Forecast the price of a share in the stock exchange

student_compute

Hello
I want to predict the growth or decrease of the price of a share of a particular symbol in the stock exchange
For this purpose, I collected data for a symbol for a month
But I do not know how to predict the price of a week in algorithms such as svm?
Can someone help
Thankful

hughesfleming68

Take a look at the Neural Market Trends tutorials on time series modeling.  https://www.neuralmarkettrends.com/tutorials/

Thomas Ott covers a lot of material on this. They may be for an earlier version of Rapidminer but the concepts are the same. For newer versions of Rapidminer, you can look at the ARIMA tutorials.

Forecasting stock prices is very tricky. You would need to keep your expectations in check.

MartinLiebig

Hi,
check out this article: https://towardsdatascience.com/time-series-forecasting-for-optimal-gas-refill-88650b1cf029

It's basically the same thing.

BR,
MArtin

student_compute

Hello
Thank you both dear friends
I have a history of a specific symbol from the stock market for 6 months in an excel file
Columns with names
Open Closed . High . Low and so on.
But I do not know which should be labeled?
And how can I anticipate the price of that symbol for a month?
Thanks for giving me some experience
Thanks
Yours sincerely

hughesfleming68

You have some choices to make. The close is the obvious label but you could also look at the percentage change day to day. It is worth transforming the change in close values to a binary classification as it will allow you to use learners that might not be suitable for pure regression tasks. You will also have to consider how to make your time series stationary. That is a topic all by itself. It is worth reading up on machine learning regression, ARIMA, linear regression etc.

You would have better success trying to predict if the price will be higher or lower rather than an actual value. Stock prices are highly unstable.

student_compute

Hello

Thank you very much

Sorry i take your time And again I ask a question

(Should the data be preprocessed?)

look

I did not notice these sentences

"You also have to consider how to make your time series stationary."

And

"You could also look at the percentage change day to day."

And that

I considered the tag as "close" and then applied svm, but it has an error .. !!!!!! look

How can I report how stock icons will be for a week? Grows or decreases?

May I send a sample process to me?

I hope you understand my question.

And you can help me

Thanks

hughesfleming68

You are getting that error because the role is not set. Look for the set role operator and set the close values as the label. Did you look at the tutorial links? I think you should work through them. If your data is weekly data then forecasting the next week is just a matter of predicting the next value. If your data is daily data then you are actually looking at predicting five steps ahead. That is a different problem.

student_compute

Hi, I want to change the tag. But in the  "set role"  operator, none of the fields will be displayed until I select. Actually empty.

And I wrote it myself in the "close" field manual. And was given an error in the run.

I used the tutorials I mentioned. I have these two problems

look


please help me..
tanks

hughesfleming68

Did you confirm that you are actually getting data from Yahoo? The last time I tried this it was broken. I suggest that you download the data to a csv file and read from the csv rather than use the Finance and Economics extension.There are couple of different workarounds for this. My preferred way is to use the Execute R operator and use the quantmod package. That is well supported and easy to use. It will also give you some flexibility with regards to how you adjust the data for splits and dividends.

My guess is that you are running into trouble because you are not actually bringing in any data. In the future, post the xml of your process here. It would have been easier to help you. There is also a tutorial on how to bring in data from Alpha Vantage. I hope that made the transition to the new forum.

regards,

Alex

student_compute

Hello
I collected data from a stock market symbol for six months
I sent an example of my data
How should I use the site's training to predict the price of the symbol over the next week?
Can anyone help me?
I sent my sample data as well as site training in this field
Thankful

hughesfleming68

For starters your data is not in an ideal format. All you need is Date,Open, High, Low and Close starting from the oldest date to the newest. Learn how to predict one day ahead. Understand windowing and sliding window validation. I looked at your process. Rather than try and covert daily data to weekly data, start with just daily data and nothing else. Add additional attributes later to see what effects they have on your prediction accuracy. You should take a look at the tutorial links above and work from there. Know the difference between univariate predictions and multivariate.

student_compute

Hello
Dear Professor
I'm really confused. And I'm totally new in this regard.
Do you have that? Send me an example of my prediction?
I know there is a lot of demand. But now I have been involved with this topic for three weeks and I have no results ...  
Thankful  

hughesfleming68

Do this tutorial. https://www.neuralmarkettrends.com/rapidminer-5-0-video-tutorial-9-financial-time-series-modeling-part-1/

When you are finished, do the next one. Don't just look at the video, sit down and reproduce it. I have read many of your 43 posts and it is not clear to me that you understand the basics. Go step by step and see what each operator does. You can ask me questions via private mail and I will help you. There is no need to make this thread any longer.

student_compute

Thank you so much
Sure
I do and I ask in private
Thank you