 122/500
89/89 [==============================] - 1s 9ms/step - loss: 892.5366 - val_loss: 984.4285
Epoch 123/500
89/89 [==============================] - 1s 9ms/step - loss: 921.8589 - val_loss: 1402.7405
Epoch 124/500
89/89 [==============================] - 1s 9ms/step - loss: 875.1534 - val_loss: 1046.4728
Epoch 125/500
89/89 [==============================] - 1s 8ms/step - loss: 849.3232 - val_loss: 961.7246
Epoch 126/500
89/89 [==============================] - 1s 9ms/step - loss: 886.5376 - val_loss: 1025.4518
Epoch 127/500
89/89 [==============================] - 1s 9ms/step - loss: 905.3434 - val_loss: 1285.8262
Epoch 128/500
89/89 [==============================] - 1s 9ms/step - loss: 904.4112 - val_loss: 1076.7793
Epoch 129/500
89/89 [==============================] - 1s 9ms/step - loss: 893.3470 - val_loss: 1285.9775
Epoch 130/500
89/89 [==============================] - 1s 9ms/step - loss: 871.8917 - val_loss: 1979.8333
Epoch 131/500
89/89 [==============================] - 1s 9ms/step - loss: 951.7941 - val_loss: 875.3125
Epoch 132/500
89/89 [==============================] - 1s 9ms/step - loss: 902.2049 - val_loss: 986.6984
Epoch 133/500
89/89 [==============================] - 1s 9ms/step - loss: 887.7725 - val_loss: 887.2075
Epoch 134/500
89/89 [==============================] - 1s 8ms/step - loss: 897.8741 - val_loss: 1218.7233
Epoch 135/500
89/89 [==============================] - 1s 9ms/step - loss: 910.2280 - val_loss: 873.9462
Epoch 136/500
89/89 [==============================] - 1s 9ms/step - loss: 898.3157 - val_loss: 1551.2761
Epoch 137/500
89/89 [==============================] - 1s 8ms/step - loss: 902.2271 - val_loss: 1377.0199
Epoch 138/500
89/89 [==============================] - 1s 9ms/step - loss: 865.7802 - val_loss: 1155.8024
Epoch 139/500
89/89 [==============================] - 1s 9ms/step - loss: 875.7888 - val_loss: 963.3873
Epoch 140/500
89/89 [==============================] - 1s 9ms/step - loss: 878.3383 - val_loss: 978.4841
Epoch 141/500
89/89 [==============================] - 1s 9ms/step - loss: 857.4650 - val_loss: 988.0334
Epoch 142/500
89/89 [==============================] - 1s 8ms/step - loss: 884.1348 - val_loss: 1442.5347
Epoch 143/500
89/89 [==============================] - 1s 8ms/step - loss: 852.4373 - val_loss: 1036.8562
Epoch 144/500
89/89 [==============================] - 1s 8ms/step - loss: 864.8729 - val_loss: 875.8694
Epoch 145/500
89/89 [==============================] - 1s 9ms/step - loss: 885.4371 - val_loss: 1141.1495
Epoch 146/500
89/89 [==============================] - 1s 9ms/step - loss: 859.5933 - val_loss: 1095.2089
Epoch 147/500
89/89 [==============================] - 1s 9ms/step - loss: 858.5456 - val_loss: 1033.9537
Epoch 148/500
89/89 [==============================] - 1s 8ms/step - loss: 887.5352 - val_loss: 1058.0504
Epoch 149/500
89/89 [==============================] - 1s 8ms/step - loss: 926.1168 - val_loss: 1259.0009
Epoch 150/500
89/89 [==============================] - 1s 9ms/step - loss: 885.6328 - val_loss: 1103.0502
Epoch 151/500
89/89 [==============================] - 1s 9ms/step - loss: 831.0637 - val_loss: 1859.7780
Epoch 152/500
89/89 [==============================] - 1s 9ms/step - loss: 883.6222 - val_loss: 926.1632
Epoch 153/500
89/89 [==============================] - 1s 10ms/step - loss: 865.0244 - val_loss: 883.8705
Epoch 154/500
89/89 [==============================] - 1s 10ms/step - loss: 900.2144 - val_loss: 965.1517
