How muralidanda0/crypto-price-forecasting works
CryptoVision serves cryptocurrency market data, technical indicators, and rolling forecasts through a Flask API to an interactive dashboard. The main workflow runs from a dashboard request through Flask, CoinGecko data retrieval, preprocessing, and—when forecasting—model loading and prediction, then returns JSON for visualization. A separate retraining workflow fetches history, prepares sequences, trains the three model types, and saves metrics and models. The README documents caching, synthetic-data fallback, trading insights, and background retraining; unsampled route and template details are represented at subsystem level.
Main entry points: pretrain.py Training Pipeline · app.js Dashboard Client · coingecko_api.py CoinGecko Wrapper · app.py Flask API
Components 21
Dashboard and API
- Dashboard Client
static/js/app.js - Dashboard Layout
templates/index.html - Flask API
app.py - Market Endpoints
app.py - Forecast Endpoint
app.py - Retraining Control
app.py - Trading Insights
static/js/app.js - Interactive Charts
static/js/app.js
Market Data
- CoinGecko Wrapper
api/coingecko_api.py - History Processing
ml/preprocess.py - Technical Indicators
ml/preprocess.py
Forecasting and Training
- Training Pipeline
pretrain.py - Sequence Windows
ml/preprocess.py - LSTM Model
ml/lstm_model.py - Regression Models
ml/regression_model.py - Rolling Forecasts
ml/predict.py
Local State
- Disk Cache
api/coingecko_api.py - Saved Models
ml/lstm_model.py - Saved Metrics
pretrain.py
Other components
- Dashboard User
- CoinGecko API
Connections 28
- Dashboard User → Dashboard Client: uses
README.md - Dashboard Client → Dashboard Layout: renders
README.md - Dashboard Client → Flask API: requests data
static/js/app.js - Flask API → Market Endpoints: routes requests
app.py - Flask API → Forecast Endpoint: routes requests
app.py - Flask API → Retraining Control: routes requests
app.py - Market Endpoints → CoinGecko Wrapper: fetches market data
app.py - Market Endpoints → History Processing: builds history
app.py - Market Endpoints → Technical Indicators: calculates indicators
app.py - CoinGecko Wrapper → CoinGecko API: requests data
README.md - CoinGecko Wrapper → Disk Cache: reads and writes
README.md - Forecast Endpoint → CoinGecko Wrapper: fetches history
app.py - Forecast Endpoint → History Processing: builds history
app.py - Forecast Endpoint → Rolling Forecasts: generates forecasts
app.py - Rolling Forecasts → History Processing: loads scaler
ml/predict.py - Rolling Forecasts → LSTM Model: loads model
ml/predict.py - Rolling Forecasts → Regression Models: loads models
ml/predict.py - Retraining Control → Training Pipeline: starts retraining
README.md - Training Pipeline → CoinGecko Wrapper: fetches history
pretrain.py - Training Pipeline → History Processing: prepares data
pretrain.py - Training Pipeline → Sequence Windows: creates windows
pretrain.py - Training Pipeline → LSTM Model: trains model
pretrain.py - Training Pipeline → Regression Models: trains models
pretrain.py - LSTM Model → Saved Models: saves model
ml/lstm_model.py - Regression Models → Saved Models: saves models
ml/regression_model.py - Training Pipeline → Saved Metrics: saves metrics
pretrain.py - Dashboard Client → Interactive Charts: plots forecasts
static/js/app.js - Dashboard Client → Trading Insights: derives signals
README.md