File size: 11,098 Bytes
7302617
 
 
 
 
 
 
 
 
 
 
 
eb4abb8
7302617
 
 
 
 
 
 
 
 
 
eb4abb8
 
7302617
eb4abb8
7302617
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
eb4abb8
 
 
 
 
 
 
 
 
7302617
eb4abb8
 
 
7302617
 
eb4abb8
7302617
 
 
 
 
 
 
 
 
 
 
 
 
eb4abb8
7302617
 
 
 
 
 
 
eb4abb8
7302617
 
 
 
 
 
 
 
 
eb4abb8
7302617
 
 
 
 
 
b4db9bd
 
 
 
 
 
 
 
eb4abb8
b4db9bd
 
 
eb4abb8
b4db9bd
 
 
 
 
 
 
 
 
 
 
eb4abb8
b4db9bd
 
 
 
 
 
 
 
eb4abb8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7302617
eb4abb8
7302617
eb4abb8
 
7302617
 
 
 
 
 
 
eb4abb8
 
 
 
 
 
7302617
eb4abb8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b4db9bd
7302617
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
eb4abb8
 
 
7302617
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
eb4abb8
 
 
 
 
 
 
 
 
 
 
 
 
7302617
eb4abb8
7302617
eb4abb8
 
 
 
 
 
 
 
 
7302617
eb4abb8
 
 
 
7302617
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
"""
Feature importance and leave-one-feature-out ablation for the 10 face_orientation features.
Run: python -m evaluation.feature_importance

Outputs:
- XGBoost gain-based importance (from trained checkpoint)
- Leave-one-feature-out LOPO F1 (ablation): drop each feature in turn, report mean LOPO F1.
- Writes evaluation/feature_selection_justification.md
"""

import os
import sys
import argparse

import numpy as np
from sklearn.preprocessing import StandardScaler
from sklearn.metrics import f1_score
from xgboost import XGBClassifier

_PROJECT_ROOT = os.path.abspath(os.path.join(os.path.dirname(__file__), ".."))
if _PROJECT_ROOT not in sys.path:
    sys.path.insert(0, _PROJECT_ROOT)

from data_preparation.prepare_dataset import get_default_split_config, load_per_person, SELECTED_FEATURES
from models.xgboost.config import XGB_BASE_PARAMS, build_xgb_classifier, get_xgb_params

_, SEED = get_default_split_config()
FEATURES = SELECTED_FEATURES["face_orientation"]


def _resolve_xgb_path():
    return os.path.join(_PROJECT_ROOT, "checkpoints", "xgboost_face_orientation_best.json")


def xgb_feature_importance():
    """Load trained XGBoost and return gain-based importance for the 10 features."""
    path = _resolve_xgb_path()
    if not os.path.isfile(path):
        print(f"[WARN] No XGBoost checkpoint at {path}; skip importance.")
        return None
    model = XGBClassifier()
    model.load_model(path)
    imp = model.get_booster().get_score(importance_type="gain")
    # Booster uses f0, f1, ...; we use same order as FEATURES (training order)
    by_idx = {int(k.replace("f", "")): v for k, v in imp.items() if k.startswith("f")}
    order = [by_idx.get(i, 0.0) for i in range(len(FEATURES))]
    return dict(zip(FEATURES, order))


def _make_eval_model(seed: int, quick: bool):
    if not quick:
        return build_xgb_classifier(seed, verbosity=0)

    params = get_xgb_params()
    params["n_estimators"] = 200
    params["random_state"] = seed
    params["verbosity"] = 0
    return XGBClassifier(**params)


def run_ablation_lopo(by_person, persons, quick: bool):
    """Leave-one-feature-out: for each feature, train XGBoost on the other 9 with LOPO, report mean F1."""
    results = {}
    for drop_feat in FEATURES:
        print(f"  -> dropping {drop_feat} ({len(results)+1}/{len(FEATURES)})")
        idx_keep = [i for i, f in enumerate(FEATURES) if f != drop_feat]
        f1s = []
        for held_out in persons:
            train_X = np.concatenate([by_person[p][0] for p in persons if p != held_out])
            train_y = np.concatenate([by_person[p][1] for p in persons if p != held_out])
            X_test, y_test = by_person[held_out]

