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.
This commit is contained in:
@@ -0,0 +1,44 @@
|
||||
DeepSeek deepseek-v4-flash, prompt d32e04c6; budget 200 calls
|
||||
pools: precise 6600, broad 0 unlabelled; picked 5000
|
||||
selected 5000 unlabelled headlines
|
||||
250 labelled, 10 calls (0 failed), 0 from cache
|
||||
500 labelled, 20 calls (0 failed), 0 from cache
|
||||
DeepSeek batch of 25 failed (attempt 1/3, None) — retrying
|
||||
750 labelled, 31 calls (1 failed), 0 from cache
|
||||
1000 labelled, 41 calls (1 failed), 0 from cache
|
||||
1250 labelled, 51 calls (1 failed), 0 from cache
|
||||
1500 labelled, 61 calls (1 failed), 0 from cache
|
||||
1750 labelled, 71 calls (1 failed), 0 from cache
|
||||
2000 labelled, 81 calls (1 failed), 0 from cache
|
||||
2250 labelled, 91 calls (1 failed), 0 from cache
|
||||
2500 labelled, 101 calls (1 failed), 0 from cache
|
||||
2750 labelled, 111 calls (1 failed), 0 from cache
|
||||
3000 labelled, 121 calls (1 failed), 0 from cache
|
||||
3250 labelled, 131 calls (1 failed), 0 from cache
|
||||
3500 labelled, 141 calls (1 failed), 0 from cache
|
||||
3750 labelled, 151 calls (1 failed), 0 from cache
|
||||
4000 labelled, 161 calls (1 failed), 0 from cache
|
||||
4250 labelled, 171 calls (1 failed), 0 from cache
|
||||
4500 labelled, 181 calls (1 failed), 0 from cache
|
||||
4750 labelled, 191 calls (1 failed), 0 from cache
|
||||
4975 labelled, 200 calls (1 failed), 0 from cache
|
||||
this run: 200 calls, 1 failed, 4975 headlines labelled
|
||||
14625 labelled headlines
|
||||
year n bear irrel bull relevant mean xau
|
||||
2009 10 0 10 0 0.0% +0.000
|
||||
2010 10 0 10 0 0.0% +0.000
|
||||
2011 10 0 10 0 0.0% +0.000
|
||||
2012 2 0 2 0 0.0% +0.000
|
||||
2015 13 0 13 0 0.0% +0.000
|
||||
2016 6 0 6 0 0.0% +0.000
|
||||
2017 33 0 33 0 0.0% +0.000
|
||||
2018 19 0 19 0 0.0% +0.000
|
||||
2019 1003 78 648 184 26.1% +0.056
|
||||
2020 1211 80 762 219 24.7% +0.064
|
||||
2021 1264 104 876 151 20.2% +0.018
|
||||
2022 1536 145 930 284 27.9% +0.045
|
||||
2023 1691 145 1082 295 26.0% +0.046
|
||||
2024 2047 132 1325 408 26.4% +0.071
|
||||
2025 3071 222 1671 829 34.2% +0.109
|
||||
2026 2699 345 1379 635 36.3% +0.047
|
||||
(relevant = |xau| >= 0.25, irrel = |xau| <= 0.10, the rest is the grey zone)
|
||||
Reference in New Issue
Block a user