Reference: https://www.geeksforgeeks.org/stock-price-prediction-using-machine-learning-in-python/

In [41]:

%pip install pandas -q
%pip install numpy -q
%pip install seaborn  -q
%pip install matplotlib -q
%pip install yfinance -q
%pip install scikit-learn -q
%pip install yellowbrick

[1m[[0m[34;49mnotice[0m[1;39;49m][0m[39;49m A new release of pip is available: [0m[31;49m23.0.1[0m[39;49m -> [0m[32;49m23.2.1[0m
[1m[[0m[34;49mnotice[0m[1;39;49m][0m[39;49m To update, run: [0m[32;49mpip install --upgrade pip[0m
Note: you may need to restart the kernel to use updated packages.

[1m[[0m[34;49mnotice[0m[1;39;49m][0m[39;49m A new release of pip is available: [0m[31;49m23.0.1[0m[39;49m -> [0m[32;49m23.2.1[0m
[1m[[0m[34;49mnotice[0m[1;39;49m][0m[39;49m To update, run: [0m[32;49mpip install --upgrade pip[0m
Note: you may need to restart the kernel to use updated packages.

[1m[[0m[34;49mnotice[0m[1;39;49m][0m[39;49m A new release of pip is available: [0m[31;49m23.0.1[0m[39;49m -> [0m[32;49m23.2.1[0m
[1m[[0m[34;49mnotice[0m[1;39;49m][0m[39;49m To update, run: [0m[32;49mpip install --upgrade pip[0m
Note: you may need to restart the kernel to use updated packages.

[1m[[0m[34;49mnotice[0m[1;39;49m][0m[39;49m A new release of pip is available: [0m[31;49m23.0.1[0m[39;49m -> [0m[32;49m23.2.1[0m
[1m[[0m[34;49mnotice[0m[1;39;49m][0m[39;49m To update, run: [0m[32;49mpip install --upgrade pip[0m
Note: you may need to restart the kernel to use updated packages.

[1m[[0m[34;49mnotice[0m[1;39;49m][0m[39;49m A new release of pip is available: [0m[31;49m23.0.1[0m[39;49m -> [0m[32;49m23.2.1[0m
[1m[[0m[34;49mnotice[0m[1;39;49m][0m[39;49m To update, run: [0m[32;49mpip install --upgrade pip[0m
Note: you may need to restart the kernel to use updated packages.

[1m[[0m[34;49mnotice[0m[1;39;49m][0m[39;49m A new release of pip is available: [0m[31;49m23.0.1[0m[39;49m -> [0m[32;49m23.2.1[0m
[1m[[0m[34;49mnotice[0m[1;39;49m][0m[39;49m To update, run: [0m[32;49mpip install --upgrade pip[0m
Note: you may need to restart the kernel to use updated packages.
Collecting yellowbrick
  Using cached yellowbrick-1.5-py3-none-any.whl (282 kB)
Requirement already satisfied: scipy>=1.0.0 in ./venv/lib/python3.10/site-packages (from yellowbrick) (1.11.1)
Requirement already satisfied: cycler>=0.10.0 in ./venv/lib/python3.10/site-packages (from yellowbrick) (0.11.0)
Requirement already satisfied: numpy>=1.16.0 in ./venv/lib/python3.10/site-packages (from yellowbrick) (1.25.2)
Requirement already satisfied: matplotlib!=3.0.0,>=2.0.2 in ./venv/lib/python3.10/site-packages (from yellowbrick) (3.7.2)
Requirement already satisfied: scikit-learn>=1.0.0 in ./venv/lib/python3.10/site-packages (from yellowbrick) (1.3.0)
Requirement already satisfied: kiwisolver>=1.0.1 in ./venv/lib/python3.10/site-packages (from matplotlib!=3.0.0,>=2.0.2->yellowbrick) (1.4.4)
Requirement already satisfied: fonttools>=4.22.0 in ./venv/lib/python3.10/site-packages (from matplotlib!=3.0.0,>=2.0.2->yellowbrick) (4.42.0)
Requirement already satisfied: contourpy>=1.0.1 in ./venv/lib/python3.10/site-packages (from matplotlib!=3.0.0,>=2.0.2->yellowbrick) (1.1.0)
Requirement already satisfied: pyparsing<3.1,>=2.3.1 in ./venv/lib/python3.10/site-packages (from matplotlib!=3.0.0,>=2.0.2->yellowbrick) (3.0.9)
Requirement already satisfied: pillow>=6.2.0 in ./venv/lib/python3.10/site-packages (from matplotlib!=3.0.0,>=2.0.2->yellowbrick) (10.0.0)
Requirement already satisfied: python-dateutil>=2.7 in ./venv/lib/python3.10/site-packages (from matplotlib!=3.0.0,>=2.0.2->yellowbrick) (2.8.2)
Requirement already satisfied: packaging>=20.0 in ./venv/lib/python3.10/site-packages (from matplotlib!=3.0.0,>=2.0.2->yellowbrick) (23.1)
Requirement already satisfied: threadpoolctl>=2.0.0 in ./venv/lib/python3.10/site-packages (from scikit-learn>=1.0.0->yellowbrick) (3.2.0)
Requirement already satisfied: joblib>=1.1.1 in ./venv/lib/python3.10/site-packages (from scikit-learn>=1.0.0->yellowbrick) (1.3.2)
Requirement already satisfied: six>=1.5 in ./venv/lib/python3.10/site-packages (from python-dateutil>=2.7->matplotlib!=3.0.0,>=2.0.2->yellowbrick) (1.16.0)
Installing collected packages: yellowbrick
Successfully installed yellowbrick-1.5

