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:
Harwinder Sekhon
2026-09-27 19:22:30 -04:00
commit 01baef48ff
98 changed files with 1719757 additions and 0 deletions
+20
View File
@@ -0,0 +1,20 @@
growth: 94 windows of 30 days x 20 queries; 4245 done, 1880 to go
100 requests, window 2023-09-13, 38522 unique headlines, 26 req/min
200 requests, window 2020-02-01, 38742 unique headlines, 26 req/min
300 requests, window 2019-02-06, 39126 unique headlines, 26 req/min
400 requests, window 2022-08-19, 39452 unique headlines, 26 req/min
500 requests, window 2025-09-02, 39819 unique headlines, 26 req/min
600 requests, window 2023-11-12, 40298 unique headlines, 26 req/min
700 requests, window 2026-03-31, 40734 unique headlines, 26 req/min
800 requests, window 2025-07-04, 41074 unique headlines, 26 req/min
900 requests, window 2020-09-28, 41336 unique headlines, 26 req/min
1000 requests, window 2024-10-07, 41810 unique headlines, 26 req/min
1100 requests, window 2026-06-29, 42163 unique headlines, 26 req/min
1200 requests, window 2020-05-01, 42407 unique headlines, 26 req/min
1300 requests, window 2021-02-25, 42948 unique headlines, 26 req/min
1400 requests, window 2023-12-12, 43263 unique headlines, 26 req/min
1500 requests, window 2020-06-30, 43490 unique headlines, 26 req/min
1600 requests, window 2021-12-22, 43819 unique headlines, 26 req/min
1700 requests, window 2022-09-18, 44095 unique headlines, 26 req/min
1800 requests, window 2024-05-10, 44388 unique headlines, 26 req/min
done: 1880 requests this run, 44529 unique headlines in /home/harwinder/Development/ML-NEWS/data/history.db