            X_tr = train_X[:, idx_keep]
            X_te = X_test[:, idx_keep]
            scaler = StandardScaler().fit(X_tr)
            X_tr_sc = scaler.transform(X_tr)
            X_te_sc = scaler.transform(X_te)

            xgb = _make_eval_model(SEED, quick)
            xgb.fit(X_tr_sc, train_y)
            pred = xgb.predict(X_te_sc)
            f1s.append(f1_score(y_test, pred, average="weighted"))
        results[drop_feat] = np.mean(f1s)
    return results


def run_baseline_lopo_f1(by_person, persons, quick: bool):
    """Full 10-feature LOPO mean F1 for reference."""
    f1s = []
    for held_out in persons:
        train_X = np.concatenate([by_person[p][0] for p in persons if p != held_out])
        train_y = np.concatenate([by_person[p][1] for p in persons if p != held_out])
        X_test, y_test = by_person[held_out]
        scaler = StandardScaler().fit(train_X)
        X_tr_sc = scaler.transform(train_X)
        X_te_sc = scaler.transform(X_test)
        xgb = _make_eval_model(SEED, quick)
        xgb.fit(X_tr_sc, train_y)
        pred = xgb.predict(X_te_sc)
        f1s.append(f1_score(y_test, pred, average="weighted"))
    return np.mean(f1s)


# Channel subsets for ablation (subset name -> list of feature names)
CHANNEL_SUBSETS = {
    "head_pose": ["head_deviation", "s_face", "pitch"],
    "eye_state": ["ear_left", "ear_avg", "ear_right", "perclos"],
    "gaze": ["h_gaze", "gaze_offset", "s_eye"],
}


def run_channel_ablation(by_person, persons, quick: bool, baseline: float):
    """LOPO XGBoost with head-only, eye-only, gaze-only, and all 10. Returns dict subset_name -> mean F1."""
    results = {}
    for subset_name, feat_list in CHANNEL_SUBSETS.items():
        print(f"  -> channel {subset_name}")
        idx_keep = [FEATURES.index(f) for f in feat_list]
        f1s = []
        for held_out in persons:
            train_X = np.concatenate([by_person[p][0] for p in persons if p != held_out])
            train_y = np.concatenate([by_person[p][1] for p in persons if p != held_out])
            X_test, y_test = by_person[held_out]
            X_tr = train_X[:, idx_keep]
            X_te = X_test[:, idx_keep]
            scaler = StandardScaler().fit(X_tr)
            X_tr_sc = scaler.transform(X_tr)
            X_te_sc = scaler.transform(X_te)
            xgb = _make_eval_model(SEED, quick)
            xgb.fit(X_tr_sc, train_y)
            pred = xgb.predict(X_te_sc)
            f1s.append(f1_score(y_test, pred, average="weighted"))
        results[subset_name] = np.mean(f1s)
    results["all_10"] = baseline
    return results


def _parse_args():
    parser = argparse.ArgumentParser(description="Feature importance + LOPO ablation")
    parser.add_argument(
        "--quick",
        action="store_true",
        help="Use fewer trees (200) for faster iteration.",
    )
    parser.add_argument(
        "--skip-lofo",
        action="store_true",
        help="Skip leave-one-feature-out ablation.",
    )
    parser.add_argument(
        "--skip-channel",
        action="store_true",
        help="Skip channel ablation.",
    )
    return parser.parse_args()


def main():
    args = _parse_args()
    print("=== Feature importance (XGBoost gain) ===")
    if args.quick:
        print("Running in quick mode (n_estimators=200).")
    imp = xgb_feature_importance()
    if imp:
        for name in FEATURES:
            print(f"  {name}: {imp.get(name, 0):.2f}")
        order = sorted(imp.items(), key=lambda x: -x[1])
        print("  Top-5 by gain:", [x[0] for x in order[:5]])

    print("\n[DATA] Loading per-person splits once...")
    by_person, _, _ = load_per_person("face_orientation")
    persons = sorted(by_person.keys())

    print("\n=== Baseline LOPO (all 10 features) ===")
    baseline = run_baseline_lopo_f1(by_person, persons, quick=args.quick)
    print(f"  Baseline (all 10 features) mean LOPO F1: {baseline:.4f}")