Epoch 155/500
89/89 [==============================] - 1s 9ms/step - loss: 857.1143 - val_loss: 1058.2616
Epoch 156/500
89/89 [==============================] - 1s 9ms/step - loss: 879.2825 - val_loss: 1146.2948
Epoch 157/500
89/89 [==============================] - 1s 9ms/step - loss: 852.4259 - val_loss: 1225.2124
Epoch 158/500
89/89 [==============================] - 1s 8ms/step - loss: 909.5094 - val_loss: 1384.2289
Epoch 159/500
89/89 [==============================] - 1s 9ms/step - loss: 844.4752 - val_loss: 851.6119
Epoch 160/500
89/89 [==============================] - 1s 9ms/step - loss: 871.4876 - val_loss: 1137.0043
Epoch 161/500
89/89 [==============================] - 1s 9ms/step - loss: 873.1281 - val_loss: 839.7911
Epoch 162/500
89/89 [==============================] - 1s 9ms/step - loss: 882.4369 - val_loss: 958.2319
Epoch 163/500
89/89 [==============================] - 1s 8ms/step - loss: 897.3118 - val_loss: 976.2214
Epoch 164/500
89/89 [==============================] - 1s 9ms/step - loss: 863.2068 - val_loss: 1056.6986
Epoch 165/500
89/89 [==============================] - 1s 9ms/step - loss: 900.8514 - val_loss: 1004.8099
Epoch 166/500
89/89 [==============================] - 1s 8ms/step - loss: 894.6444 - val_loss: 988.3345
Epoch 167/500
89/89 [==============================] - 1s 8ms/step - loss: 889.9005 - val_loss: 1560.9814
Epoch 168/500
89/89 [==============================] - 1s 9ms/step - loss: 882.0018 - val_loss: 872.4977
Epoch 169/500
89/89 [==============================] - 1s 13ms/step - loss: 887.3175 - val_loss: 1488.1952
Epoch 170/500
89/89 [==============================] - 1s 9ms/step - loss: 827.1350 - val_loss: 870.1962
Epoch 171/500
89/89 [==============================] - 1s 8ms/step - loss: 857.9397 - val_loss: 934.2194
Epoch 172/500
89/89 [==============================] - 1s 9ms/step - loss: 853.0777 - val_loss: 920.4064
Epoch 173/500
89/89 [==============================] - 1s 10ms/step - loss: 884.1110 - val_loss: 1187.6647
Epoch 174/500
89/89 [==============================] - 1s 13ms/step - loss: 866.8095 - val_loss: 888.5013
Epoch 175/500
89/89 [==============================] - 1s 8ms/step - loss: 852.5687 - val_loss: 1314.4453
Epoch 176/500
89/89 [==============================] - 1s 9ms/step - loss: 859.4567 - val_loss: 1150.2583
Epoch 177/500
89/89 [==============================] - 1s 9ms/step - loss: 848.0142 - val_loss: 1070.6479
Epoch 178/500
89/89 [==============================] - 1s 10ms/step - loss: 839.4297 - val_loss: 938.0947
Epoch 179/500
89/89 [==============================] - 1s 8ms/step - loss: 865.0428 - val_loss: 1112.5199
Epoch 180/500
89/89 [==============================] - 1s 9ms/step - loss: 890.2850 - val_loss: 1593.5288
Epoch 181/500
89/89 [==============================] - 1s 9ms/step - loss: 890.6694 - val_loss: 1251.8132
Epoch 182/500
89/89 [==============================] - 1s 9ms/step - loss: 865.7590 - val_loss: 1131.9238
Epoch 183/500
89/89 [==============================] - 1s 9ms/step - loss: 897.3402 - val_loss: 1227.0776
Epoch 184/500
89/89 [==============================] - 1s 9ms/step - loss: 847.9206 - val_loss: 965.4744
Epoch 185/500
89/89 [==============================] - 1s 9ms/step - loss: 833.6210 - val_loss: 1284.7535
Epoch 186/500
89/89 [==============================] - 1s 8ms/step - loss: 853.7308 - val_loss: 1018.0238
Epoch 186: early stopping





