Table 1 The formula used in the taxonomic concept-based patient similarity
| Â | # | Formula | Reference | |
|---|---|---|---|---|
Information Content (IC) | 1 | levels(a → r) | [13] | |
2 | \( -\log \left(\frac{\frac{\left| leaves(a)\right|}{\left| subsumers(a)\right|}+1}{\left| leaves(r)\right|+1}\right) \) | [14] | ||
Code-level Similarity (CS) | 1 | \( \left\{\begin{array}{c}0,\kern0.5em if\ a=b\\ {}1,\kern0.5em otherwise\end{array}\right. \) | – | |
2 | \( 1-\frac{2 IC(c)}{IC(a)+ IC(b)} \) | |||
3 | \( 1-{e}^{\alpha \left( IC(a)+ IC(b)-2 IC(c)\right)}\bullet \frac{e^{\beta IC(c)}-{e}^{-\beta IC(c)}}{e^{\beta IC(c)}+{e}^{-\beta IC(c)}} \) | [17] | ||
4 | \( \frac{IC(l)- IC(c)}{IC(l)} \) | – | ||
Set-level similarity (SS) | 1 | Dice | \( 1-\frac{2\mid A\cap B\mid }{\mid A\mid +\mid B\mid } \) | – |
2 | Jaccard | \( 1-\frac{\mid A\cap B\mid }{\mid A\cup B\mid } \) | – | |
3 | Cosine | \( 1-\frac{\mid A\cap B\mid }{\sqrt{\mid A\mid \bullet \mid B\mid }} \) | – | |
4 | Overlap | \( 1-\frac{\mid A\cap B\mid }{\min \left\{|A|,|B|\right\}} \) | – | |
5 | \( \frac{1}{\mid A\mid +\mid B\mid}\bullet \left(\sum \limits_{a\in A}\underset{b\in B}{\min } CS\left(a,b\right)+\sum \limits_{b\in B}\underset{a\in A}{\min } CS\left(b,a\right)\right) \) | [17] | ||
6 | \( \frac{1}{\mid A\cup B\mid}\bullet \left(\sum \limits_{a\in A\setminus B}\frac{1}{\mid B\mid}\sum \limits_{b\in B} CS\left(a,b\right)+\sum \limits_{b\in B\setminus A}\frac{1}{\mid A\mid}\sum \limits_{a\in A} CS\left(b,a\right)\right) \) | [18] | ||