source: orange/orange/doc/datasets/bridges.htm @ 1760:9d4bb141fb0e

Revision 1760:9d4bb141fb0e, 4.5 KB checked in by blaz <blaz.zupan@…>, 9 years ago (diff)

data info file

Line 
1<html>
2<head>
3<title>Pittsburgh Bridges Data Base</title>
4</head>
5<body>
6<h1>Info on Pittsburgh Bridges Data Base</h1>
7<pre>
81. Title: Pittsburgh bridges
9
102. Sources:
11   -- Yoram Reich & Steven J. Fenves
12      Department of Civil Engineering
13      and
14      Engineering Design Research Center
15      Carnegie Mellon University
16      Pittsburgh, PA 15213
17
18      Compiled from various sources.
19
20   -- Donor: Yoram Reich (yoram.reich@cs.cmu.edu)
21   -- Date: 1 August 1990
22
233. Past Usage:
24
25   -- Reich & Fenves (1989). Incremental Learning for Capturing Design
26      Expertise. Technical Report: EDRC 12-34-89, Engineering Design
27      Research Center, Carnegie Mellon University, Pittsburgh, PA.
28      -- Qualitative results and runs with original ordering of examples.
29         using COBWEB.
30
31   -- Reich (1989). Converging to ``Ideal'' Design Knowledge by Learning,
32      Proceedings of the First International Workshop on Formal Methods in
33      Engineering Design, pp: 330-349, Colorado Springs, CO, January 1990.
34      -- Describes a new design method with Bridger (variant of COBWEB) using
35     this domain. (Also an EDRC report: 12-35-89)
36
37   -- Reich (1989) Combining Nominal and Continuous Properties in an
38      Incremental Learning System for Design. Technical Report: EDRC 12-33-89.
39      -- Comparison of performance of Bridger when running on both versions
40     (V1 and V2) of the database
41
42   -- Reich (1989) Incremental Concept Formation with Mixed Property Types
43      Unpublished Manuscript.
44      -- Results using 10 random 10-fold cross-validation test with Bridger
45     (relative error rate):
46     Version V1 of the database:
47     MATERIAL 18.4%, REL-L 38.7%, SPAN 42.7%, T-OR-D 14.7%, TYPE 47.6%.
48     Version V2 of the database:
49     MATERIAL 24.2%, REL-L 41.7%, SPAN 39.9%, T-OR-D 14.7%, TYPE 56.5%.
50
51   -- Quinlan (1989) Personal communication.
52      -- Results of a 10-fold cross-validation test with C4.5, and with
53         a separate decision tree for each design property obtained the
54     following error rates on version V1 of the database:
55     MATERIAL 15%, REL-L 32%, SPAN 32%, T-OR-D 15%, TYPE 44%.
56
574. Number of instances: 108
58
595. Relevant Information:
60   
61   There are two versions to the database:
62      V1 contains the original examples and
63      V2 contains descriptions after discretizing numeric properties.
64
65   There are no ``classes'' in the domain. Rather this is a DESIGN domain where
66   5 properties (design description) need to be predicted based on 7
67   specification properties.
68
696. Number of Attributes: 13: 7 specifications, 5 design description, and 1
70   identifier (not used for the classification)
71
727. Attribute Information: 
73
74   The type field state whether a property is continuous/integer (c)
75                              or nominal (n).
76   For properties with c,n type, the range of continuous numbers is given
77   first and the possible values of the nominal follow the semi-colon.
78
79
80       name     type    possible values     comments
81   ------------------------------------------------------------------------
82   1.  IDENTIF  -   -           identifier of the examples
83   2.  RIVER    n   A, M, O
84   3.  LOCATION n       1 to 52
85   4.  ERECTED  c,n 1818-1986 ; CRAFTS, EMERGING, MATURE, MODERN
86   5.  PURPOSE  n   WALK, AQUEDUCT, RR, HIGHWAY
87   6.  LENGTH   c,n 804-4558 ; SHORT, MEDIUM, LONG
88   7.  LANES    c,n 1, 2, 4, 6 ; 1, 2, 4, 6
89   8.  CLEAR-G  n   N, G
90   9.  T-OR-D   n   THROUGH, DECK
91   10. MATERIAL n   WOOD, IRON, STEEL
92   11. SPAN n   SHORT, MEDUIM, LONG
93   12. REL-L    n   S, S-F, F
94   13. TYPE n   WOOD, SUSPEN, SIMPLE-T, ARCH, CANTILEV, CONT-T
95
96
978. More complicated attributes:
98
99   One can use a hierarchical structure for the Type property. There are two
100   options.
101
102    option 1 (use examples without modification)
103        --------
104
105            Type
106         /      /        \     \
107       /       /          \      \
108    wood    suspen      arch    truss
109                       /  |    \
110                     /    |      \
111                cantilev  cont-t   simple
112
113
114    option 2 (requires changes in the Type property - specified bellow)
115    --------
116
117                Type
118
119         /      /        |              \   
120        /     /          |               \
121        wood   suspen   arch              truss
122                / \            /  |  \    \
123                  /     \        /    |   \ \
124            tied-a    not-tied  cantilev cont-t simple arch-t
125
126
127    Change the Type  property of the following examples (in both V1 and V2):
128        E28   ->  arch-t
129        E91,E90,E84,E83,E73  -> tied-a
130        E97,E78,E77,E75,E66,E64,E43  -> not-tied
131
132
1339. Missing Attribute Values:
134   Attribute #:     # instances with missing values:
135    2            1
136    6           27
137    7           16
138    8            2
139    9            6
140   10            2
141   11           16
142   12            5
143   13            3
144<pre>
145</body>
146</html>
Note: See TracBrowser for help on using the repository browser.