Prediction of Car Sales using Regression Algorithm

The Project aims at predicting the price of used cars by building a Machine Learning model using Linear Regression algorithm.

Steps to run the files:

1. Make sure to create a virtual env with the required packages mentioned here .
2. To run the python file, In the src directory run the command:

python regression_ml_pyt.py

3. To run Pytest on the test file, In the tests directory run the command:

pytest test_reg_ml.py

4. To run Pylint and verify its score, In the src directory run the command:

pylint regression_ml_pyt.py

About the Dataset:

The dataset contains around 300 entries made during the period 1992 - 2020. The various attributes present in the dataset are:

1. Car_Name
2. Year
3. Selling_Price
4. Present_Price
5. Kms_Driven
6. Fuel_Type
7. Seller_Type
8. Transmission
9. Owner

Attribute	Description
Car_Name	Contains the name/model of the car sold
Year	Contains the year in which the car was bought
Selling_Price	Contains the price the owner wished to sell the car
Present_Price	Contains the current ex-showroom price of the car
Kms_Driven	Contains the distance completed by the car in kilometers
Fuel_Type	Contains the Fuel type of the car
Seller_Type	Contains the type of seller such as a dealer or an individual
Transmission	Contains the type of transmission such as the car is manual or automatic
Owner	Contains the number of previous owners of the car

Project Implementation:

• The required python libraries for data analysis and plots such as numpy, pandas, matplotlib are imported.

• The dataset is imported using the pandas library and stored into a dataframe.

• All the features of the data are extracted and analysed.

• This data can now be visualized using bar plots, pie charts etc.

• The next step is to perform data cleansing. It is done by converting any categorical variables of string type into int values.

• The required libraries for linear regression are imported to train the model.

• The dataset now needs to be split into training and testing set, so that once the model is trained, it's functionality can be verified using the test data.

• In the project, the linear regression is computed using two methods. They are:

  • By computing the R-Squared Error Value
  • Using the Ordinary Least Squares Method(OLS)

• The project is implemented using the conda virtualenv 4.10.3 running on python 3.8.10