20.500.12556/RUL-183212 Active feature acquisition by prediction explanations Aktivno pridobivanje znacilk z razlago napovedi V mnogih prakticnih aplikacijah strojnega u ˇ cenja, npr. v ˇ medicinski diagnostiki, je pridobivanje dodatnih znacilk lahko ˇ drag in zamuden postopek. Aktivno pridobivanje znacilk ˇ (AFA) naslavlja ta problem, a enostavne strategije pogosto uporabljajo staticne, globalne metrike, ki so neodvisne od ˇ konteksta posameznega primera in zato suboptimalne. V clanku predlagamo novo strategijo, ki jo vodijo razlage ˇ SHAP, personalizirane za posamezen primer. Izvedli smo sistematicno primerjavo strategij za aktivno pridobivanje ˇ znacilk (naklju ˇ cna, stati ˇ cna in na razlagah osnovana) na ˇ desetih podatkovnih zbirkah, z mehanizmoma manjkajocih ˇ podatkov MCAR in MNAR. Empiricni rezultati ka ˇ zejo, da je ˇ strategija, osnovana na razlagah, znacilno bolj ˇ sa od stati ˇ cnih ˇ metod, zlasti v kompleksnih in realisticnih pogojih MNAR. ˇ Personalizirane razlage predstavljajo mocan in prilagodljiv ˇ nacin za u ˇ cinkovito pridobivanje dodatnih zna ˇ cilk. In many real-world machine learning applications, particularly in resource-constrained domains such as medical diagnostics, acquiring feature values is a costly and often sequential process. Active Feature Acquisition (AFA) addresses this by selecting the most informative subset of features to acquire for a given instance, balancing predictive accuracy against acquisition costs. Conventional AFA strategies often rely on static, global feature importance metrics, which are instance-agnostic and can be suboptimal when feature relevance is context-dependent. We investigate a practically relevant two-stage acquisition scenario, where an initial subset of features is observed, and a subsequent non-overlapping subset is selected for acquisition. We propose a novel AFA strategy that leverages instance-specific model explanations, specifically Shapley additive explanations (SHAP), to guide the selection process. We conduct a systematic comparative study, evaluating three distinct acquisition strategies: random acquisition (as a baseline), acquisition guided by static global feature importance lists, and the proposed SHAP-based approach. Using XGBoost-trained models, these methods are evaluated across ten benchmark datasets under two missingness scenarios: Missing Completely at Random (MCAR) and Missing Not at Random (MNAR). The empirical results demonstrate that the SHAP-based strategy significantly outperforms static global feature importance methods in complex settings, particularly under MNAR missingness, where the context of observed features is critical for effective decision-making. While it also performs strongly in simpler MCAR scenarios (especially at high missingness rates), its robustness in more realistic settings suggests that utilizing instance-specific explanations provides a powerful and adaptive mechanism for personalized and effective feature acquisition. aktivno pridobivanje značilk Shapleyeve razlage manjkajoči podatki strojno učenje active feature acquisition Shapley additive explanations missing data machine learning true false true Slovenski jezik Angleški jezik Neznano 2026-06-08 12:43:39 2026-06-08 12:43:46 2026-06-09 14:07:46 0000-00-00 00:00:00 2026 0 0 Str. 14-31 št. 1-2 Letn. 93 2026 0000-00-00 Zaloznikova Objavljeno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 004.852 0013-5852 280806915 280806915_Active_feature_acqusition_by_prediction_explanations_Matjaz_Kukar.pdf 280806915_Active_feature_acqusition_by_prediction_explanations_Matjaz_Kukar.pdf 1 E5EEACC9627EBD92310E41428FE2EFE0 71230310236286c3aef71c0d8dd136ffb855c576b13947e893f048f4c9fa52bd eadd1efd-632d-11f1-9b0d-0050569b8976 https://repozitorij.uni-lj.si/Dokument.php?lang=slv&id=235099 https://ev.fe.uni-lj.si/1-2-2026/Kukar.pdf 0 c58be34e-63fb-11f1-9b0d-0050569b8976 https://repozitorij.uni-lj.si/Dokument.php?lang=slv&id=235208 https://doi.org/10.5281/zenodo.18879292 1 edc0e30d-6326-11f1-9b0d-0050569b8976 https://repozitorij.uni-lj.si/Dokument.php?lang=slv&id=235095 Fakulteta za elektrotehniko Fakulteta za računalništvo in informatiko Elektrotehniški vestnik 0 0 0