You are on page 1of 6

PREDICTING-USED-CAR-

PRICES-BASED-ON-
CURRENT-CAR-STATUS
Blueprint of Project

V. Subbarao
ML Intern-Technocolabs Software
 The used car market is quite active in USA. Therefore
predicting car prices is highly variable. In market, it can be
difficult to make a decision whether a used car is worth the
posted price or not when we see listings online. Several
features of cars like mileage, make, model, year, etc. can
influence the actual price of a car.
 The main focus area of this project is by using the existing
current data we are going to develop machine learning
models which can fit & accurately predict the price of a used
INTRODUCTI
car based on its features, in order to make informed better
purchases of cars
ON
 We plan to utilize Deep Neural Networks to improve our
prediction performance while avoiding overfitting, along with
we implement and evaluate various learning methods on a
dataset consisting of the sale prices of different makes and
models across cities in the United States.
 Features of data are key parameters for prediction models, In
this project, we are going use the dataset of used cars sales &
with different parameters from all over the United States
related to company called True Car which has 852123 data
DATASET
variables from 2010 to 2017.
 Features of dataset
&
 1. Mileage FEATURE
S
 2. Make
 3. Model
 4. Year
 5.State
 6.City
MODEL
 We can utilize several classic and state-of-the-art methods,
including ensemble learning techniques, with a 9O-1O split
on test and training data
 Linear Regression
 2.Gradient Boost IMPLE
 3.Random Forest
 4.XGBoost
MENTA
 5.KMeans + Linear Regression Ensemble:
 6.Light GBM
TION
 Finally at the end, we use F1 score as the metric to compare
the performance of these models.
 Yet to be done
RESULTS
CONCLUSI
 Yet to be done ON
&
FUTURE

WORK

You might also like