Clustering and HTM for Stock Prediction

Hello, we have recently used clustering and HTM methods to predict the price of Chinese stocks. Our work is as follows: Stock Price Prediction Based on Morphological Similarity Clustering and Hierarchical Temporal Memory | IEEE Journals & Magazine | IEEE Xplore.

However, the results of the experiment were not particularly satisfactory. Is there any better way to improve the prediction accuracy? Is our method correct?


You know that it is well-known that you can’t predict the market :wink:


That overstates the premise of the Efficient Market.

You cannot predict the market if you rely on the same data, use the same tools and work within the same time constraints as everyone else. I already know multiple ways to predict the markets, and people who make a living at it. It’s a boring job, if rather well paid.

But market efficiency relies on rewarding those who find and trade on inefficiencies, and AI tools are in big demand to do just that. Markets rely on the flow of time and HTM is better than most in that department, so it wouldn’t surprise me at all if there were HTM practitioners out there doing just that.


Here is your proof:

Commercial use of HTM
Numenta has three commercial partners using HTM. […] The third partner, Intelletic Trading Systems, is a fintech startup. The company has developed a platform for autonomous trading of futures and other financial assets using HTM. The described commercial usage documents HTM’s practical relevance.

1 Like

I think it’s limiting to look at the stock market as a regression problem or time series problem.

There are distinct price structures that cause significant effects on the price movement. Some of these structures, price patterns, have extreme high statistical movement direction.

Segmenting price structure into distinct categories could be the key to high performing models. HTM is great with pattern detection.

Although there are data that are difficult to procure that will produce highly accurate long term directional forecast using regression techniques. I think that is a much better focus than pure price prediction.

1 Like