victor369basu/Real-time-stock-market-prediction

The architecture implements parallel processing through dual Kafka topics—one streaming to MongoDB for storage and training, another to TensorFlow.js for real-time inference using pre-trained model weights. Data preprocessing uses MinMax-Scalar normalization on time-series stock data, with the model trained on 80/20 split and performing 8-step-ahead predictions by consuming 7 prior observations. Node.js producers/consumers enable simultaneous model training and live predictions across the pipeline without blocking I/O.

648 stars. No commits in the last 6 months.