[1m[[0m[34;49mnotice[0m[1;39;49m][0m[39;49m A new release of pip is available: [0m[31;49m23.0.1[0m[39;49m -> [0m[32;49m23.2.1[0m
[1m[[0m[34;49mnotice[0m[1;39;49m][0m[39;49m To update, run: [0m[32;49mpip install --upgrade pip[0m
Note: you may need to restart the kernel to use updated packages.

In [59]:

import pandas as pd
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
import yfinance as yf

from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from sklearn.linear_model import LogisticRegression
from sklearn.svm import SVC
from sklearn.metrics import classification_report, accuracy_score
from yellowbrick.classifier import ConfusionMatrix

Load dataset of bank stock price

In [3]:

ticker = "ITUB4.SA"
data = yf.download(ticker, start="2008-01-01", progress=False)

In [4]:

data.head()

Out [4]:

	Open	High	Low	Close	Adj Close	Volume
Date
2008-01-02	15.100806	15.223335	14.306027	14.411997	8.390958	6442543
2008-01-03	14.372258	14.736532	14.117267	14.140448	8.232860	7212266
2008-01-04	14.315962	14.431867	13.799355	14.239795	8.290697	7374122
2008-01-07	14.292780	14.570953	14.007985	14.239795	8.290697	7597580
2008-01-08	14.534526	14.703416	14.256353	14.372258	8.367819	5372057

Data analysis

Visualize dataset info

In [5]:

data.shape

Out [5]:

(3876, 6)

In [6]:

data.describe()

Out [6]:

	Open	High	Low	Close	Adj Close	Volume
count	3876.000000	3876.000000	3876.000000	3876.000000	3876.000000	3.876000e+03
mean	21.517539	21.799286	21.222586	21.512633	16.731696	2.267386e+07
std	7.123009	7.186481	7.053429	7.115916	7.935137	1.393839e+07
min	7.244082	7.653890	6.999853	7.244082	4.217649	0.000000e+00
25%	15.766475	15.967374	15.577260	15.775082	9.843326	1.333177e+07
50%	19.382919	19.682645	19.144298	19.423195	13.558575	1.949876e+07
75%	26.915833	27.295000	26.580000	26.912500	24.145989	2.864823e+07
max	38.669998	39.790001	38.400002	39.689999	33.508636	1.606699e+08

In [7]:

data.info()

<class 'pandas.core.frame.DataFrame'>
DatetimeIndex: 3876 entries, 2008-01-02 to 2023-08-11
Data columns (total 6 columns):
 #   Column     Non-Null Count  Dtype  
---  ------     --------------  -----  
 0   Open       3876 non-null   float64
 1   High       3876 non-null   float64
 2   Low        3876 non-null   float64
 3   Close      3876 non-null   float64
 4   Adj Close  3876 non-null   float64
 5   Volume     3876 non-null   int64  
dtypes: float64(5), int64(1)
memory usage: 212.0 KB

Exploratory Data Analysis

Now let's look at how quarterly results can affect stock prices over time.

