Data
kdd_coil_7

kdd_coil_7

active ARFF Publicly available Visibility: public Uploaded 03-10-2014 by Joaquin Vanschoren
0 likes downloaded by 0 people , 0 total downloads 0 issues 0 downvotes
Issue #Downvotes for this reason By


Loading wiki
Help us complete this description Edit
Author: Source: Unknown - Date unknown Please cite: %%%%%%%%%%%%%%%%%%% Data-Description % %%%%%%%%%%%%%%%%%%% COIL 1999 Competition Data Data Type multivariate Abstract This data set is from the 1999 Computational Intelligence and Learning (COIL) competition. The data contains measurements of river chemical concentrations and algae densities. Sources Original Owner [1]ERUDIT European Network for Fuzzy Logic and Uncertainty Modelling in Information Technology Donor Jens Strackeljan Technical University Clausthal Institute of Applied Mechanics Graupenstr. 3, 38678 Clausthal-Zellerfeld, Germany [2]tmjs@itm.tu-clausthal.de Date Donated: September 9, 1999 Data Characteristics This data comes from a water quality study where samples were taken from sites on different European rivers of a period of approximately one year. These samples were analyzed for various chemical substances including: nitrogen in the form of nitrates, nitrites and ammonia, phosphate, pH, oxygen, chloride. In parallel, algae samples were collected to determine the algae population distributions. Other Relevant Information The competition involved the prediction of algal frequency distributions on the basis of the measured concentrations of the chemical substances and the global information concerning the season when the sample was taken, the river size and its flow velocity. The competition [3]instructions contain additional information on the prediction task. Data Format There are a total of 340 examples each containing 17 values. The first 11 values of each data set are the season, the river size, the fluid velocity and 8 chemical concentrations which should be relevant for the algae population distribution. The last 8 values of each example are the distribution of different kinds of algae. These 8 kinds are only a very small part of the whole community, but for the competition we limited the number to 7. The value 0.0 means that the frequency is very low. The data set also contains some empty fields which are labeled with the string XXXXX. The training data are saved in the file: analysis.data (ASCII format). Table 1: Structure of the file analysis.data A K a g CC[1,1] CC[1,11] AG[1,1] AG[1,7] CC[200,1] CC[200,11] AG[200,1] AG[200,7] Explanation: CC[i,j]: Chemical concentration or river characteristic AG[i,j]: Algal frequency The chemical parameters are labeled as A, ..., K. The columns of the algaes are labeled as a, ..,g. Past Usage [4]The Third (1999) International COIL Competition Home Page _________________________________________________________________ [5]The UCI KDD Archive [6]Information and Computer Science [7]University of California, Irvine Irvine, CA 92697-3425 Last modified: October 13, 1999 References 1. http://www.erudit.de/ 2. mailto:tmjs@itm.tu-clausthal.de 3. file://localhost/research/ml/datasets/uci/raw/data/ucikdd/coil/instructions.txt 4. http://www.erudit.de/erudit/activities/ic-99/index.htm 5. http://kdd.ics.uci.edu/ 6. http://www.ics.uci.edu/ 7. http://www.uci.edu/ %%%%%%%%%%%%%%%%%%% Task-Description % %%%%%%%%%%%%%%%%%%% Third International Competition Protecting rivers and streams by monitoring chemical concentrations and algae communities. Intelligent Techniques for Monitoring Water Quality using chemical indicators and algae population Recent years have been characterised by increasing concern at the impact man is having on the environment. The impact on the environment of toxic waste, from a wide variety of manufacturing processes, is well known. More recently, however, it has become clear that the more subtle effects of nutrient level and chemical balance changes arising from farming land run-off and sewage water treatment also have a serious, but indirect, effect on the states of rivers, lakes and even the sea. In temperate climates across the world summers are characterized by numerous reports excessive summer algae growth resulting in poor water clarity, mass deaths of river fish from reduced oxygen levels and the closure of recreational water facilities on account of the toxic effects of this annual algal bloom. Reducing the impact of these man-made changes in river nutrient levels has stimulated much biological research