Data Mining Novel Chart Patterns With Python | Algorithmic Trading Strategy

Using perceptually important points combined with unsupervised learning to find unique chart patterns for trading using python. We cluster the price structure patterns and select the high performing patterns using the martin ratio as an objective function. We perform a monte carlo permutation test to verify the results. We also perform a walkforward test.

This video has a detailed explanation of the perceptually important points algorithm.

Links

Citations
Chung, F.L., Fu, T.C., Luk, R., Ng, V., Flexible Time Series Pattern Matching Based on
Perceptually Important Points. In: Workshop on Learning from Temporal and Spatial Data
at IJCAI (2001) 1-7

Keogh, E., Lin, J., Truppel, W.: Clustering of Time Series Subsequences is Meaningless:
Implications for Previous and Future Research. Proc. of ICDM, (2003) 115-122

Fu, Tc., Chung, Fl., Luk, R., Ng, Cm. (2005). Preventing Meaningless Stock Time Series Pattern Discovery by Changing Perceptually Important Point Detection. In: Wang, L., Jin, Y. (eds) Fuzzy Systems and Knowledge Discovery. FSKD 2005. Lecture Notes in Computer Science(), vol 3613. Springer, Berlin, Heidelberg.

Peter Rousseeuw. Silhouettes: A graphical aid to the interpretation and validation of cluster analysis. J. Comput. Appl. Math., 20(1):53–65, November 1987.

The content covered on this channel is NOT to be considered as any financial or investment advice. Past results are not necessarily indicative of future results. This content is purely for education/entertainment.

[cheesecake_mafia]

Mate, I had developed something similar but very crude compared to this and was pretty happy with myself until I saw this. Wooh!! I need to get to work. This video is one of the best examples of how to think critically and systematically in quantitative finance. I raise my hat to you, sir! You are doing an awesome job!! Kudos!!

[jeremycarroll]

Your skill, patience and academic rigour are greatly appreciated. Please continue making these videos; it's unbelievably helpful and educational.

[justjoy]

Trying to stop watching and go to bed but the monotonous tones of your voice combined with very cool lines moving along matplotlib, followed by well explained code keeps me hooked.

[mozkhiyar]

My man i dont know why ypu have to low subs compared to the effort and knowledge that you are sharing. I loved what you are doing and keep it up. This channel is gold.

[Asparuh.Emilov]

I love your videos! You are an absolute Data Science Master! Absolute professional! Amazing Job! Thanks for sharing this information! You are the absolute best algorithmic trader I have seen so far!

[sergioerm]

Man, it's just a pity that very few people can even follow the basics of what you're sharing... Thank YOU, THANK YOU!

[hellskreamer]

This is an absolute goldmine, i love how its so well explained theoretically, that it is very applicable! Been learning coding for a while, and this is just blowing my mind in terms of both content and availability. Having a conclusion in the end of the video is very handy! Thank you fr your hard work!

[mkl]

Applying selected patterns across various timeframes can enhance results significantly. For instance, identifying a pattern across both 1-hour and 1-day timeframes can act as an amplifier.

[pauldacus]

Something I use for finding the optimal KMeans cluster number is the elbow method, in the _kneed_ python package.

[Edbrad]

Just found you. This is all amazing. I don’t even care about getting so detailed. I do wonder if getting so microscopic is worth it because unless it gives you substantially more profits, it seems inefficient. I feel like it would be better to make the less zoomed in strategy better. Like why trade 10, 000 times on a 1 min chart to make the same as 5 trades on a 1 hour timeframe ?

[dimitrisUK]

this is spectacular. Thank you for sharing!

[UnbenutzerKanalname]

excellent video and channel overall, thanks! if you are looking for ideas, a method to determine the support and resistance zones which the highest probability of price making a meaningful reversion from would be very interesting. meaningful could be defined as a certain distance between the zones, e.g., a Fib or something like that.

[scGR]

also wondering, does standardizing pattern make them representable? or perhaps change data point to % change can visualize better?

[probablyonthemoon]

Great video. Have you tried extending this paper to n points? Thinking that if n=7 it might be able to see channels

[sambakker]

What IDE do you use? And where do you retrieve your historical data?

[nicholasgrayam]

Can volume be added with each candle price as another dimension in price shape? (Volume/Price shape) I would imagine clustering would be a similar process.

[scGR]

just curious, have you tried different clustering methods?

[mohamedabass]

Good job
I wish to apply hidden Markov chain to algorithmic trading

[cybergrimreaper]

Sir, what is the sequence for watching your videos

[pratyushsethi]

I was doing a project, and wanted to ask something, i am new to this, but have one idea for pattern recognition. (for flag, wedge, head and shoulder....). If free we can chat over the email.