Car Price Prediction

Hello Everyone,

Here is my regression project based on predicting the price of used cars using Linear Regression.

Dataset

I used Honda Used Car Selling Dataset which is one of my own dataset uploaded on Kaggle.

Link to the Dataset : Car Price Dataset

Problem Statement

• To develop a Machine Learning Model that can accurately predict the prices of used cars based on various features and attributes.

• The predicted prices will assist both buyers and sellers in making informed decisions, ensuring fair transactions in the used car market.

Streamlit Web App

• For my project, I have created a Streamlit Web App for predicting the prices of cars in more interactive and user friendly way.

• This web app allows you to predict the prices of the cars by just selecting some of its features and fill in some details.

• These all are the features you need to select or enter before pressing the predict button :

  • Year : Select the manufacturing year of the car.

  • kms Driven : Enter the total distance covered by the car.

  • Fuel Type : Choose the fuel type of the car.

  • Suspension : Pick the type of suspension.

  • Car Model : Select your car model from the available options.

• After selecting all these features, Just hit the 'Predict' Button.

• This web app also has multiple constraints in the input fields.

• I have named it AutoValuate.

Link to the Web App : Car Price Prediction App

Table of Contents

• Setting up the Enviroment
• Libraries required for the Project
• Getting started with Repository
• Steps involved in the Project
• Conclusion

Setting up the Enviroment

Jupyter Notebook is required for this project and you can install and set it up in the terminal.

• Install the Notebook

pip install notebook

• Run the Notebook

jupyter notebook

Libraries required for the Project

Pandas

• Go to the terminal and run this code

pip install pandas

• Go to Jupyter Notebook and run this code from a cell

!pip install pandas

Matplotlib

• Go to the terminal and run this code

pip install matplotlib

• Go to Jupyter Notebook and run this code from a cell

!pip install matplotlib

Seaborn

• Go to the terminal and run this code

pip install seaborn

• Go to Jupyter Notebook and run this code from a cell

!pip install seaborn

Sklearn

• Go to the terminal and run this code

pip install scikit-learn

• Go to Jupyter Notebook and run this code from a cell

!pip install scikit-learn

Getting Started

• Clone this repository to your local machine by using the following command :

git clone https://github.com/TheMrityunjayPathak/CarPricePrediction.git

Steps involved in the Project

Data Cleaning

• fuel_type, suspension and car_model has extra whitespaces which is removed by str.strip() method.

• Removed kms suffix from kms_driven column by using str.split() method and keeping only numeric part of the column.

• After that we can convert kms_driven column to 'int' DataType.

• Modifying price column from 6.45 Lakh to 645000 and convering it into integer by using a custom made function.

• From car_model column we will keep only first 3 words of car name and removing the rest.

Data Visualization

• Visualizing year with price by using sns.swarmplot()

• Visualizing kms_driven with price by using sns.relplot()

• Visualizing car_model with price by using sns.relplot() and suspension as 'hue' parameter

Dummy Variable

• We first create dummy variable column based on the text column.

• Then we transform it into a DataFrame.

• After that we will merge the dummies DataFrame and our orignal DataFrame.

• Finally we will drop the text column from our dataset.

Outlier Removal

• After describing the dataset I noticed that in kms_driven column, 75% of cars has travelled 85000 kms and our maximum value in kms_driven is 11 Lakh kms which is an outlier.

• And similarly, In our price column, 75% of cars has price 7 Lakh and our maximum price is 26 Lakh which is an outlier.

Model Building

• Firstly I have definied dependent and independent variables for our traning and testing.

• I have splitted data into traning and testing set by using train_test_split.

• Then I fit the model with X_train and y_train and checked the score.

• After that I used k-fold cross-validation for measuring accuracy of our model.

• So I cheked the cross_val_score for measuring the best score of our model and then I have taken mean of all the scores.

• And, Finally I predicted the result from our trained model.

Conclusion

• Developed a highly accurate Linear Regression Model using various features and attributes to predict used car prices, achieving an average prediction accuracy of 82%.

• Further model showcased its robustness by undergoing rigorous k-fold cross-validation, resulting in a mean cross-val score of 83%.

^ Scroll to Top ^