In [8]:

plt.figure(figsize=(15, 5))
plt.plot(data["Adj Close"])
plt.title("ITUB4 close price", fontsize=24)
plt.ylabel("Price (in R$)")
plt.show()

Let"s check if the "Close" and "Adj Close" have the same values, so that we can remove possibly redundant data.

In [9]:

data

Out [9]:

	Open	High	Low	Close	Adj Close	Volume
Date
2008-01-02	15.100806	15.223335	14.306027	14.411997	8.390958	6442543
2008-01-03	14.372258	14.736532	14.117267	14.140448	8.232860	7212266
2008-01-04	14.315962	14.431867	13.799355	14.239795	8.290697	7374122
2008-01-07	14.292780	14.570953	14.007985	14.239795	8.290697	7597580
2008-01-08	14.534526	14.703416	14.256353	14.372258	8.367819	5372057
...	...	...	...	...	...	...
2023-08-07	27.930000	28.280001	27.620001	27.660000	27.660000	33001900
2023-08-08	27.360001	28.010000	27.110001	27.600000	27.600000	44517500
2023-08-09	27.500000	27.600000	26.920000	27.500000	27.500000	35060900
2023-08-10	27.650000	27.940001	27.520000	27.590000	27.590000	30620600
2023-08-11	27.629999	27.799999	27.459999	27.639999	27.639999	19424700

3876 rows × 6 columns

In [10]:

data[data["Close"] == data["Adj Close"]].shape

Out [10]:

(9, 6)

As we can notice, only 8 rows the "Close" and "Adj Close" have the same value, so we can't remove it.

Before moving on, let's check the dataset for null values.

In [11]:

data.isnull().sum()

Out [11]:

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

In [12]:


features = ["Open", "High", "Low", "Close", "Adj Close", "Volume"]

plt.subplots(figsize=(20, 10))

for i, column in enumerate(features):
    plt.subplot(2, 3, i+1)

    # https://stackoverflow.com/questions/67638590/emulating-deprecated-seaborn-distplots
    _, FD_bins = np.histogram(data[column], bins="fd")
    bin_nr = min(len(FD_bins)-1, 50)
    sns.histplot(data=data, x=column, bins=bin_nr,
                 stat="density", alpha=0.4, kde=True, kde_kws={"cut": 3})

plt.show()

/tmp/ipykernel_22901/980192968.py:6: MatplotlibDeprecationWarning: Auto-removal of overlapping axes is deprecated since 3.6 and will be removed two minor releases later; explicitly call ax.remove() as needed.
  plt.subplot(2, 3, i+1)

After that, let's create the box plot to identify outliers.

In [13]:

plt.subplots(figsize=(20,10))

for i, column in enumerate(features):
  plt.subplot(2,3,i+1)
  sns.boxplot(data[column]).set_title(column)

plt.show()

/tmp/ipykernel_22901/2157267840.py:4: MatplotlibDeprecationWarning: Auto-removal of overlapping axes is deprecated since 3.6 and will be removed two minor releases later; explicitly call ax.remove() as needed.
  plt.subplot(2,3,i+1)

We can conclude that only in the volume column have outliers.

Feature Engineering

Let's add 3 new columns "day", "month" and "year" from column "Date".

In [18]:

data['day'] = data.index.day.astype('int')
data['month'] = data.index.month.astype('int')
data['year'] = data.index.year.astype('int')
 
data.head()

Out [18]:

	Open	High	Low	Close	Adj Close	Volume	day	month	year
Date
2008-01-02	15.100806	15.223335	14.306027	14.411997	8.390958	6442543	2	1	2008
2008-01-03	14.372258	14.736532	14.117267	14.140448	8.232860	7212266	3	1	2008
2008-01-04	14.315962	14.431867	13.799355	14.239795	8.290697	7374122	4	1	2008
2008-01-07	14.292780	14.570953	14.007985	14.239795	8.290697	7597580	7	1	2008
2008-01-08	14.534526	14.703416	14.256353	14.372258	8.367819	5372057	8	1	2008

Now we create an is_quarter column to use later in the algorithm because every quarter the companies publish the quarterly results.

