One trick worth trying. Fit the model on the recent prices. Then get the model predictions in-sample. Compare the current actual price with the model's in-sample prediction of the current price. The difference between the actual and predicted might be effective as a mean reversion indicator. I've used idea this with ARMA and Holt-Winters. I got better results doing that then using the future predictions from those models. Never used prophet, seasonality in market prices is usually pretty weak though.
Hey thanks for sharing. Interesting idea, didn't think of it this way, it might be a good indicator, still have to find a way how to use it or to interpret it. ARMA family (SARIMAX...) are on my list when time and life will allow. Great channel by the way I liked your videos :) I will sub
Good video, if Prophet is good at predicting Seasonality then it should work good on stocks or sectors . because companies sales and profit are Seasonal according to their industry.
Hello friend, im the one who propesed it. You may look into the following hyperparameter tunings: weekly_seasonality=X, daily_seasonality = Y, yearly_seasonality = Z whereas X, Y and Z are the Fourier Order as a number. Additionally you may want to set changepoints on the opening of USA Session. The most important change i did, was to define the chanepoints manually on the date the FED rises interest rates or something unusual happend
Hey thank you again for this. Interesting idea to set changepoints according to financial calendar, but this depends also on the type of the event and its effect on the price which can't be predicted somehow?
@@CodeTradingCafe i coded a loop that went through several hundred stocks, changed changepoint in various ways and measured MAPE, and averaged the MAPE results. Same with every other hyperparameter. You can code it in 2-3 hours and let run over night. Maybe it helps. GL
Would including indicator outputs from Pandas TA improve the predictions? Here Prophet is using a x/y dataset to predict future range, but would including things like Bollinger, Stochastic, and/or MACD data on the "x" side of the model produce a better fitting prediction?
Hi, I have to look into prophet and see how it processes additional features. That being said, my answer is maybe using a smoother indicator like the moving average might help prophet since it eliminates the noise. And what I would be hoping for from prophet is to just detect the trend, if this is done correctly I guess it's a big help already.
There was a paper that predicted S&P500's next day price decently by the previous day's price, the futures' price, and other technical indicators. But they decompose it using ICA, and selected the least noisy components.
Outstanding video and work. Congratulations!!! At 9:42, where you first present the plot routine, there is a slight mistake where the parameter "yhatlow" should read "yhat" and, at same time, where it reads "yhat" it should read "yhatlow". Without fixing this it just plots funny. This mistake is also present in the code. But thank you for your work. Will subscribe.
by using the rnn and providing the recent business news source to system ,the rnn can scan weather it is a good news or bad and on the basis of technicals the system will predict to where the markets will hit from the news. just a idea
copy trading I am not sure is possible unless you get access to a particular account you need to copy. But for binance you can use their API just like I did using Oanda's in this video ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-WcfKaZL4vpA.html
Hello good day, I would like to know if you can track a file CVS with Python well yes for sure, but I want to know if that is possible with the results of that file to do an indicator base only on entries thank you in advance if you can make a video will be great, I'm one of your followers as well I program mql4 mql5 and python my friend excellent videos keep it up.