Report
Aleksandra Matlingiewicz
Daniel Imiołek
Dawid Hanke
Wojciech Hermansa




Answers

SECTION  A

2.A 150  objects
2.B 5 attributes
2.C 3 decision classes

3.Different aggregates that we can use.
For Pie chart we can use Average, Maximum,Median,Sum etc.
3A  Yes, Dataset is perfectly balanced
3B  The best for separation would be  Petal Width  and Petal Length


SECTION B

1A 41188 objects
1B 19 normal attributes and 2 special attributes
1C 2 Decision Classes
1D NO=36548  YES=4640 Dataset is not balanced

2A 3 missing values in duration
2B The identifier is ? Row number 56,84 126
3C  We replaced  ? with 258

3
The distrubution ratio is  11:89
There is 2059 objects left
ratio is still  10:90

4A Additional attribute  is outlier
 We found 10 outliers

The number of outliers is different in comparison to other section, because the other section could have other samples choosen after reduction.

0.51 0.62  sqrt((x1-x2)^2*(y1-y2)^2*(z1-z2)^2*(p1-p2)^2)

5A min 0.634 max 5.045
5B After normalization min 0 max 1

5C Every of type Real: Duration, Campaign, Pdays, Previous, Emp.var.rate, cons.price.idx, cons.conf.idx , euribor3m, nr.emplyed


6A  9 normal attributes
6B  2 attributes
6C 17  attributes, we remove them because they give us the same information
6D  nr.employed and euribor3m , We removed only 2 features so it didnt gave us much better results. Normally we would like to have 3-4  features and make work on them.

C and B were higlhy corelated

7
7A 618
7B 1441
Training Set 133:1308 8:92
Test Set 70:548 11:89
On automatic we have the same ratio as original distribution ratio, the same in stratified sample.
Default is ;