In [25]:

data['is_quarter_end'] = np.where(data['month'] % 3 == 0, 1, 0)
data.head()

Out [25]:

	Open	High	Low	Close	Adj Close	Volume	day	month	year	is_quarter_end
Date
2008-01-02	15.100806	15.223335	14.306027	14.411997	8.390958	6442543	2	1	2008	0
2008-01-03	14.372258	14.736532	14.117267	14.140448	8.232860	7212266	3	1	2008	0
2008-01-04	14.315962	14.431867	13.799355	14.239795	8.290697	7374122	4	1	2008	0
2008-01-07	14.292780	14.570953	14.007985	14.239795	8.290697	7597580	7	1	2008	0
2008-01-08	14.534526	14.703416	14.256353	14.372258	8.367819	5372057	8	1	2008	0

In [29]:

data_grouped = data.groupby('year').mean()
plt.subplots(figsize=(20,15))
 
for i, column in enumerate(['Open', 'High', 'Low', 'Close']):
  plt.subplot(2,2,i+1)
  data_grouped[col].plot.bar().set_title(column)

plt.show()

/tmp/ipykernel_22901/1494818884.py:5: MatplotlibDeprecationWarning: Auto-removal of overlapping axes is deprecated since 3.6 and will be removed two minor releases later; explicitly call ax.remove() as needed.
  plt.subplot(2,2,i+1)

In [31]:

data.groupby('is_quarter_end').mean()

Out [31]:

	Open	High	Low	Close	Adj Close	Volume	day	month	year
is_quarter_end
0	21.562839	21.844573	21.272231	21.561341	16.760581	2.262913e+07	15.810131	5.978732	2015.344934
1	21.426729	21.708502	21.123064	21.414992	16.673791	2.276353e+07	15.598450	7.327907	2015.277519

With the above DataFrame we can make some importante observations:

The price ("Adj Close") is slightly lower when it is at the end of a quarter
The volume has a small increase at the end of a quarter

In [32]:

data['open-close']  = data['Open'] - data['Adj Close']
data['low-high']  = data['Low'] - data['High']
data['target'] = np.where(data['Adj Close'].shift(-1) > data['Adj Close'], 1, 0) # "1" if the "Adj Close" of the previous day is higher than the current "Adj Close"

The "target" is the most importante feature, because it will be used to help model to buy or not.

In [33]:

plt.pie(
  data['target'].value_counts().values,
  labels=[0, 1], 
  autopct='%1.1f%%'
)
plt.show()

In [35]:

plt.figure(figsize=(10, 10))
 
# As our concern is with the highly
# correlated features only so, we will visualize
# our heatmap as per that criteria only.
sns.heatmap(data.corr() > 0.9, annot=True, cbar=False)
plt.show()

Data Splitting and Normalization

In [94]:

features = data[['open-close', 'low-high', 'is_quarter_end']]
target = data['target']
 
scaler = StandardScaler()
features = scaler.fit_transform(features)
 
X_train, X_test, y_train, y_test = train_test_split(
    features, target, test_size=0.1, random_state=2022
)

X_train.shape, X_test.shape

Out [94]:

((3488, 3), (388, 3))

Model Development and Evaluation

In [116]:

model = SVC(kernel='sigmoid', C=1, probability=True)
 
model.fit(X_train, y_train)
 
predicts = model.predict(X_test)

In [117]:

cm = ConfusionMatrix(model)
cm.fit(X_train, y_train)
cm.score(X_test, y_test)

Out [117]:

0.5180412371134021

In [118]:

print(classification_report(y_test, predicts))

              precision    recall  f1-score   support

           0       0.48      0.52      0.50       181
           1       0.55      0.51      0.53       207

    accuracy                           0.52       388
   macro avg       0.52      0.52      0.52       388
weighted avg       0.52      0.52      0.52       388

In [119]:

accuracy_score(Y_valid, predicts)

Out [119]:

0.5180412371134021

In [ ]:

451 KiB Raw Blame History Unescape Escape