Promise 2009


ARFF Format

The recommended format for public domain data sets is the Attribute-Relation File Format (ARFF).

Other Formats

If ARFF is not an appropriate format for your data, please provide detailed description of your data format in the paper.

Data needs comments

Regardless of the format, the data must be commented. If you are using an alternate format that does not support comments in the dataset, provide this information in a separate file with extensions .desc, and submit the URL of this file. The commenting guidelines below are from ftp://ftp.ics.uci.edu/pub/machine-learning-databases/DOC-REQUIREMENTS

On Documenting Databases: Optimal Information

1. Title of Database: Indicate the central topic of the domain
2. Sources:
(a) Original owners of database (name/phone/snail address/email address)
(b) Donor of database (name/phone/snail address/email address)
(c) Date received (databases may change over time without name change!)
3. Past Usage:
(a) Complete reference of article where it was described/used
(b) Indication of what attribute(s) were being predicted
(b) Indication of study's results (i.e. Is it a good domain to use?)
4. Relevant Information Paragraph:
-- whatever is not covered elsewhere that should be stated
5. Number of Instances
6. Number of Attributes
-- if this varies (is "non-standard"), give details
-- for example, suppose that only some (not all) features are present
for each instance: then carefully note how the missing values should
be interpreted (i.e., as unknown information, don't cares, negatives,
or whatever)
7. For Each Attribute: (please give both acronym and full name if both exist)
(a) Type of domain:
(i) either numeric or non-numeric
(ii) if numeric, note whether continuous, integers only, etcetera
(iii) if non-numeric, label as "linear", "structured", or "nominal"
Be careful to distinguish numeric values from symbolic-valued attributes
that happen to be encoded numerically!
(b) Statistics for numeric domains:
-- Min, Max, Mean, SD, Correlation with predicted attribute
(c) Statistics for non-numeric domains
-- where possible, list all attribute values that occur
8. Missing Attribute Values: how many per each attribute?
9. Class Distribution: number of instances per class

Please put the documentation before the data.
Please leave the data in the following format:
(a) only one instance per line
(b) place commas between attribute values
(c) missing values should be denoted by "?"
(d) no spaces should occur between attribute values
(e) please state any exceptions to these format instructions

Example Documentation: --------------------------------------------------

1. Title: Relative CPU Performance Data

2. Source Information
-- Creators: Phillip Ein-Dor and Jacob Feldmesser
-- Ein-Dor: Faculty of Management; Tel Aviv University; Ramat-Aviv;
Tel Aviv, 69978; Israel
-- Donor: David W. Aha (aha@ics.uci.edu) (714) 856-8779
-- Date: October, 1987

3. Past Usage:
1. Ein-Dor and Feldmesser (CACM 4/87, pp 308-317)
-- Results:
-- linear regression prediction of relative cpu performance
-- Recorded 34% average deviation from actual values
2. Kibler,D. & Aha,D. (1988). Instance-Based Prediction of
Real-Valued Attributes. In Proceedings of the CSCSI (Canadian
AI) Conference.
-- Results:
-- instance-based prediction of relative cpu performance
-- similar results; no transformations required
- Predicted attribute: cpu relative performance (numeric)

4. Relevant Information:
-- The estimated relative performance values were estimated by the authors
using a linear regression method. See their article (pp 308-313) for
more details on how the relative performance values were set.

5. Number of Instances: 209

6. Number of Attributes: 10 (6 predictive attributes, 2 non-predictive,
1 goal field, and the linear regression's guess)

7. Attribute Information:
1. vendor name: 30
(adviser, amdahl,apollo, basf, bti, burroughs, c.r.d, cambex, cdc, dec,
dg, formation, four-phase, gould, honeywell, hp, ibm, ipl, magnuson,
microdata, nas, ncr, nixdorf, perkin-elmer, prime, siemens, sperry,
sratus, wang)
2. Model Name: many unique symbols
3. MYCT: machine cycle time in nanoseconds (integer)
4. MMIN: minimum main memory in kilobytes (integer)
5. MMAX: maximum main memory in kilobytes (integer)
6. CACH: cache memory in kilobytes (integer)
7. CHMIN: minimum channels in units (integer)
8. CHMAX: maximum channels in units (integer)
9. PRP: published relative performance (integer)
10. ERP: estimated relative performance from the original article (integer)

8. Missing Attribute Values: None

9. Class Distribution: the class value (PRP) is continuously valued.
PRP Value Range: Number of Instances in Range:
0-20 31
21-100 121
101-200 27
201-300 13
301-400 7
401-500 4
501-600 2
above 600 4

Summary Statistics:
Min Max Mean SD PRP Correlation
MCYT: 17 1500 203.8 260.3 -0.3071
MMIN: 64 32000 2868.0 3878.7 0.7949
MMAX: 64 64000 11796.1 11726.6 0.8630
CACH: 0 256 25.2 40.6 0.6626
CHMIN: 0 52 4.7 6.8 0.6089
CHMAX: 0 176 18.2 26.0 0.6052
PRP: 6 1150 105.6 160.8 1.0000
ERP: 15 1238 99.3 154.8 0.9665

Promise 2009

Promise Deadlines