reports/feature_selection.md

Feature Selection — MuleGuard

• Full feature space after preprocessing: 3879 columns
• Selected for modeling: 103
• Includes the fused anomaly_score (Isolation Forest) and retained domain priors.

Top selected features by CV selection frequency

Feature	Selection freq	Domain prior?
anomaly_score	1.00	✅
F3908	1.00
F3811	1.00
F3493	1.00
F3898	1.00
F3805	1.00
F2637	1.00
F1921	0.90
F270	0.90
F3801	0.90
F2489	0.90
F3484	0.90
F3914	0.80
F3804	0.80
F3831	0.80
F1819	0.80
F3872	0.80
F3873	0.80
F1597	0.80
F1813	0.80
F3799	0.70
F3806	0.70
F3810	0.70
F3499	0.70
F3748	0.70
F3813	0.70
F3105	0.70
F114	0.70
F2486	0.70
F3825	0.70
F3907	0.70
F3922	0.70
F156	0.70
F2344	0.70
F2182	0.70
F3535	0.60
F3532	0.60
F421	0.60
F3526	0.60
F3913	0.60
F1602	0.60
F3894	0.60	✅
F1821	0.60
F3812	0.60
F3753	0.60
F3751	0.60
F2638	0.50
F949	0.50
F1814	0.50
F3055	0.50
F3537	0.50
F364	0.50
F2581	0.50
F3798	0.50
F2543	0.40
F2949	0.40
F1166	0.40
F2076	0.40
F3800	0.40
F1057	0.40
F3640	0.40
F2737	0.40	✅
F3056	0.40
F950	0.40
F3919	0.40
F1815	0.40
F64	0.40
F267	0.40
F3920	0.40
F2293	0.40
F422	0.40
F2350	0.40
F3807	0.40
F2891	0.40
F3871	0.30
F3817	0.30
F3819	0.30
F1825	0.30
F1812	0.30
F1863	0.30
F115	0.00	✅
F1692	0.00	✅
F2082	0.00	✅
F2122	0.00	✅
F2582	0.10	✅
F2678	0.10	✅
F2956	0.00	✅
F3043	0.20	✅
F321	0.20	✅
F3836	0.00	✅
F3887	0.00	✅
F3888_age_days	0.00	✅
F3889_recency_ord	0.00	✅
F3891_agriculture	0.00	✅
F3891_housewife	0.00	✅
F3891_others	0.00	✅
F3891_retired	0.00	✅
F3891_salaried	0.00	✅
F3891_selfemployed	0.00	✅
F3891_student	0.00	✅
F527	0.00	✅
F531	0.00	✅
F670	0.00	✅