    ablation = None
    worst_drop = None
    if args.skip_lofo:
        print("\n=== Leave-one-feature-out ablation (LOPO mean F1) ===")
        print("  skipped (--skip-lofo)")
    else:
        print("\n=== Leave-one-feature-out ablation (LOPO mean F1) ===")
        ablation = run_ablation_lopo(by_person, persons, quick=args.quick)
        for feat in FEATURES:
            delta = baseline - ablation[feat]
            print(f"  drop {feat}: F1={ablation[feat]:.4f} (Δ={delta:+.4f})")
        worst_drop = min(ablation.items(), key=lambda x: x[1])
        print(f"  Largest F1 drop when dropping: {worst_drop[0]} (F1={worst_drop[1]:.4f})")

    channel_f1 = None
    if args.skip_channel:
        print("\n=== Channel ablation (LOPO mean F1) ===")
        print("  skipped (--skip-channel)")
    else:
        print("\n=== Channel ablation (LOPO mean F1) ===")
        channel_f1 = run_channel_ablation(by_person, persons, quick=args.quick, baseline=baseline)
        for name, f1 in channel_f1.items():
            print(f"  {name}: {f1:.4f}")

    out_dir = os.path.join(_PROJECT_ROOT, "evaluation")
    out_path = os.path.join(out_dir, "feature_selection_justification.md")
    lines = [
        "# Feature selection justification",
        "",
        "The face_orientation model uses 10 of 17 extracted features. This document summarises empirical support.",
        "",
        "## 1. Domain rationale",
        "",
        "The 10 features were chosen to cover three channels:",
        "- **Head pose:** head_deviation, s_face, pitch",
        "- **Eye state:** ear_left, ear_right, ear_avg, perclos",
        "- **Gaze:** h_gaze, gaze_offset, s_eye",
        "",
        "Excluded: v_gaze (noisy), mar (rare events), yaw/roll (redundant with head_deviation/s_face), blink_rate/closure_duration/yawn_duration (temporal overlap with perclos).",
        "",
        "## 2. XGBoost feature importance (gain)",
        "",
        f"Config used: `{XGB_BASE_PARAMS}`.",
        "Quick mode: " + ("yes (200 trees)" if args.quick else "no (full config)"),
        "",
        "From the trained XGBoost checkpoint (gain on the 10 features):",
        "",
        "| Feature | Gain |",
        "|---------|------|",
    ]
    if imp:
        for name in FEATURES:
            lines.append(f"| {name} | {imp.get(name, 0):.2f} |")
        order = sorted(imp.items(), key=lambda x: -x[1])
        lines.append("")
        lines.append(f"**Top 5 by gain:** {', '.join(x[0] for x in order[:5])}.")
    else:
        lines.append("(Run with XGBoost checkpoint to populate.)")
    lines.extend([
        "",
        "## 3. Leave-one-feature-out ablation (LOPO)",
        "",
        f"Baseline (all 10 features) mean LOPO F1: **{baseline:.4f}**.",
        "",
    ])
    if ablation is None:
        lines.append("Skipped in this run (`--skip-lofo`).")
    else:
        lines.extend([
            "| Feature dropped | Mean LOPO F1 | Δ vs baseline |",
            "|------------------|--------------|---------------|",
        ])
        for feat in FEATURES:
            delta = baseline - ablation[feat]
            lines.append(f"| {feat} | {ablation[feat]:.4f} | {delta:+.4f} |")
        lines.append("")
        lines.append(f"Dropping **{worst_drop[0]}** hurts most (F1={worst_drop[1]:.4f}), consistent with it being important.")

    lines.append("")
    lines.append("## 4. Channel ablation (LOPO)")
    lines.append("")
    if channel_f1 is None:
        lines.append("Skipped in this run (`--skip-channel`).")
    else:
        lines.append("| Subset | Mean LOPO F1 |")
        lines.append("|--------|--------------|")
        for name in ["head_pose", "eye_state", "gaze", "all_10"]:
            lines.append(f"| {name} | {channel_f1[name]:.4f} |")
    lines.append("")
    lines.append("## 5. Conclusion")
    lines.append("")
    if ablation is None:
        lines.append("Selection is supported by (1) domain rationale (three attention channels), (2) XGBoost gain importance, and (3) channel ablation. Run without `--skip-lofo` for full leave-one-out ablation.")
    else:
        lines.append("Selection is supported by (1) domain rationale (three attention channels), (2) XGBoost gain importance, and (3) leave-one-out ablation. SHAP or correlation-based pruning can be added in future work.")
    lines.append("")
    with open(out_path, "w", encoding="utf-8") as f:
        f.write("\n".join(lines))
    print(f"\nReport written to {out_path}")


if __name__ == "__main__":
    main()