Machine learning for beginners: Build your first ML project.
Aren't you amazed by what actually “A storyteller who writes software that tells story” means? Well, neither it’s that deep :)) nor this is the correct place and time to discuss such “humanities student” stuff.
The purpose of attaching the snapshot of my LinkedIn is just to make you feel secure that you are going to learn something from someone who does have some credibility ( a little, also not very active on LinkedIn). LinkedIn profile https://www.linkedin.com/in/sumit-sinha-a5a86a1a6/
Today, the blog is dedicated to AIML enthusiasts who aren't able to join the race of Artificial intelligence. This single blog for sure won’t be enough for anyone to learn machine learning and get into FAANG or win some big hackathons, but this blog will teach you how you can build a simple regression model in machine learning with just basic knowledge of Python. Understanding linear algebra, statistics, and probability will be a plus.
We are going to build a used car prediction model from scratch. For this, you need to know Python, high school maths, equations of line, and how to type. Out of 4 you already know 2 things, congratulations you completed the project by 50%, and the rest 50% starts from here.
I will share the code snippet and you can just run those snippets into a separate jupyter notebook, better if you use Google Colab.
Import useful libraries:
#import required libraries
import pandas as pd
import matplotlib.pyplot as plt
%matplotlib inline
import seaborn as sns
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from sklearn.linear_model import LinearRegression
from sklearn import metrics
from sklearn.metrics import mean_absolute_error, mean_squared_error, r2_score
import ipywidgets as widgets
from IPython.display import displayImport data:
#read_csv() takes the path of the csv file using pandas pd.read_csv() convert csv into dataframe
car_data = pd.read_csv('/content/car_data.csv')
#head() is used to display the first 5 rows of the dataframe
car_data.head()Check data:
#info() provides summary of the information present in dataframe
car_data.info()#sum of null values for each column
car_data.isnull().sum()
#Describe the statistical values
car_data.describe()
#Returns the list of columns
car_data.columnsData visualization process:
#to count the varieties and print
print("---------------------------")
print(car_data['Fuel_Type'].value_counts())
print("---------------------------")
print(car_data['Seller_Type'].value_counts())
print("---------------------------")
print(car_data['Transmission'].value_counts())
print("---------------------------")Next,
fuel_type = car_data['Fuel_Type']
seller_type = car_data['Seller_Type']
transmission_type = car_data['Transmission']
selling_price = car_data['Selling_Price']Plot graph to understand data visually:
# Set the style of the plot to 'ggplot' for a specific visual style.
style.use('ggplot')
# Create a figure with a specified size.
fig = plt.figure(figsize=(15,5))
# Set the overall title for the figure.
fig.suptitle('Visualizing categorical data columns')
# Create the first subplot in a 1x3 grid.
plt.subplot(1,3,1)
# Create a bar plot for fuel_type against selling_price with a specified color.
plt.bar(fuel_type, selling_price, color='royalblue')
# Set labels for the x and y axes.
plt.xlabel("Fuel Type")
plt.ylabel("Selling Price")
# Create the second subplot in a 1x3 grid.
plt.subplot(1,3,2)
# Create a bar plot for seller_type against selling_price with a specified color.
plt.bar(seller_type, selling_price, color='red')
# Set labels for the x axis.
plt.xlabel("Seller Type")
# Create the third subplot in a 1x3 grid.
plt.subplot(1,3,3)
# Create a bar plot for transmission_type against selling_price with a specified color.
plt.bar(transmission_type, selling_price, color='purple')
# Set label for the x axis.
plt.xlabel('Transmission type')
# Display the plot.
plt.show()Encoding methods:
There are 2 ways to encode the data
- Manual encoding: In this manual encodings i.e. 0,1… so on are assigned to different values.
- One hot encoding: This is an automatic way to encode the values.
