CodeDevX/future-prediction-multi-domain-lstm

Future Prediction Models (6 topics + Unified)

An end-to-end multi-domain AI forecasting system. Real datasets, two model modes, ChatGPT-style predictions.

Models: separate vs unified

• Separate (model_<topic>.pt) — 6 dedicated 2-layer LSTMs, one per topic
• Unified (unified_model.pt) — One combined model for all 6 domains: shared LSTM backbone + per-domain embedding + per-domain heads, trained jointly from merged separate models

The unified model merges weights from all 6 trained separate models into a single checkpoint, enabling cross-domain knowledge transfer and a single model that can predict any of the 6 topics.

Topics & datasets (all real, fetched automatically)

Topic	Asset	Source	Points
AI	NVIDIA daily close	Yahoo Finance	6,938
Programming	Daily react npm downloads	npm registry API	547
Finance	Bitcoin BTC-USD close	Yahoo Finance (Hugging Face fallback)	4,252
Sports	ATP world #1 Elo rating	Hugging Face tennis (93,028 matches, Elo computed from results)	10,944
Weather	Daily mean temperature (any city)	Open-Meteo archive	4,248
Economy	S&P 500 daily close	Yahoo Finance	6,699

Weather city is configurable. Every topic falls back gracefully: Hugging Face search -> API -> synthetic data if fully offline.

Validated accuracy (held-out test set, no data leakage)

Separate per-topic models:

Topic	MAE	H1 MAPE	vs naive baseline
AI (NVIDIA)	$2.17	2.26%	~naive
Economy (S&P 500)	$37.37	0.71%	beats naive by 0.5%
Finance (Bitcoin)	$1,437.46	1.61%	~naive
Programming (react)	1.17M downloads	7.12%	beats naive by 77.5%
Sports (ATP #1 Elo)	3.40 Elo	0.06%	~naive
Weather (Chennai temp)	0.99 K	0.18%	beats naive by 1.9%

Unified single model (merged from 6 separate models):

Topic	MAE	H1 MAPE
AI	$3.40	2.36%
Economy	$72.21	0.75%
Finance	$2,863.45	1.74%
Programming	1.14M downloads	5.42%
Sports	2.75 Elo	0.04%
Weather	0.94 K	0.18%

Markets behave near a random walk, so 100% accuracy is impossible — these are honest, validated numbers. No model can guarantee the future.

Architecture (normal mode shows only the final response)

User question -> intent detection -> prediction engine -> response formatter -> chat output

• Chat CLI — only the final assistant response reaches the user
• Intent detection + orchestration (topic, location, horizon, style)
• Prediction engine — runs the LSTM, returns structured metrics: net change, forecast range, current-to-forecast, direction, primary forecast
• Response formatter — presents metrics naturally; opens with a one-sentence ChatGPT-style summary; never invents confidence, probability, or values
• Multi-topic data pipeline (fetchers + caching)
• 2-layer LSTM, Huber loss, AdamW, early stopping, temporal split
• Per-topic model loader, rollout forecast, chart + text report

Ask it anything (ChatGPT-style)

Interactive chat with natural language queries:

• "what will bitcoin do next week?"
• "compare all topics"
• "predict programming for 14 days"
• "predict ai for 3 days"

Example response:

In short, the model expects Bitcoin to stay roughly stable next week, hovering near $63,228.

Bitcoin Forecast
August 18-24, 2026

The model forecasts relatively sideways movement for Bitcoin next week, with an
estimated level around **$63,228**.

Outlook: Sideways
Predicted level (2026-08-24): ~$63,228
Expected change: <0.01%
Forecast range: <0.01%
From latest observed value ($63,229, 2026-08-17): <0.01%

This is a model-generated forecast, and actual market behavior may differ.

Weather for any city

Configure city via environment variables or city registry, then query weather predictions.

Train / refresh

Run training for all topics or individual domains. Outputs: model checkpoints, configs, predictions, charts, and results.

Add a new domain

1. Add entry to topics configuration (label, unit, asset name)
2. Add a fetcher returning date + value columns
3. Run training — everything else is automatic

Hugging Face Repository

All models and configs are available at: https://huggingface.co/CodeDevX/future-prediction-multi-domain-lstm