Scribd
Upload
338 views33 pages

Car Evaluation

This document summarizes a project evaluating cars using an artificial neural network model. It discusses how car sales data can be used to build a classification model. The dataset contains information on 1728 cars rated based on attributes like price, maintenance costs, doors, etc. The authors transform the data for use in their neural network model and test different network architectures with varying numbers of hidden layers and nodes. Their best performing model achieves an accuracy of over 80% in classifying cars after 2000 epochs of training.

Uploaded by

brayan0428
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
338 views33 pages

Car Evaluation

This document summarizes a project evaluating cars using an artificial neural network model. It discusses how car sales data can be used to build a classification model. The dataset contains information on 1728 cars rated based on attributes like price, maintenance costs, doors, etc. The authors transform the data for use in their neural network model and test different network architectures with varying numbers of hidden layers and nodes. Their best performing model achieves an accuracy of over 80% in classifying cars after 2000 epochs of training.

Uploaded by

brayan0428
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 33

Car Evaluation

Shivam Porwal(1611CS14)
Debanjan Sarkar(1611CS17)
Background and Motivation

• We all Like Cars.


• Car Sell:
• Around 3,50,000 cars are sold in India every year.
• 5,15,000 cars are sold in Taiwan, 2005(which was highest in 10 years).
Introduction of Dataset

• Car Evaluation Database:


• UCI Machine Learning Depository
• https://archive.ics.uci.edu/ml/datasets/Car+Evaluation.
• Model Evaluates the cars based on certain characteristics.
• Derived from a Simple Hierarchical decision model.
• It is a Multiclass Classification Problem.
How Much Data is Enough ?..
How Much Data is Enough ?...

• We try to Discover the useful knowledge from volumes of data.


How Much Data is Enough ?...

• We try to Discover the useful knowledge from volumes of data.


• It Means larger data we use, better result we can get.
How Much Data is Enough ?...

• We try to Discover the useful knowledge from volumes of data.


• It Means larger data we use, better result we can get.
• But some scholar think that great deal of data don't guarantee
better result than little of data.
How Much Data is Enough ?...

• We try to Discover the useful knowledge from volumes of data.


• It Means larger data we use, better result we can get.
• But some scholar think that great deal of data don't guarantee better
result than little of data.
• Because resources are limited, the large samples will result in much load
and contain lots of exceptional cases.
DataSet Information

• Dataset consist of 1728 instances.


• Each record consist of 6 attributes:
• Buying {very high, high, med, low}
• MAINT {very high, high, med, low}
• Doors {2,3,4,5-more}
• Person {2,4,more}
• LUG_BOOT {small, med, big}
• Safety {low, Med, High}
• Class {unacc, acc, good, very-good}
Know the data

Car
Know the data

Car

pric safe
e ty
com
fort
able
Know the data

Car

pric safe
e ty
com
fort
able
b_pr m_p
ice safe
rice
ty
door pers
size
s ons
Approach

• Solution with neural network.


• As we cannot work with strings in neural networks, so we map the
attribute values and classes to binary numbers.
Transformed Dataset

buying: 1,0,0,0 instead of vhigh, 0,1,0,0 instead of high, 0,0,1,0 instead of


med, 0,0,0,1 instead of low.
maint: 1,0,0,0 instead of vhigh, 0,1,0,0 instead of high, 0,0,1,0 instead of
med, 0,0,0,1 instead of low.
doors: 0,0,0,1 instead of 2, 0,0,1,0 instead of 3, 0,1,0,0 instead of 4,
1,0,0,0 instead of 5more.
persons: 0,0,1 instead of 2, 0,1,0 instead of 4, 1,0,0 instead of more.
lug_boot: 0,0,1 instead of small, 0,1,0 instead of med, 1,0,0 instead of big.
safety: 0,0,1 instead of low, 0,1,0 instead of med, 1,0,0 instead of high.
Implementation

• Language used : Python


• Libraries used : Numpy, Keras
• Backend : Theano
• Type of Model : MLP(Multi-Layer Perceptron)
Network Architecture
Literature Survey

Classification Accuracy for Decision Trees


Literature Survey

Classification Accuracy for Naive Bayesian


Literature Survey

Classification Accuracy for Artificial Neural Network (ANN)


Considerations:

• CASE : 1
• Input Neuron : 21
• Number of Hidden layers:2
• Hidden Neurons in first layer:100
• Hidden Neurons in second layer:80
• Output Neurons : 4
• Softmax function is used as activation function in both Hidden Layer as well
as output layer.
• Learning Rate is considered to be : 0.01.
• Momentum :0.7
• Number of Epochs: 50
Results:

Fig(a) : Output with Accuracy and MSE


Graphical Representation :

Fig(b) : Accuracy Plot Fig(c) : Error Plot


Considerations:

• CASE : 2
• Input Neuron : 21
• Number of Hidden layers:3
• Neurons at hidden Layer 1: 100
• Neurons at hidden Layer 2: 80
• Neurons at hidden Layer 3:70
• Output Neurons : 4
• Softmax function is used as activation function in both Hidden Layer as well
as output layer.
• Learning Rate is considered to be : 0.01.
• Momentum :0.7
• Number of Epochs: 50
Results:

Fig(d) : Output with Accuracy and MSE


Graphical Representation :

Fig(e) : Accuracy Plot Fig(f) : Error Plot


Considerations:

• CASE : 3
• Input Neuron : 21
• Number of Hidden layers:4
• Neurons at hidden Layer 1:100
• Neurons at hidden Layer 2: 80
• Neurons at hidden Layer 3: 70
• Neurons at hidden Layer 4:60
• Output Neurons : 4
• Softmax function is used as activation function in both Hidden Layer as well as
output layer.
• Learning Rate is considered to be : 0.01.
• Momentum :0.7
• Number of Epochs: 50
Results:

Fig(f) : Output with Accuracy and MSE


Graphical Representation :

Fig(h) : Accuracy Plot Fig(i) : Error Plot


Considerations:

• CASE : 4
• Input Neuron : 21
• Number of Hidden layers:4
• Neurons at hidden Layer 1:100
• Neurons at hidden Layer 2: 80
• Neurons at hidden Layer 3: 70
• Neurons at hidden Layer 4 : 60
• Output Neurons : 4
• Softmax function is used as activation function in both Hidden Layer as well as
output layer.
• Learning Rate is considered to be : 0.01.
• Momentum :0.7
• Number of Epochs: 2000
Results:

Fig(f) : Output with Accuracy and MSE


Graphical Representation :

Fig(h) : Accuracy Plot Fig(i) : Error Plot


Future Work:

• We have to add more data items to get more accurate results.


• Need more Optimization Techniques.
• Can use more complex networks.
References:

• Knowledge acquisition and explanation for multi-attribute decision


making by M Bohanec, V Rajkovic.
• Machine Learning by Function Decomposition Blaz Zupan, Marko
Bohanec, Ivan Bratko, Janez Demsar.
• https://archive.ics.uci.edu/ml/datasets/Car+Evaluation.
• http://neuroph.sourceforge.net/tutorials/carevaluation1/carevaluation1
.html

You might also like