<keras.callbacks.History at 0x7f42a1213fd0>

# So there are 162 iterations. Let's recap our training loss vs validation loss, and make it to DataFrame.

model_loss = pd.DataFrame(model.history.history)
model_loss
loss val_loss
0 6.728794e+06 7.068232e+06
1 6.723610e+06 7.057610e+06
2 6.707948e+06 7.036904e+06
3 6.682746e+06 7.005204e+06
4 6.644921e+06 6.959078e+06
... ... ...
181 8.657590e+02 1.131924e+03
182 8.973402e+02 1.227078e+03
183 8.479206e+02 9.654744e+02
184 8.336210e+02 1.284754e+03
185 8.537308e+02 1.018024e+03

186 rows × 2 columns

plt.figure(figsize = (12,6))
model_loss.plot()
plt.xlabel('n of Epochs')

Text(0.5, 0, 'n of Epochs')




<Figure size 864x432 with 0 Axes>

png

I can’t believe this our model get “just right” form along the training. We see that ‘training loss’ and ‘validation loss’ walk along the side. It’s beautiful graph that you could see.

4) Predictions

Training is done, let’s get our model to predict and compare it with y_test dataset. We will see the value of RMSE and linear regression graph.

from sklearn.metrics import mean_squared_error, explained_variance_score

predictions = model.predict(X_test)

23/23 [==============================] - 1s 3ms/step

# MSE
mean_squared_error(y_test, predictions)

1018.0237450375761

# RMSE
np.sqrt(mean_squared_error(y_test, predictions))

31.906484372891605

# Explained variance regression score function

explained_variance_score(y_test, predictions)
# Best possible score is 1.0, lower values are worse (sklearn).

0.9994461431887325

We see that we get the value of RMSE is 31.906. It’s very small according to the context of stocks. Look at the ‘explained variance score’, we get 0.9994. Our model just did a great job. So what if we plot our valid dataset with model’s prediction? Could we get some nice line of regression?

plt.figure(figsize=(12,6))
plt.scatter(y_test,predictions)
plt.plot(y_test, y_test, 'r', linewidth = 2.5)
plt.xlabel('y_test')
plt.ylabel('predictions')
plt.title('LSTM Model Prediction Evaluation')

Text(0.5, 1.0, 'LSTM Model Prediction Evaluation')

png

Still can’t believe this that our model’s predictions just fit with y_test dataset. It shows fit very well with few outliers there. It’s okay. Let’s try to predict closing price of BBRI on June 2nd.

5) Predict with Another Data on June 2nd, 2022

First we get our dataset, then prepare it up into our model’s prediction.

# BBRI
bbri2 = web.DataReader('bbri.jk', 'yahoo', '2022-06-01', '2022-06-01')
bbri2
High Low Open Close Volume Adj Close
Date
2020-06-02 3290 2950 2950 3180 467460800 2993.267334 
bbri2 = bbri2.drop('Close', axis = 1)

bbri2 = bbri2.values

bbri2 = scaler.transform(bbri2)

bbri2 = bbri2.reshape(-1, 5, 1)

model.predict(bbri2)

1/1 [==============================] - 0s 21ms/step





array([[3157.858]], dtype=float32)

Our model predicts 3124 of BBRI’s closing price on June 2nd, 2022 while the actual close price is 3180. Well, that’s not bad though since our model’s RMSE is about 31 units and the evaluation just fit between its predictions and y_test dataset.

← Previous Post Next Post →