with the aim of identifying the crucial chemical control variables for the biological processes. The data used in this problem comes from one such study. During the research study water quality samples were taken from sites on different European rivers of a period of approximately one year. These samples were analyzed for various chemical substances including: nitrogen in the form of nitrates, nitrites and ammonia, phosphate, pH, oxygen, chloride. In parallel, algae samples were collected to determine the algae population distributions. It is well known that the dynamics of the algae community is determined by external chemical environment with one or more factors being predominant. While the chemical analysis is cheap and easily automated, the biological part involves microscopic examination, requires trained manpower and is therefore both expensive and slow. Diatoms like Cymbella are major contributors to primary production throughout the world. The diatom reacts with large sensitivity to even small changes in acidity . Over a three and half billion year history algae have evolved and adapted as primary plant colonizers of almost every known habitant in terrestrial and aquatic environments. They respond very rapidly to man-made environment changes. The relationship between the chemical and biological features is complex and can be expected to need the application of advanced techniques. Typical of such real-life problems, the particular data set for the problem contains a mixture of (fuzzy) qualiative variables and numerical measurement values, with much of the data being incomplete. The competition task is the prediction of algal frequency distributions on the basis of the measured concentrations of the chemical substances and the global information concerning the season when the sample was taken, the river size and its flow velocity. The two last variables are given as linguistic variables. 340 data sets were taken and each contain 17 values. The first 11 values of each data set are the season, the river size, the fluid velocity and 8 chemical concentrations which should be relevant for the algae population distribution. The last 8 values of each data set are the distribution of different kinds of algae. These 8 kinds are only a very small part of the whole community, but for the competition we limited the number to 7. The value 0.0 means that the frequency is very low. The data set also contains some empty fields which are labeled with the string XXXXX. Each participant in the competition receives 200 complete data sets (training data) and 140 data sets (evaluation data) containing only the 11 values of the river descriptions and the chemical concentrations. This training data is to be used in obtainin a 'model' providing a prediction of the algal distributions associated with the evaluation data. The training data are saved in the file: analysis.txt (ASCII format). Structure of the file analysis.txt A K a g CC1,1 ... CC1,11 AG1,1 ... AG1,7 .... ... ... ... CC200,1 ... CC200,11 AG240,1 ... AG240,7 Explanation: CCi,j: Chemical concentration j=1,..11 AGi,k: Algal frequency k=1...7 The chemical parameters are labeled as A, ..., K. The columns of the algaes are labeled as a, ..,g. Evaluation data are saved in file eval.txt (ASCII format). Table 2: Structure of the file eval.* A K CC1,1 ... CC1,11 ..... ... CC140,1 ... CC140,11 _____________________________________________________________ Objective The objective of the competition is to provide a prediction model on basis of the training data. Having obtained this prediction model, each participant must provide the solution in the form of the results of applying this model to the evaluation data. The results obtained in this way should correspond to the results of the evaluation data (which are known to the organizer). The criteria used to evaluate the results is given below. All 7 Algae frequency distributions must be determined. For this purpose any number of partial models may be developed. _____________________________________________________________ Judgment of the results To judge the results, the sum of squared errors will be calculated. The following Table describes the results of a particular participant. Matrix of results a g Res1,1 ... Res1,7 .... ... Res140,1 Res140,7 All solutions that lead to a smallest total error will be regarded as winner of the contest. Information about the dataset CLASSTYPE: numeric CLASSINDEX: last ALGAE #: 7/7

