Interactive key Comparison Paper A Demo — P6 Analysis: Identification key optimization

Optimization efficiency

Side-by-side comparison: greedy information-gain optimized order vs. random trait order. Both use 20 of 157 traits to identify 478 species.

% % unique species

1
0.0%
1
0.0%
2
0.0%
2
0.0%
3
0.0%
3
0.0%
4
0.0%
4
0.4%
5
0.0%
5
1.1%
6
0.0%
6
1.9%
7
4.6%
7
3.4%
8
17.6%
8
4.4%
9
29.3%
9
4.6%
10
50.0%
10
6.3%
11
60.7%
11
8.6%
12
73.6%
12
10.0%
13
78.5%
13
11.9%
14
83.9%
14
15.9%
15
87.5%
15
23.0%
16
90.2%
16
31.6%
17
92.1%
17
37.7%
18
93.5%
18
42.9%
19
94.4%
19
49.6%
20
95.4%
20
50.6%
Optimized order (greedy) Random order

Optimized order (greedy)

# Trait IG % unique species Largest group
1 Crop fields, roadsides, disturbed places 1.000 0.0% 239
2 Fulla de 1 a 5cm 0.996 0.0% 130
3 Leaf without petiole 0.991 0.0% 78
4 March flowering 0.980 0.0% 44
5 Glabrous plant 0.953 0.0% 29
6 Regular flower 0.920 0.0% 18
7 Indehiscent fruit 0.834 4.6% 14
8 Yellow or cream flower 0.646 17.6% 11
9 Leaf 5 to 10cm long 0.496 29.3% 7
10 July flowering 0.379 50.0% 6
11 Petal size 5 to 10mm 0.209 60.7% 5
12 Divided leaf or leaflet margin 0.160 73.6% 5
13 Fused sepals or tepals 0.086 78.5% 5
14 White flower 0.062 83.9% 5
15 Oval to obovate leaf 0.040 87.5% 5
16 May flowering 0.031 90.2% 5
17 February flowering 0.024 92.1% 4
18 Rocky coast 0.019 93.5% 3
19 Glandular plant 0.016 94.4% 3
20 Pink or purple flower 0.014 95.4% 2

Random order

# Trait % unique species Largest group
1 Petal size 10 to 20mm 0.0% 337
2 Alternate leaves 0.0% 215
3 Non-green plant 0.0% 212
4 Petals with toothed apex 0.4% 190
5 Fused sepals or tepals 1.1% 104
6 Numerous petals or tepals 1.9% 104
7 October flowering 3.4% 78
8 Plant with bulb, corm or tuber 4.4% 73
9 Tree 4.6% 63
10 Glandular plant 6.3% 62
11 Herbaceous plant 8.6% 42
12 Silicicolous therophytic grasslands 10.0% 41
13 Parallel venation 11.9% 41
14 Rocky coast 15.9% 38
15 Yellow or cream flower 23.0% 20
16 Petal size more than 20mm 31.6% 19
17 Indehiscent fruit 37.7% 15
18 Deeply lobed leaf 42.9% 15
19 Regular flower 49.6% 13
20 Lobed or laciniate petals 50.6% 13

Key findings

Metric Optimized order (greedy) Random order
Steps to reach 50% 10 20
Steps to reach 90% 16 -
Final % unique (20 steps) 95.4% 50.6%
Advantage +44.8 pp