This page presents experimental results for the paper:
Fuzzy Multi-Instance Classifiers
S. Vluymans, D. Sánchez Tarragó, Y. Saeys, C. Cornelis, F. Herrera
IEEE Transactions on Fuzzy Systems 24(6), p.1395-1409, 2016

Corresponding author: sarah.vluymans@ugent.be

Results

Below, we present the accuracy and kappa values for all methods and all datasets. The datasets are grouped according to application domain: bio-informatics, textual data applications, inductive logic programming and image categorization. We present the mean values per group as well as the overall mean. For each dataset, the row-wise highest value is printed in bold. These two tables accompany Tables VIII and IX of the paper.



Accuracy values
Dataset BFMIC IFMIC MILES MIWrapper SimpleMI MILR BARTMIP miSVM CitationKNN
Musk1 81.522 82.609 88.043 86.957 73.913 85.870 84.783 78.261 86.957
Musk2 72.277 74.257 81.188 82.178 81.188 80.198 86.139 71.287 83.168
Atoms 73.936 73.936 75.000 73.936 67.553 72.872 73.936 66.489 71.277
Bonds 71.809 75.000 82.447 77.660 79.255 72.340 76.064 66.489 74.468
Chains 73.936 73.404 82.447 85.106 78.723 73.404 80.319 66.489 72.872
AntDrugs5 72.000 71.750 72.250 77.250 77.250 74.750 76.250 72.000 72.500
AntDrugs10 80.750 80.750 80.500 80.750 78.500 82.250 81.250 75.500 78.250
AntDrugs20 75.500 75.500 77.000 79.000 76.500 80.250 78.750 69.750 69.250
Mean Bio-IT 75.216 75.901 79.859 80.355 76.610 77.742 79.686 70.783 76.093
TREC1 90.750 92.500 94.500 90.750 94.250 84.750 79.000 88.250 59.250
TREC2 70.250 71.000 80.500 73.750 82.750 69.000 61.000 77.750 46.750
TREC3 77.750 81.000 75.500 84.750 88.500 79.500 63.750 76.750 49.750
TREC4 78.750 82.500 80.500 81.750 87.500 79.250 68.000 79.750 43.750
TREC7 76.000 75.000 78.500 74.000 76.250 72.000 60.500 72.750 44.500
TREC9 67.000 63.750 63.500 61.250 62.250 60.000 57.750 59.750 47.250
TREC10 77.750 75.250 77.500 80.500 75.500 72.250 64.500 72.750 47.000
WIR7 75.221 75.221 56.637 75.221 66.372 75.221 52.212 68.142 61.947
WIR8 77.876 72.566 55.752 71.681 69.027 71.681 53.097 58.407 66.372
WIR9 74.336 72.566 53.097 76.106 73.451 74.336 57.522 61.062 69.027
Mean Text 76.568 76.135 71.599 76.976 77.585 73.799 61.733 71.536 53.560
EastWest 80.000 70.000 75.000 50.000 65.000 65.000 70.000 50.000 50.000
WestEast 80.000 75.000 80.000 50.000 55.000 65.000 55.000 35.000 50.000
Mean LogProg 80.000 72.500 77.500 50.000 60.000 65.000 62.500 42.500 50.000
Elephant 84.500 84.500 78.500 84.500 76.500 79.000 81.500 78.500 50.000
Fox 62.500 59.500 58.500 56.000 63.500 57.000 59.500 50.000 50.000
Tiger 84.000 82.500 77.500 81.000 74.000 77.500 80.000 79.000 50.000
Corel1vs2 92.500 91.000 91.000 91.000 84.500 83.000 92.500 84.500 88.500
Corel1vs3 83.500 83.500 84.500 83.500 87.500 86.500 86.000 82.000 83.000
Corel1vs4 96.500 97.500 96.500 99.000 89.000 88.500 98.500 92.500 98.000
Corel1vs5 98.500 97.000 99.000 92.500 98.500 90.000 99.000 85.000 98.000
Corel2vs3 89.500 88.000 84.000 82.000 74.000 82.000 90.500 69.500 85.000
Corel2vs4 97.000 86.500 94.000 91.000 84.500 89.500 96.000 94.000 91.500
Corel2vs5 98.000 100.000 99.000 91.500 96.000 95.500 100.000 99.000 99.500
Corel3vs4 94.000 88.000 94.000 91.500 90.500 89.500 98.000 58.000 89.500
Corel3vs5 98.500 100.000 99.000 95.000 98.000 84.500 100.000 97.500 98.500
Corel4vs5 100.000 98.500 99.000 93.500 98.000 96.500 100.000 95.500 99.500
Mean Image 90.692 88.962 88.808 87.077 85.731 84.538 90.885 81.923 83.154
Mean 82.013 80.911 80.738 80.139 79.492 78.452 77.616 73.686 70.465




