Skip to content
Projects · Aug 2022 – Dec 2022 · PES University

Stock Market Prediction using HMMs

Time‑series modeling with HMMs to infer market regimes and forecast behavior — a probabilistic approach to noisy financial data.

Probabilistic Models Time Series State Estimation Evaluation

Overview

A probabilistic modeling project that uses Hidden Markov Models to capture latent market regimes and make regime‑aware forecasts on stock price movements.

Problem

Financial time series are noisy and non‑stationary. Simple models struggle when the underlying regime shifts — bull, bear, volatile, calm.

Approach

Preprocessed time‑series data into observation features suitable for HMM training. Trained models to infer hidden states (regimes) and used those states to drive predictions. Validated results with careful evaluation and emphasized interpretability of learned regimes.

Impact

A strong “math‑meets‑engineering” project: showcases probabilistic reasoning, time‑series intuition, and the ability to explain why a model behaves the way it does.