Files
News-ML-Decoder/eval/results/backfill_20260917_2322.json
Harwinder Sekhon 01baef48ff Initial import: XAUUSD news sentiment models, evals, and data tooling
TF-IDF baseline (news_sentiment.py, models/xauusd_sentiment.joblib) and the
two-head DeBERTa-v3-small transformer (transformer/, news_transformer.py)
with its Google News crawler and DeepSeek labeller (data_collect/), the
golden-set and archive evals (eval/), and the 2026-09-22 and 2026-09-27
results and progress notes.

Transformer weights (model.pt, model.onnx) and the crawl database are not
tracked; rebuild with transformer/train.py and transformer/export_onnx.py.
The DeepSeek label cache in data/ is tracked.
2026-09-27 19:22:30 -04:00

190 lines
3.5 KiB
JSON

{
"generated_at": "20260917_2322",
"gold_rows": 616,
"golden": {
"gold+macro (backfilled)": {
"accuracy": 0.5384615384615384,
"confusion": [
[
16,
4,
12
],
[
1,
6,
1
],
[
11,
7,
20
]
],
"directional_accuracy": 0.5142857142857142,
"directional_hits": 36,
"directional_n": 70,
"hits": 42,
"macro_f1": 0.5255712050078247,
"n": 78,
"sign_agreement": 0.6101694915254238,
"sign_hits": 36,
"sign_n": 59
},
"gold-only (shipped data)": {
"accuracy": 0.5,
"confusion": [
[
13,
2,
17
],
[
2,
1,
5
],
[
12,
1,
25
]
],
"directional_accuracy": 0.5428571428571428,
"directional_hits": 38,
"directional_n": 70,
"hits": 39,
"macro_f1": 0.3985266422953362,
"n": 78,
"sign_agreement": 0.5671641791044776,
"sign_hits": 38,
"sign_n": 67
}
},
"macro_label_distribution": {
"bearish": 350,
"bullish": 228,
"neutral": 383
},
"macro_rows": 961,
"pooled_oof": {
"gold/g+m": {
"accuracy": 0.6155202821869489,
"confusion": [
[
164,
27,
52
],
[
15,
74,
32
],
[
47,
45,
111
]
],
"directional_accuracy": 0.6165919282511211,
"directional_hits": 275,
"directional_n": 446,
"hits": 349,
"macro_f1": 0.6038188006599577,
"n": 567,
"sign_agreement": 0.7352941176470589,
"sign_hits": 275,
"sign_n": 374
},
"gold/g-only": {
"accuracy": 0.6437389770723104,
"confusion": [
[
165,
23,
55
],
[
13,
83,
25
],
[
40,
46,
117
]
],
"directional_accuracy": 0.6322869955156951,
"directional_hits": 282,
"directional_n": 446,
"hits": 365,
"macro_f1": 0.6362979163521462,
"n": 567,
"sign_agreement": 0.7480106100795756,
"sign_hits": 282,
"sign_n": 377
},
"macro/g+m": {
"accuracy": 0.5183129855715871,
"confusion": [
[
174,
127,
38
],
[
43,
244,
58
],
[
61,
107,
49
]
],
"directional_accuracy": 0.4010791366906475,
"directional_hits": 223,
"directional_n": 556,
"hits": 467,
"macro_f1": 0.47589676446146534,
"n": 901,
"sign_agreement": 0.6925465838509317,
"sign_hits": 223,
"sign_n": 322
},
"macro/g-only": {
"accuracy": 0.38401775804661487,
"confusion": [
[
137,
71,
131
],
[
70,
96,
179
],
[
65,
39,
113
]
],
"directional_accuracy": 0.44964028776978415,
"directional_hits": 250,
"directional_n": 556,
"hits": 346,
"macro_f1": 0.3833425073738865,
"n": 901,
"sign_agreement": 0.5605381165919282,
"sign_hits": 250,
"sign_n": 446
}
},
"threshold": 0.05
}