Thank you. I am going to be creating more of these videos on forecasting shortly and will be using individual companies and their respective stock symbols. What company or companies would you like to see forecasting with (AAPL, GOOGL, MSFT, AMZN, FB, BRK-B, etc...)?
I have the code all done for those videos. Now it's just getting the time to make the videos and upload them. Finding the time to make these videos is not always easy when everyone knows you are a data scientist and are a great solutions resource. I will get them up as soon as I can. Thanks again. Also, be sure and check out all the other great videos on my channel that cover everything from forecasting, predicting, regression, custom arima models and more. :)
Hey, do you have an email address or a good forum to recommend ? I work for a contact center and I ve been struggling with forecasting my call volume. there is too much variances in my data. I need some help :(
Every data set is different. This is why we do many tests. Then, in the end, you need to just look at it, take a step back and ask, "Does this make sense?" In the end, your read and opinion is what these companies will pay the big money for. Sometimes you need to hold back and rerun the process. Sometimes you need to add more external or internal data for a better read. It all depends on the data, situation, date ranges, outside events and more.
Hello, I am facing difficulty when using data from yahoo finance, how to input other data instead of'SPY'. Is it done with the help of getsymbols function, or what am i doing wrong here. Can anyone help?
Yes. Look at the other similar videos. I recently just published one on the Pakistan ETF (PAK). Any stock ticker symbol that has data in Yahoo Finance will work. Some penny stocks may not have data there, so keep that in mind.
Thanks for the beautiful video, but from my experience I think that the subject of expectation is a very difficult subject. We spend a long time learning about the model then finding the proper one, testing it and forecasting it, this is a nice game, but it doesn't bear fruit. Political and economical conditions and investor sentiment affect at least 90% of the short-term or long-term pricing, leaving only 10% for the expectations and the models that are often unconvincing.
That is why, if you watch the entire video and any of my other forecasting video series, you will see that I tell people to look at it from a more holistic, directional approach as there are numerous factors at play that we cannot control for. I teach to look at it from a longer term perspective also. Look at it as if it is trending up or down rather than an exact price.
Thank you. I am glad you liked the video. Be sure to subscribe and click the bell so you get notified every time I create and publish great new videos like this one.
Sir when i run forecast for fitA it gives me result but when i run forecast for fit B it gives me error in attr(data,"tsp") what does it means kindly guide me .
Your data most likely has formatting issues or similar. Please read this - r.789695.n4.nabble.com/How-to-setup-the-tsp-attribute-of-a-dataset-td908269.html
Very informative Sir, kindly tell me when we fit autoarima fn it gives us best fit model so why u use different ar and ma terms and on which basis u select these term as u take (1,2,4) in fitB. fitC as (5,1,4). Please guide me without using auto arima fn how can we select our model ? And if we choose on the basis of acf and pacf then whats the procedure.highly greatful.
The auto arima gives you best fit based on a pre-programmed quick analysis. This is okay, but we all know that a custom arima model when tweaked correctly can yield a better fit and higher accuracy. That is what we do in some of the other videos also. Please watch them and you will see this for yourself. Here is a great one to start with. ua-cam.com/video/UekOBfpu8m8/v-deo.html
thank you so much, this tutorial very helpful to me.. you save my college grade, |Sir :)) I promise I'm gonna subscribed and like your every "statistics coding" on your UA-cam content. Aww.. can't wait for another helpful content :D :D
Thank for the video, but i have de this error. Can you help me? > #Test findings on original XTS objects > #ADF test for p-value > print(adf.test(DJI_Close_Prices))#p-value = Error in adf.test(DJI_Close_Prices) : NAs in x
Hello Sir. What can i do to lower the p-value and as a consequence the MAPE of the forecast? You said something about diversification but i don't know what you mean by that. I have used the Data of the last 25 years of one Stock and got p-value of 0,4038 and a MAPE of the forecast of 98,6% of my auto arima. I want to predict it for the next 252 days.
The problem there is predicting stock price values for 252 days. I would say anything beyond 2 or 3 months is going to have less accuracy and should only be used in a directional, trending manner. There are so many variables that can affect the stock market and individual stocks. Even well diversified ETF's can become highly variable. Outside circumstances and news like the corona virus and the economy in China can play a huge role. Even though we can do long term predictions, I would stick with a shorter prediction. To lower the p-value you can try differencing it. Look at my other videos on custom arimas as they cover this well. :)
It all depends upon your data. That is why I have a deeper, more thorough process that does mord testing anc ix mord advanced in the videos on thd channel. Look for the multi-part arima video series. :)
There is no file to download or send with this one. If you load the libraries in the video and then add that one line of code you can then get any publicly traded stock data for the period of time you choose. Please re-watch and look at the first 5-10 lines of code I show. Thank you. :)
Awesome, what a succinct and to the point tutorial.
Thank you. I am going to be creating more of these videos on forecasting shortly and will be using individual companies and their respective stock symbols. What company or companies would you like to see forecasting with (AAPL, GOOGL, MSFT, AMZN, FB, BRK-B, etc...)?
Great video!
I'm waiting the other methods.
I have the code all done for those videos. Now it's just getting the time to make the videos and upload them. Finding the time to make these videos is not always easy when everyone knows you are a data scientist and are a great solutions resource. I will get them up as soon as I can. Thanks again. Also, be sure and check out all the other great videos on my channel that cover everything from forecasting, predicting, regression, custom arima models and more. :)
Hey, do you have an email address or a good forum to recommend ? I work for a contact center and I ve been struggling with forecasting my call volume. there is too much variances in my data. I need some help :(
Thanks for this video, really helped me a lot to develop the code based on your methods
Glad you liked the video. :)
Great video.. thanks!
