Author(s): Shenggang Li

Originally published on Towards AI.

A Novel Approach Combining Markov Decision Theory and Neural State Space Models for Stock Price Prediction

This member-only story is on us. Upgrade to access all of Medium.

Photo by Ella Jardim on Unsplash

Imagine you’re trying to predict stock prices, but instead of just guessing whether the price will go up or down tomorrow, you’re asking a smarter question: What happens if tomorrow’s price crosses a certain threshold? For instance, if a stock price drops below a key support level, what’s likely to happen in the following days? This kind of conditional forecasting is not only more insightful but also mirrors real-world decisions made in financial markets.

The problem is that traditional time series models aren’t built for these “what if” scenarios. That’s where Markov Decision Theory meets neural state space models like Mamba to create something new. By extending the classic state space framework, we can bake future conditions — like “tomorrow’s price event” — directly into the prediction process. Think of it as giving the model a crystal ball, allowing it to consider not just the past but also what might happen next.

In this new approach, we explore how adding event-driven dynamics to Mamba state space models unlocks exciting possibilities for forecasting. We connect these ideas to Markov Decision Processes (MDPs) to show why they’re so… Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI