Car Evaluation Database
 


 

1. Sources:

M. Bohanec and V. Rajkovic: Knowledge acquisition and explanation for multi-attribute decision making. In 8th Intl Workshop on Expert Systems and their Applications, Avignon, France. pages 59-78, 1988.

2. Information:

class N                             N[%]

-------------------------------------------------

unacc                             1210 (70.023 %)

acc                                 384 (22.222 %)

good                               69 ( 3.993 %)

v-good                            65 ( 3.762 %)
 

3. Attribute values:

buying             v-high(0), high(2), med(4), low(6)

maint               v-high(0), high(2), med(4), low(6)

doors              2, 3, 4, 5-more(5)

persons           2, 4, more(6)

lug_boot         small(0), med(1), big(2)

safety             low(0), med(1), high(2)

class               unacc(0), acc(2), good(4), v-good(6)
 
 

4. Database: 1728 instances.

Car Database file has 1728 - 20 = 1708 instances. The data required for 100% accuracy is:

Class             #Required             #Actual             Accuracy
                        Instance                 Instance         Expected

0 = unacc         800                     1210                 100%

2 = acc             800                     384

4 = good          800                     69

6 = v-good       800                     65

Total                 3200                 1728

The most important feature here is to understand the DecisionMaker will be expected to predict class 0 (and only class 0) with 100% accuracy. The data can be used to estimate class 2. For class 4 and 6, the DM can be used at most for educated guess.

As we will see, the DecisionMaker 2.5 will have an accuracy rate of 100% for class 0. Below are the first 10 instances:

Car Database File:

0 0 2 2 0 0 0
0 0 2 2 0 1 0
0 0 2 2 0 2 0
0 0 2 2 1 0 0
0 0 2 2 1 1 0
0 0 2 2 1 2 0
0 0 2 2 2 0 0
0 0 2 2 2 1 0
0 0 2 2 2 2 0
0 0 2 4 0 0 0

Below are the last 20 instances, which will be used to test the DecisionMaker 's accuracy:

Question             Answer

6 6 5 2 2 1             0
6 6 5 2 2 2             0
6 6 5 4 0 0             0
6 6 5 4 0 1             2
6 6 5 4 0 2             4
6 6 5 4 1 0             0
6 6 5 4 1 1             4
6 6 5 4 1 2             6
6 6 5 4 2 0             0
6 6 5 4 2 1             4
6 6 5 4 2 2             6
6 6 5 6 0 0             0
6 6 5 6 0 1             2
6 6 5 6 0 2             4
6 6 5 6 1 0             0
6 6 5 6 1 1             4
6 6 5 6 1 2             6
6 6 5 6 2 0             0
6 6 5 6 2 1             4
6 6 5 6 2 2             6
 
 

5. Results.

Click "Integer/+ Predict" to get the following answer file:
 
 

=================== End ==========================

6 6 5 2 2 1

Possibility Confidence*Probability

0 37076

------------------------------------------------------

0

6 6 5 2 2 2

Possibility Confidence*Probability

0 35858

------------------------------------------------------

0

6 6 5 4 0 0

Possibility Confidence*Probability

0 27976

------------------------------------------------------

0

6 6 5 4 0 1

Possibility Confidence*Probability

0 18207

2 16790

------------------------------------------------------

1

6 6 5 4 0 2

Possibility Confidence*Probability

0 12485

2 8549

4 14800

------------------------------------------------------

2

6 6 5 4 1 0

Possibility Confidence*Probability

0 34997

------------------------------------------------------

0

6 6 5 4 1 1

Possibility Confidence*Probability

0 22119

2 9388

4 13312

------------------------------------------------------

2

6 6 5 4 1 2

Possibility Confidence*Probability

0 14960

2 11261

6 14772

4 4520

------------------------------------------------------

3

... ...

=================== End ==========================
 
 

6. Analysis

Remember the 4 classes and what each class stand for:

class             code
-------------------------------------------------
unacc             0
acc                2
good             4
v-good             6
 

#             Correct Answer         Predicted Answer

1                 0                             0, 100%
2                 0                             0, 100%
3                 0                             0, 100%
4                 2                             2, 48%
5                 4                             4, 41%
6                 0                             0, 100%
7                 4                             0, 49%; 4, 23%; wrong
8                 6                             0, 36%; 6, 32%; wrong

As expected, the classification for class 0 (unacceptable) is 100% accurate. Class 4 and 6 do not have enough instances (69 instances of class 4 and 65 instances of class 6), thereby severely limiting the DecisionMaker 's ability to provide you with 100% accuracy.