Stock Outperformers — AI Signals
How our AI turns raw market data into ranked stock signals
v2025.10Owner: Stock Outperformers Team
Strategies
5
Model families
7
Latency (typical)
Real‑time (ms‑level)
How it works
1. Ingest
Import CSVs from data partners (per strategy).
→
2. Clean & Normalize
Fix decimals, strip symbols, handle ‘crap values’, drop constant/date/crypto cols, dedupe by primary key.
→
3. Type Detect
Detect numeric vs categorical; enforce per‑strategy types; impute nulls.
→
4. Feature Build
Compose training tables; compute profit label from latest vs prior price.
→
5. Train/Assemble
CatBoost Classifier/Regressor, rule combinations, and specialty models (Achievers, Owner Increase, AutoML).
→
6. Score & Rank
Predict, then convert to a 0–100 rank so lists are easy to compare across sources.
→
7. Publish
Securely send signals to our private systems (no public endpoints).
→
8. Govern
Schema checks, calibration view, drift monitoring, documented limits.
What you get
Daily ranked stock opportunities scored from 0–100.
Probability estimates of profit vs loss (when model supports it).
Delivered instantly to your internal applications or dashboards.
How to read a signal
Rank (0–100): percentile within today's universe. 100 = top of the list.
Profit/Loss probability: where supported, CatBoost classifier outputs probabilities; use them with thresholds, not as guarantees.
Source: signals are aggregated across Data Provider A–E; names are masked by design.
Privacy & masking
Provider identities and internal endpoints are intentionally redacted in this public view.
Generated overview • Data flows → Models → Signals • For details, contact Stock Outperformers Team