omg this is something i will look the entire serie, ty man
@eminsh
9 ай бұрын
So glad you enjoyed it, thanks for watching
@peralser
4 ай бұрын
Great Video! Thanks for sharing your knowledge with us!!
@elbozo5723
6 ай бұрын
I’m not gonna lie, this seems poorly thought out. You’re using a basic multilayer perceptron on sequential data; when more complex models like RNNs, LSTMs, and Transformers are specifically designed for this purpose. A basic google search of “stock prediction using neural networks” basically confirms this theory, as almost all use LSTMs or RNNs. Do you have any justification for using a basic MLP here?
@ouanesachouche6785
6 ай бұрын
You forgot to mention the quality of that yahoo finance data 😂 , at least clean it a bit and add some features
@jaydy71
5 ай бұрын
To be fair, he did present it as merely an example to get started and he did point out that this particular model was unsuccessful. LSTMs etc are indeed designed for sequential data but in my personal experience I was never successful using them for market predictions in a trading bot. Many online articles that demonstrate LSTM for price predictions look very promising at face value, but if you look closely they never seem able to actually predict anything with any kind of usable accuracy. I have successfully used LSTMs in other domains, but not price predictions. My most successful trading bot effort has been actually trading for over 6 months now (and with pretty consistent profits) and it is using an MLP, so don't disregard them out of hand. The most common trap I've seen people fall into (including me) is throwing lots of candlestick data at a ML model and hope for it to become Warren Buffett, because that won't work, no matter what kind of machine learning model you use.
@maxfrenk
2 ай бұрын
the justification is - it's a training video that's not meant for production :)
@elbozo5723
2 ай бұрын
@@maxfrenk but training for what? It’s supposed to be about deep neural networks and day trading AIs, but he didn’t come close to correctly implementing either
@maxfrenk
2 ай бұрын
@@elbozo5723 I mean, it's meant to get views on KZitem 😂 that's the primary purpose
@iG_GeneRal
2 күн бұрын
Any one here that talk bad about this video should go and do its own, this is a massive work to expose, thank you very this information, by the way how can use the code
@theos9207
4 ай бұрын
Thank you so much for sharing
@DalazG
8 ай бұрын
This is brilliant. I'm learning coding myself so this is great help! Wondering is it possible to integrate this trading ai into mql for forex trading?
@eminsh
8 ай бұрын
Lots of viewers interested in forex. I plan on making videos on forex, futures, and options but research takes time.
@leonxger
5 ай бұрын
I just stumbled upon this amazing video. This is so interesting to see. I'm a developer myself and have always wanted to get into ML and AI programming. I'm new to Python, but I'm learning. Which algorithm would you recommend me to use for market data prediction? Im preparing a dataset with lots of good indicators right low and cant wait to backtest it. Thanks in advance! 😀
@eminsh
5 ай бұрын
That's great to hear! market data prediction is a pretty broad term, but my best results have come from an ensemble model where you combine different models to balance out their predictive weaknesses. I have a video of it on my channel.
@leonxger
5 ай бұрын
Amazing! Definitely going to check that out! I just built a Python script that fetched me every bitcoin tick for the last 4.5 years, giving me 150 million rows of potential practice data 😂@@eminsh
@amuga_1
10 ай бұрын
❤
@BoxingBytes-y9i
8 ай бұрын
Hi, thanks for the video. What would you say would be the next steps for someone looking to learn& implement a strategy in production? There are so many things to try, different ML algorithms, using different lags, different granularity in data, different data... Sounds like too much to try
@eminsh
8 ай бұрын
I've been playing around with RL in finance for almost a year now trying different stuff and sharing what I think is interesting and I still haven't felt like anything is worth deploying. I'd be flat broke if I gave any of the models actual money to trade. That said, I think the only way to find an actual application is to test the many different variables, so I would suggest just starting testing something.
@iainstruthers3559
9 ай бұрын
Hi, thanks so much for posting this. Do you think you would you be able to improve the performance on an out of sample set by training the model with multiple training sets?
@eminsh
9 ай бұрын
I read a paper once where the writers would retrain their model every quarter because they found, what they called "alpha decay", where performance dropped as time went on because the model didn't have context to current data like a human would gain from experience. I think this approach would maybe be better than training on larger and larger datasets going farther and farther into the past.
@govardhanab7223
8 ай бұрын
Hi Emin shall , please tell success rate of the prediction by AI/ ML models for index options trading for day trading work ? if yes which models and how and why and if no please share
@eminsh
8 ай бұрын
I don't think I've ever tried to use a model with options trading before so I'm not really sure what the success is there. I'll try to wrangle a bunch of data for it if I can build/find an environment for that asset class.
@govardhanab7223
8 ай бұрын
@@eminsh also how to feed real time ticker data unlike historical data?
@eminsh
8 ай бұрын
I made a video about how to stream data using websockets here: kzitem.info/news/bejne/2YFry4CCkJeXg44
@elixir_0072
2 ай бұрын
Good, stuff I just finished an internship on DS and I'm trying to put my skills to the test on predictive modeling for XAUUSD. My objective is to forecast the high and low on a daily basis. I noticed the features you used were based on the Adj.close and engineered lags, why don't you consider using OHLC features as well with engineered, macd, ma, ema, std, rsi. Etc? I would like to rub minds and share my notes with you if possible...
@triviagames6507
10 ай бұрын
Can you share the code and training data for this?
@eminsh
9 ай бұрын
I'm currently working on building a place to access all my code from these videos. Just taking me a little longer than expected
@heuksalman
6 ай бұрын
how to make it profitable then?
@eminsh
5 ай бұрын
to be honest I don't really know, I have hypothesis but I'll share it here if I figure it out
@ConnorDavis-jo2qb
10 ай бұрын
Please share code and data
@eminsh
9 ай бұрын
Organizing my code now, will be posting links in the description of each video moving forward shortly.
@MuratDagcan
10 күн бұрын
Garcia Shirley Young Shirley Perez Donna
@capitola
7 ай бұрын
can you generate the requirements.txt file for packages versions? and which version on pythoy are you using?
Пікірлер: 36