12 features

algae_7 (target)numeric61 unique values
0 missing
seasonnominal4 unique values
0 missing
river_sizenominal3 unique values
0 missing
fluid_velocitynominal3 unique values
0 missing
concentration_1numeric96 unique values
2 missing
concentration_2numeric103 unique values
2 missing
concentration_3numeric272 unique values
16 missing
concentration_4numeric283 unique values
2 missing
concentration_5numeric270 unique values
2 missing
concentration_6numeric252 unique values
2 missing
concentration_7numeric286 unique values
7 missing
concentration_8numeric194 unique values
23 missing

107 properties

316
Number of instances (rows) of the dataset.
12
Number of attributes (columns) of the dataset.
0
Number of distinct values of the target attribute (if it is nominal).
56
Number of missing values in the dataset.
34
Number of instances with at least one value missing.
9
Number of numeric attributes.
3
Number of nominal attributes.
2.04
Second quartile (Median) of skewness among attributes of the numeric type.
Kappa coefficient achieved by the landmarker weka.classifiers.trees.REPTree -L 3
Kappa coefficient achieved by the landmarker weka.classifiers.trees.DecisionStump
160.45
Maximum of means among attributes of the numeric type.
Minimal mutual information between the nominal attributes and the target attribute.
18.43
Second quartile (Median) of standard deviation of attributes of the numeric type.
Area Under the ROC Curve achieved by the landmarker weka.classifiers.trees.RandomTree -depth 1
0.04
Number of attributes divided by the number of instances.
Maximum mutual information between the nominal attributes and the target attribute.
3
The minimal number of distinct values among attributes of the nominal type.
0
Percentage of binary attributes.
Third quartile of entropy among attributes.
Error rate achieved by the landmarker weka.classifiers.trees.RandomTree -depth 1
Number of attributes needed to optimally describe the class (under the assumption of independence among attributes). Equals ClassEntropy divided by MeanMutualInformation.
4
The maximum number of distinct values among attributes of the nominal type.
-0.89
Minimum skewness among attributes of the numeric type.
10.76
Percentage of instances having missing values.
11.87
Third quartile of kurtosis among attributes of the numeric type.
-1.62
Average class difference between consecutive instances.
Kappa coefficient achieved by the landmarker weka.classifiers.trees.RandomTree -depth 1
Area Under the ROC Curve achieved by the landmarker weka.classifiers.trees.J48 -C .00001
4.17
Maximum skewness among attributes of the numeric type.
0.59
Minimum standard deviation of attributes of the numeric type.
1.48
Percentage of missing values.
88.65
Third quartile of means among attributes of the numeric type.
Area Under the ROC Curve achieved by the landmarker weka.classifiers.trees.DecisionStump -E "weka.attributeSelection.CfsSubsetEval -P 1 -E 1" -S "weka.attributeSelection.BestFirst -D 1 -N 5" -W
Area Under the ROC Curve achieved by the landmarker weka.classifiers.trees.RandomTree -depth 2
Error rate achieved by the landmarker weka.classifiers.trees.J48 -C .00001
187.1
Maximum standard deviation of attributes of the numeric type.
Percentage of instances belonging to the least frequent class.
75
Percentage of numeric attributes.
Third quartile of mutual information between the nominal attributes and the target attribute.
Error rate achieved by the landmarker weka.classifiers.trees.DecisionStump -E "weka.attributeSelection.CfsSubsetEval -P 1 -E 1" -S "weka.attributeSelection.BestFirst -D 1 -N 5" -W
Error rate achieved by the landmarker weka.classifiers.trees.RandomTree -depth 2
Kappa coefficient achieved by the landmarker weka.classifiers.trees.J48 -C .00001
Average entropy of the attributes.
Number of instances belonging to the least frequent class.
25
Percentage of nominal attributes.
2.84
Third quartile of skewness among attributes of the numeric type.
Kappa coefficient achieved by the landmarker weka.classifiers.trees.DecisionStump -E "weka.attributeSelection.CfsSubsetEval -P 1 -E 1" -S "weka.attributeSelection.BestFirst -D 1 -N 5" -W
Kappa coefficient achieved by the landmarker weka.classifiers.trees.RandomTree -depth 2
Area Under the ROC Curve achieved by the landmarker weka.classifiers.trees.J48 -C .0001
6.32
Mean kurtosis among attributes of the numeric type.
Area Under the ROC Curve achieved by the landmarker weka.classifiers.bayes.NaiveBayes
First quartile of entropy among attributes.
85.41
Third quartile of standard deviation of attributes of the numeric type.
Area Under the ROC Curve achieved by the landmarker weka.classifiers.bayes.NaiveBayes -E "weka.attributeSelection.CfsSubsetEval -P 1 -E 1" -S "weka.attributeSelection.BestFirst -D 1 -N 5" -W
