abhinavxox / car-price-prediction Goto Github PK

CAR PRICE PREDICTION

B.Tech. Computer Science and Engineering

Artificial Intelligence

Submitted by:

July 2021

DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING AMRITA

VISHWA VIDYAPEETHAM

(Estd. U/S 3 of the UGC Act 1956)

Amritapuri Campus Kollam

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING AMRITA VISHWA

VIDYAPEETHAM

(Estd. U/S 3 of the UGC Act 1956)

Amritapuri Campus Kollam

DECLARATION

We, Abhinav Pandey, Abishek A., Anfas Hassan, Aravind MJ, Rithesh R, D.S.S.

Sandeep Chandra hereby declare that this project entitled CAR PRICE PREDICTION is

a record of the original work done by us under the guidance of DR GEORG GUTJAHR

and DR GOPAKUMAR G , Dept. of Computer Science andEngineering, Amrita Vishwa

Vidyapeetham, that this work has not formed the basis for any

degree/diploma/associations/fellowship or similar awards to any candidate in any university

to the best of our knowledge.

Place: Amritapuri Date: 12 July 2022

Signature of the student Signature of the Project Guide

ACKNOWLEDGEMENTS

Humble pranams at the lotus feet of Amma, Sri Mata

Amritanandamayi devi

We thank Dr Georg Gutjahr, Department of Computer Science and Engineering, Amrita

Vishwa Vidyapeetham, for her timely advice and constructive suggestions. We thank the

faculty of the CSE Department for giving us an opportunity to explore our capabilities and

widen our knowledge. We also thank our team members for their cooperation and hard

work put into this project. We also thank our batch mates for their support and our parents

for their support and encouragement.

INTRODUCTION

An automobile company is entering the market and it is planning to set up a

manufacturing unit here. It also plans to produce cars locally to give competition to

existing companies. They have planned to do a contract with an automobile consulting

company to understand the factors on which the pricing of the cars depends.

Specifically, they want to understand the main features about a car that has more effect

on its pricing.

The company wants to know:

● Which variables are significant in predicting the price of a car

● How well those variables describe the price of a car

In our project we have planned to solve the same problem. We want to model the price

of cars with the available independent variables. We have split every model into a

training and testing set. We have used independent variables in the training set to

predict the price and correspondingly check the accuracy of the models with the actual

price in the testing set. We have made 12 different models and fit single as well as

multi variables in that model according to analysis. We have used many python

libraries such as sklearn, matplot, seaborn, etc.

This provides a solution for the company to manage and understand the variation of

prices with the variables. They can accordingly manipulate the design of the cars, the

business strategy etc to meet certain price levels. Further, the model will be a good

way for management to understand the pricing dynamics of a new market.

DATA SET

Our dataset consists of 205 rows and 26 columns. The rows include the no of data per

car. The column includes independent variables or features of the car. The variables

are:

METHODOLOGY

Data is collected from a dataset from Kaggle. It consists of 205 rows each of the car

and its information. It includes columns such as fueltype, horsepower, cylinder

number etc. But this data was not ready to be used. We have to filter and process the

data before we put regression on it. We first understood the structure and features of

the data.

Firstly, the company names of the cars were very vast. So, we have filtered the names

of cars with the actual names of the company. For instance,alfa-romeo giulia and

alfa-romeo stelvio arecars from same company but different name. We have only

considered the company name like alfa romeo.

Second task was to remove any errors. The company names were misspelt which

made them different data. So, we corrected some names as show:

Third task was to remove the duplicates from the dataset if there were any. Fourth and

main task was to encode our data. We have a lot of strings in our data, but our

regression model cannot take strings as an input. So, we used labelEncoder to

transform string into int value according to the set. Now our data is filtered and ready

to be used.

We move on to data analysis and visualisation. We crossed checked every independent

variable and how it varies in the data set with the price. We have visualised the spread

of the variable with the price and accordingly selected the variable for the model.

● The plot seemed to be right-skewed, meaning that the most prices in

the dataset are low(Below 15,000).

● There is a significant difference between the mean and the median of

the price distribution.

● The data points are far spread out from the mean, which indicates a

high variance in the car prices.(85% of the prices are below 18,500,

whereas the remaining 15% are between 18,500 and 45,400.

In the histogram bar graph, we can see that there is much variation in the company

name but the fuel type histogram is not useful as there are only two types petrol and

diesel.

Here we can analyse how much the different types of engines are varying the prices.

Third engine type is the most used of all the data. But the fifth is the one which has

highest price range and others have low price range.

Door-number variable is not affecting the price much. There is no significant

difference between the categories in it. It seems aspiration with turbo has a higher

price range than the std(though it has some high values outside the whiskers)

carwidth, carlength and curbweight seems to have a positive correlation with

price. carheight doesn't show any significant trend with price.

Now that we’ve analysed the data, we have to fit the data in the models. We’ve

used 12 models. 6 of them are of single variables like highwaympg, citympg,

horsepower, enginesize, wheelbase and boreratio. Some showed less accuracy, so

we are going to further use the ones which show more accuracy and less MSE.

Then there are 5 models which are made to test combined variables.

We have used the sklearn python library to split the testing and training set. There

is a separate table for actual price for testing of the models. We first define the

model, put our variable or combined variables in it and then determine the size of

the testing set. We have also imported LinearRegression from sklearn and used

that to predict the prices from the model. Then we test it in the testing set. We

calculate the mean squared error of the predicted and actual prices of cars. We

also determine the accuracy of the model.

Example:

Model(Prediction using enginesize):

RESULTS

● Prediction using highway mpg

● Prediction using citympg

● Prediction using horsepower

● Prediction using enginesize

● Prediction using wheelbase

● Prediction using boreratio

● Prediction using enginesize and boreratio

● Prediction using enginesize and horsepower

● Prediction using stroke, horsepower, enginesize

● Prediction using boreratio, horsepower and enginesize

● Prediction using boreratio, horsepower, enginesize and stroke

CONCLUSION

Car price prediction can be a challenging task due to the high number of attributes

that should be considered for the accurate prediction. The major step in the

prediction process is collection and preprocessing of the data. In this research,

PHP scripts were built to normalise, standardise and clean data to avoid

unnecessary noise for linear regression.Then after analysing the fitting the data in

the models, we used it in the testing and training sets and analysed which

performed better.

REFERENCES