TABLE IV:
Feature weighting method comparison based on MAP
| BioASQ test datasets | MAP for learning-to-rank algorithms | ||||||
|---|---|---|---|---|---|---|---|
| ARS | MART | RankBoost | AdaRank | CoordAscent | LambdaMART | RandForests | |
| year 2016, batch 1 | 0.4438 | 0.4181 | 0.3731 | 0.3792 | 0.4296 | 0.4025 | 0.4175 |
| year 2016, batch 2 | 0.4780 | 0.4625 | 0.3698 | 0.4396 | 0.4493 | 0.4497 | 0.4736 |
| year 2016, batch 3 | 0.4534 | 0.4198 | 0.3366 | 0.4009 | 0.4274 | 0.4026 | 0.4417 |
| year 2016, batch 4 | 0.4388 | 0.4036 | 0.3490 | 0.3813 | 0.4127 | 0.4022 | 0.4296 |
| year 2016, batch 5 | 0.3722 | 0.3563 | 0.2869 | 0.3263 | 0.3551 | 0.3314 | 0.3729 |
| year 2017, batch 1 | 0.4075 | 0.3843 | 0.2616 | 0.1233 | 0.3786 | 0.3517 | 0.3975 |
| year 2017, batch 2 | 0.4363 | 0.4334 | 0.3300 | 0.1457 | 0.4299 | 0.4227 | 0.4263 |
| year 2017, batch 3 | 0.4534 | 0.4377 | 0.3223 | 0.1536 | 0.4456 | 0.4105 | 0.4434 |
| year 2017, batch 4 | 0.3891 | 0.3693 | 0.2598 | 0.1193 | 0.3763 | 0.3362 | 0.3791 |
| year 2017, batch 5 | 0.2316 | 0.2068 | 0.1226 | 0.0793 | 0.2170 | 0.1887 | 0.2216 |