Could you share the code you used?
never worked with R ... interesting ... thank you
How do you change the time axis to correspond to the actual time variable? Thanks!
Thanks for the video. Why did the auto arima suggest 3 AR lags (p=3) when the pacf test only indicated 1? And which should you trust more?
Every data set is different. This is why we do many tests. Then, in the end, you need to just look at it, take a step back and ask, "Does this make sense?" In the end, your read and opinion is what these companies will pay the big money for. Sometimes you need to hold back and rerun the process. Sometimes you need to add more external or internal data for a better read. It all depends on the data, situation, date ranges, outside events and more.
Great video, helped alot!
Glad it helped!
Hello, I am facing difficulty when using data from yahoo finance, how to input other data instead of'SPY'. Is it done with the help of getsymbols function, or what am i doing wrong here. Can anyone help?
Yes. Look at the other similar videos. I recently just published one on the Pakistan ETF (PAK). Any stock ticker symbol that has data in Yahoo Finance will work. Some penny stocks may not have data there, so keep that in mind.
Thanks for the beautiful video, but from my experience I think that the subject of expectation is a very difficult subject. We spend a long time learning about the model then finding the proper one, testing it and forecasting it, this is a nice game, but it doesn't bear fruit. Political and economical conditions and investor sentiment affect at least 90% of the short-term or long-term pricing, leaving only 10% for the expectations and the models that are often unconvincing.
That is why, if you watch the entire video and any of my other forecasting video series, you will see that I tell people to look at it from a more holistic, directional approach as there are numerous factors at play that we cannot control for. I teach to look at it from a longer term perspective also. Look at it as if it is trending up or down rather than an exact price.
You are a wizard.
Thank you. I am glad you liked the video. Be sure to subscribe and click the bell so you get notified every time I create and publish great new videos like this one.
HI! If I want to include the high, low, open or other variables in the model? What shall be changed in the model/approach?
Sir when i run forecast for fitA it gives me result but when i run forecast for fit B it gives me error in attr(data,"tsp") what does it means kindly guide me .
Your data most likely has formatting issues or similar. Please read this - r.789695.n4.nabble.com/How-to-setup-the-tsp-attribute-of-a-dataset-td908269.html
Very informative Sir, kindly tell me when we fit autoarima fn it gives us best fit model so why u use different ar and ma terms and on which basis u select these term as u take (1,2,4) in fitB. fitC as (5,1,4).
Please guide me without using auto arima fn how can we select our model ? And if we choose on the basis of acf and pacf then whats the procedure.highly greatful.
The auto arima gives you best fit based on a pre-programmed quick analysis. This is okay, but we all know that a custom arima model when tweaked correctly can yield a better fit and higher accuracy. That is what we do in some of the other videos also. Please watch them and you will see this for yourself. Here is a great one to start with. ua-cam.com/video/UekOBfpu8m8/v-deo.html
Thanks alot Sir
thank you so much, this tutorial very helpful to me.. you save my college grade, |Sir :)) I promise I'm gonna subscribed and like your every "statistics coding" on your UA-cam content. Aww.. can't wait for another helpful content :D :D
Glad my tutorials on data science were helpful to you. Please feel free to share this channel and videos so that others can benefit. Thanks again. :)
You're a legend...
Superb Video Sir, But when I run forecast for fitA it gives me a constant result for 100 days ( same value for all 100 days ) How can I solve this ?
Thank for the video, but i have de this error. Can you help me?
> #Test findings on original XTS objects
> #ADF test for p-value
> print(adf.test(DJI_Close_Prices))#p-value =
Error in adf.test(DJI_Close_Prices) : NAs in x
I have the same problem and i don't know what should i do. Can you help me?
May I know how to get the ARIMA(p,d,q)... how to find the value of p d and q?
Testing models, ACF, PACF
Hi I code as u for accuracy but I got error as argument predicted is missing with no default
What is do
Pls give the solution
You can post your data here or at StackOverflow. Someone will be able to help you. Look at your datatypes and what you are trying to do.
Hello Sir. What can i do to lower the p-value and as a consequence the MAPE of the forecast? You said something about diversification but i don't know what you mean by that. I have used the Data of the last 25 years of one Stock and got p-value of 0,4038 and a MAPE of the forecast of 98,6% of my auto arima. I want to predict it for the next 252 days.
The problem there is predicting stock price values for 252 days. I would say anything beyond 2 or 3 months is going to have less accuracy and should only be used in a directional, trending manner. There are so many variables that can affect the stock market and individual stocks. Even well diversified ETF's can become highly variable. Outside circumstances and news like the corona virus and the economy in China can play a huge role. Even though we can do long term predictions, I would stick with a shorter prediction. To lower the p-value you can try differencing it. Look at my other videos on custom arimas as they cover this well. :)
where can i get the code
Can you add the script in the description?
Hi. I just want to point out that your data is not stationary and you have not done difference yet at 6:47 min
Could you please share the code file for the same
data is not stationary in the mean... didnt take D1
It all depends upon your data. That is why I have a deeper, more thorough process that does mord testing anc ix mord advanced in the videos on thd channel. Look for the multi-part arima video series. :)
could you send me that file R studio to my email sir ?
There is no file to download or send with this one. If you load the libraries in the video and then add that one line of code you can then get any publicly traded stock data for the period of time you choose. Please re-watch and look at the first 5-10 lines of code I show. Thank you. :)
@@techknowhow4802 but it will be better if you share code / github file