MarketPulse AI — Broker Market Analysis & Forecast

Problem

Traders on the South Korean equity market face information overload — live prices, technicals, and short-horizon decisions without a clear, model-backed forecast or confidence signal.

Challenges

• Live broker market data via KIS API with reliable connection and market-session awareness
• Multi-horizon price forecasts (seconds to minutes) without lagging the UI
• Transparent Bullish / Bearish / Neutral probabilities plus actionable HOLD/BUY/SELL guidance
• Explainable evidence from EMA, MACD, VWAP and rule/ML hybrid signals

Solution

MarketPulse AI is a financial data analysis platform for the South Korean market: real-time broker feeds (KIS API), candlestick charts with EMA overlays, AI prediction probabilities, multi-horizon forecasts, and confidence-scored recommendations with rule-based evidence.

Architecture

Broker-agnostic market data layer connects to KIS for Korean equities; streaming ticks update watchlist and chart; technical indicators compute in near real time; prediction engine outputs Bullish/Bearish/Neutral probabilities and horizon forecasts (e.g. 1m / 5m / 10m / 30m); dashboard surfaces recommendation (HOLD/BUY/SELL) with confidence and evidence chips.

Timeline

Production dashboard (v0.1+)

Lessons learned

• Traders trust forecasts more when probabilities and technical evidence are shown together
• Broker session state (open/closed/connected) must be first-class in the UI
• Short-horizon forecasts need streaming latency budgets, not batch jobs