Quantopian's community platform is shutting down. Please read this post for more information and download your code.
Back to Community
Time series pattern recognition/identification

What is the easiest way to construct (identify) time series technical trading patterns on Quantopian? such as "head and shoulder","cup and handle", "higher high", etc. Are there any existing Python libraries? Thanks.

3 responses

You may want to look at this post. Excellent. https://www.quantopian.com/posts/an-empirical-algorithmic-evaluation-of-technical-analysis

Use the research environment to experiment with different 'signals'. Categorize the results by different features. Look at 'alphalens' to determine the effectiveness across different categories. https://www.quantopian.com/posts/alphalens-a-new-tool-for-analyzing-alpha-factors . The notebook compares efficacy of a factor vs stock sector. However it is very informative to change the grouping and look across other categories such as different market caps (small, mid, large) or across different betas (high, low, inverse, correlation to SPY), or across different recent returns (high momentum or low momentum).

That may help?

Dan,
Thanks you vert much for pointing to the post. It is very helpful.

With the recent development in AI, it will be close to obtain comparable recognition ability to that of a human trader.