Area Under the ROC Curve achieved by the landmarker weka.classifiers.trees.RandomTree -depth 3
Error rate achieved by the landmarker weka.classifiers.trees.J48 -C .0001
46
Mean of means among attributes of the numeric type.
Error rate achieved by the landmarker weka.classifiers.bayes.NaiveBayes
1.27
First quartile of kurtosis among attributes of the numeric type.
Area Under the ROC Curve achieved by the landmarker weka.classifiers.trees.REPTree -L 1
Error rate achieved by the landmarker weka.classifiers.bayes.NaiveBayes -E "weka.attributeSelection.CfsSubsetEval -P 1 -E 1" -S "weka.attributeSelection.BestFirst -D 1 -N 5" -W
Error rate achieved by the landmarker weka.classifiers.trees.RandomTree -depth 3
Kappa coefficient achieved by the landmarker weka.classifiers.trees.J48 -C .0001
Average mutual information between the nominal attributes and the target attribute.
Kappa coefficient achieved by the landmarker weka.classifiers.bayes.NaiveBayes
5.49
First quartile of means among attributes of the numeric type.
First quartile of mutual information between the nominal attributes and the target attribute.
Error rate achieved by the landmarker weka.classifiers.trees.REPTree -L 1
Kappa coefficient achieved by the landmarker weka.classifiers.bayes.NaiveBayes -E "weka.attributeSelection.CfsSubsetEval -P 1 -E 1" -S "weka.attributeSelection.BestFirst -D 1 -N 5" -W
Kappa coefficient achieved by the landmarker weka.classifiers.trees.RandomTree -depth 3
Area Under the ROC Curve achieved by the landmarker weka.classifiers.trees.J48 -C .001
An estimate of the amount of irrelevant information in the attributes regarding the class. Equals (MeanAttributeEntropy - MeanMutualInformation) divided by MeanMutualInformation.
0
Number of binary attributes.
0.08
First quartile of skewness among attributes of the numeric type.
Kappa coefficient achieved by the landmarker weka.classifiers.trees.REPTree -L 1
Area Under the ROC Curve achieved by the landmarker weka.classifiers.lazy.IBk -E "weka.attributeSelection.CfsSubsetEval -P 1 -E 1" -S "weka.attributeSelection.BestFirst -D 1 -N 5" -W
0.58
Standard deviation of the number of distinct values among attributes of the nominal type.
Error rate achieved by the landmarker weka.classifiers.trees.J48 -C .001
3.33
Average number of distinct values among the attributes of the nominal type.
2.25
First quartile of standard deviation of attributes of the numeric type.
Area Under the ROC Curve achieved by the landmarker weka.classifiers.trees.REPTree -L 2
Error rate achieved by the landmarker weka.classifiers.lazy.IBk -E "weka.attributeSelection.CfsSubsetEval -P 1 -E 1" -S "weka.attributeSelection.BestFirst -D 1 -N 5" -W
Area Under the ROC Curve achieved by the landmarker weka.classifiers.lazy.IBk
Kappa coefficient achieved by the landmarker weka.classifiers.trees.J48 -C .001
1.61
Mean skewness among attributes of the numeric type.
Second quartile (Median) of entropy among attributes.
Error rate achieved by the landmarker weka.classifiers.trees.REPTree -L 2
Kappa coefficient achieved by the landmarker weka.classifiers.lazy.IBk -E "weka.attributeSelection.CfsSubsetEval -P 1 -E 1" -S "weka.attributeSelection.BestFirst -D 1 -N 5" -W
Error rate achieved by the landmarker weka.classifiers.lazy.IBk
Percentage of instances belonging to the most frequent class.
47.88
Mean standard deviation of attributes of the numeric type.
4.07
Second quartile (Median) of kurtosis among attributes of the numeric type.
Kappa coefficient achieved by the landmarker weka.classifiers.trees.REPTree -L 2
Entropy of the target attribute values.
Kappa coefficient achieved by the landmarker weka.classifiers.lazy.IBk
Number of instances belonging to the most frequent class.
Minimal entropy among attributes.
12.86
Second quartile (Median) of means among attributes of the numeric type.
Area Under the ROC Curve achieved by the landmarker weka.classifiers.trees.REPTree -L 3
Area Under the ROC Curve achieved by the landmarker weka.classifiers.trees.DecisionStump
Maximum entropy among attributes.
0.51
Minimum kurtosis among attributes of the numeric type.
Second quartile (Median) of mutual information between the nominal attributes and the target attribute.
Error rate achieved by the landmarker weka.classifiers.trees.REPTree -L 3
Error rate achieved by the landmarker weka.classifiers.trees.DecisionStump
19.01
Maximum kurtosis among attributes of the numeric type.
2.18
Minimum of means among attributes of the numeric type.

5 tasks

12 runs - estimation_procedure: 10-fold Crossvalidation - evaluation_measure: mean_absolute_error - target_feature: algae_7
0 runs - estimation_procedure: 10 times 10-fold Crossvalidation - evaluation_measure: mean_absolute_error - target_feature: algae_7
0 runs - estimation_procedure: 50 times Clustering
0 runs - estimation_procedure: 50 times Clustering
Define a new task