Kappa values
Dataset BFMIC IFMIC MILES MIWrapper SimpleMI MILR BARTMIP miSVM CitationKNN
Musk1 0.627 0.650 0.760 0.739 0.478 0.717 0.695 0.564 0.739
Musk2 0.475 0.500 0.601 0.620 0.597 0.586 0.716 0.391 0.643
Atoms 0.408 0.464 0.454 0.351 0.332 0.319 0.403 0.000 0.294
Bonds 0.260 0.462 0.617 0.499 0.567 0.256 0.448 0.000 0.389
Chains 0.334 0.292 0.614 0.671 0.511 0.341 0.531 0.000 0.374
AntDrugs5 0.437 0.432 0.444 0.543 0.545 0.493 0.524 0.437 0.449
AntDrugs10 0.602 0.602 0.605 0.603 0.564 0.635 0.615 0.499 0.556
AntDrugs20 0.493 0.493 0.535 0.572 0.528 0.599 0.568 0.381 0.375
Mean Bio-IT 0.455 0.487 0.579 0.575 0.515 0.493 0.562 0.284 0.477
TREC1 0.815 0.850 0.890 0.815 0.885 0.695 0.580 0.765 0.185
TREC2 0.405 0.420 0.610 0.475 0.655 0.380 0.220 0.555 -0.065
TREC3 0.555 0.620 0.510 0.695 0.770 0.590 0.275 0.535 -0.005
TREC4 0.575 0.650 0.610 0.635 0.750 0.585 0.360 0.595 -0.125
TREC7 0.520 0.500 0.570 0.480 0.525 0.440 0.210 0.455 -0.110
TREC9 0.340 0.275 0.270 0.225 0.245 0.200 0.155 0.195 -0.055
TREC10 0.555 0.505 0.550 0.610 0.510 0.445 0.290 0.455 -0.060
WIR7 0.506 0.508 0.128 0.503 0.328 0.505 0.052 0.368 0.232
WIR8 0.559 0.455 0.114 0.434 0.381 0.435 0.059 0.180 0.326
WIR9 0.488 0.455 0.056 0.522 0.467 0.488 0.152 0.232 0.380
Mean Text 0.532 0.524 0.431 0.539 0.552 0.476 0.235 0.434 0.070
EastWest 0.600 0.400 0.500 0.000 0.300 0.300 0.400 0.000 0.000
WestEast 0.600 0.500 0.600 0.000 0.100 0.300 0.100 -0.300 0.000
Mean LogProg 0.600 0.450 0.550 0.000 0.200 0.300 0.250 -0.150 0.000
Elephant 0.690 0.690 0.570 0.690 0.530 0.580 0.630 0.570 0.000
Fox 0.250 0.190 0.170 0.120 0.270 0.140 0.190 0.000 0.000
Tiger 0.680 0.650 0.550 0.620 0.480 0.550 0.600 0.580 0.000
Corel1vs2 0.850 0.820 0.820 0.820 0.690 0.660 0.850 0.690 0.770
Corel1vs3 0.670 0.670 0.690 0.670 0.750 0.730 0.720 0.640 0.660
Corel1vs4 0.930 0.950 0.930 0.980 0.780 0.770 0.970 0.850 0.960
Corel1vs5 0.970 0.940 0.980 0.850 0.970 0.800 0.980 0.700 0.960
Corel2vs3 0.790 0.760 0.680 0.640 0.480 0.640 0.810 0.390 0.700
Corel2vs4 0.940 0.730 0.880 0.820 0.690 0.790 0.920 0.880 0.830
Corel2vs5 0.960 1.000 0.980 0.830 0.920 0.910 1.000 0.980 0.990
Corel3vs4 0.880 0.760 0.880 0.830 0.810 0.790 0.960 0.160 0.790
Corel3vs5 0.970 1.000 0.980 0.900 0.960 0.690 1.000 0.950 0.970
Corel4vs5 1.000 0.970 0.980 0.870 0.960 0.930 1.000 0.910 0.990
Mean Image 0.814 0.779 0.776 0.742 0.715 0.691 0.818 0.638 0.663
Mean 0.628 0.611 0.610 0.595 0.586 0.554 0.545 0.443 0.398
Sarah VluymansSarah Vluymans