#manual encoding
car_data.replace({'Fuel_Type':{'Petrol':0, 'Diesel':1, 'CNG':2}}, inplace=True)
#one hot encoding
car_data = pd.get_dummies(car_data, columns=['Seller_Type', 'Transmission'], drop_first=True)
car_data.head()Heat Map visualization:
This is used to understand the correlation between different feature sets.
plt.figure(figsize=(10,7))
sns.heatmap(car_data.corr(), annot=True)
plt.title('Correlation between the columns')
plt.show()Draw regression plot:
fig=plt.figure(figsize=(7,5))
plt.title('Correlation between present price and selling price')
sns.regplot(x='Present_Price', y='Selling_Price', data=car_data)Training and testing data preparation:
#cretaing dependent and independent variables. y has the dependent variable
X = car_data.drop(['Car_Name','Selling_Price'], axis=1)
y = car_data['Selling_Price']
#shape() tells the rows and columns
print("Shape of X is: ",X.shape)
print("Shape of y is: ", y.shape)Train test split:
#train test split of data
#x holds independent variables for training and testing
#y hold dependent variable for training and testing
X_train, X_test, y_train, y_test = train_test_split(X,y, test_size=0.3, random_state=42)Shape of the data:
print("X_test shape:", X_test.shape)
print("X_train shape:", X_train.shape)
print("y_test shape: ", y_test.shape)
print("y_train shape:", y_train.shape)Quick tip - No. of rows of train and test data should be equal for eg. X_train has n rows so y_train should also have n rows.
Note: This blog doesn't cover data fitting or data scaling methods. It is always good to scale down the values. Below I am providing the code for scaling you can use if you wish to, but the prediction would be scaled too.
Standard scaling method:
scaler = StandardScaler()
X_train = scaler.fit_transform(X_train)
X_test = scaler.transform(X_test)Building a Linear regression model:
model = LinearRegression()
#fit() is used to train the model using the data given as parameters to it
model.fit(X_train, y_train)Congratulations you just trained your first machine learning model which is a linear regression model.
That simple? Umm.. yesss. But it is just the beginning so keep exploring. But wait the project isn’t complete yet. A good machine learning engineer should know everything about his model including the error and accuracy, so let’s calculate that.
Predict Output:
pred = model.predict(X_test)Important metrics:
These topics are more toward statistics, and will be covered in another article.
print("MAE: ", (metrics.mean_absolute_error(pred, y_test)))
print("MSE: ", (metrics.mean_squared_error(pred, y_test)))
print("R2 score: ", (metrics.r2_score(pred, y_test)))Actual vs Prediction:
sns.regplot(x=pred, y=y_test, scatter_kws={'color': 'blue'}, line_kws={'color': 'red'})
plt.xlabel("Predicted Price")
plt.ylabel('Actual Price')
plt.title("ACtual vs predicted price")
plt.show()Beautifying the project within Colab:
# Create input widgets for different types of features
year_widget = widgets.IntSlider(description='Year:', min=1970, max=2023, step=1, value=2010)
present_price_widget = widgets.FloatSlider(description='Present Price:', min=0.2, max=100, step=0.1, value=9.83)
kilometer_driven_widget = widgets.IntSlider(description='Kilometer Driven:', min=1000, max=100000, step=1000, value=2071)
fuel_type_widget = widgets.Dropdown(description='Fuel Type:', options=['Petrol', 'Diesel', 'CNG'], value='Petrol')
owner_widget = widgets.Dropdown(description='Owner:', options=['First', 'Second', 'Third'], value='First')
seller_type_widget = widgets.Dropdown(description='Seller Type:', options=['Individual', 'Dealer'], value='Individual')
transmission_widget = widgets.Dropdown(description='Transmission:', options=['Manual', 'Automatic'], value='Manual')
# Function to get input values and make a prediction
def make_prediction(button):
fuel_type_encoded = 0 if fuel_type_widget.value == 'Petrol' else 1 if fuel_type_widget.value == 'Diesel' else 3
owner_encoded = 0 if owner_widget.value == 'First' else 1 if owner_widget.value == 'Second' else 2
seller_type_encoded = 0 if seller_type_widget.value == 'Individual' else 1
transmission_encoded = 0 if transmission_widget.value == 'Manual' else 1
new_data_point = [
[year_widget.value, present_price_widget.value, kilometer_driven_widget.value,
fuel_type_encoded, owner_encoded, seller_type_encoded, transmission_encoded]
]
prediction = model.predict(new_data_point)
print("Selling Price in Lakhs:", prediction)
# Create a button to trigger the prediction
predict_button = widgets.Button(description='Predict')
predict_button.on_click(make_prediction)
# Display the input widgets and the button
display(year_widget, present_price_widget, kilometer_driven_widget,
fuel_type_widget, owner_widget, seller_type_widget, transmission_widget,
predict_button)With this, you have successfully built and evaluated a machine-learning model from scratch. Hope you like this article, the link to the notebook is attached: Click to view now.
