Car-Price-Prediction

Make a regression predictive model to predict the car price. Data source : https://www.kaggle.com/hellbuoy/car-price-prediction?select=CarPrice_Assignment.csv

Project Overview

• Seek insight from the dataset with Exploratory Data Analysis
• Performed data cleansing, data processing, data engineering to prepare data before modeling
• Built a model to predict the car price

Exploratory Data Analysis

Variable	Type	Definition
id	Numeric	Unique car ID
symbol	Categorical	Assigned insurance risk rating (+3 is risky, -3 is safe)
name	Categorical	Name of car company
fuel	Categorical	Car fuel type
aspiration	Categorical	Aspiration used in a car
door	Categorical	Number of doors
body	Categorical	Body of car
wheel	Categorical	Type of drive wheel
engineloc	Categorical	Location of car engine
wheelbase	Numeric	Wheelbase of car
length	Numeric	Length of car
width	Numeric	Width of car
height	Numeric	Height of car
weight	Numeric	Weight of a car without occupants or baggage
engine	Categorical	Type of engine
cylinder	Categorical	Cylinder number placed in the car
enginesize	Numeric	Size of car
fuelsystem	Categorical	Fuel system of car
boreratio	Numeric	Boreratio of car
stroke	Numeric	Stroke or volume inside the engine
compressionratio	Numeric	Compression ratio of car
hp	Numeric	Horsepower
peakrpm	Numeric	Car peak rpm
citympg	Numeric	Mileage in city
highwaympg	Numeric	Mileage on highway
price	Numeric	Price of car

There are outliers in wheelbase, with, enginesize, stroke, compressionration, hp, peakrpm, highwaympg, and price.
Overall the features has normal distribution.

Based on fuel, door, body, wheel, engine, cylinder features, majority is a city cars.

Data Processing

• Check missing & duplicate values
• Replace all features that have an outliers with IQR limit
• Ordinal encoding for door & cylinder features
• Feature encoding the rest of categorical columns
• Drop categorical features that encoded before
• Normalize all selected features

Modeling

• Split features & target
• Split data into data train & data test
• Train model with 4 different algorithm such as Linear Regression, Random Forest, Decision Tree, and SVR
• Evaluate model with MAE Score, RMSE Score & R2 Score
• Validate model with train and test accuracy
• Hyperparameter tuning
• Select the best model

Model Evaluation

Score	Linear Regression	Decision Tree	Random Forest	SVR
MAE	0.06	0.05	0.03	0.07
RMSE	0.09	0.07	0.05	0.09
R2	0.80	0.88	0.94	0.82
Train Acc	0.93	0.99	0.98	0.87
Test Acc	0.80	0.87	0.94	0.82

Based on predictive modeling, Random Forest algorithm has the best score compared to the other algorithms.
Random Forest make a prediction with the lowest error, MAE Score 0.03, RMSE Score 0.05, and the highest R2 Score 0.95.
The algorithm is fit based on the train & test accuracy score (0.98 - 0.94).