DASH + Predictions exported to CSV + Predictions to multiple tickers

[karenjh2]

Thank you for doing this. It is impressive.

I tested LSTM with a ticker whose df_stock begins in May 2006. The predictions were for dates beginning on May 11 2006.

This should be predictions for dates starting at the end of the series, so in 2021. I will attempt to debug this. I would really like to see it working.

[DidierRLopes]

Hey,

Thank you for the feedback :)

During development I had some issues where the dataframe didn't have the order I wanted. Meaning that if selected a starting date of 2006, it would have used data from 2021 until 2006, and predict from future to past...

I've done several tests around this, and thought I had fixed it by sorting these datetime index. Probably you did some sequence of commands that allowed to expose this again.

Do you remember what you did to trigger this?
Thanks!

[karenjh2]

User error. I did not do the sort. I called the LSTM function directly. Sorry but I find that initial menu hard to work with. Having done that, the predictions are in the future. Thanks. Karen Higgins From: DidierRLopes <notifications@github.com> Sent: Monday, March 1, 2021 3:11 AM To: DidierRLopes/GamestonkTerminal <GamestonkTerminal@noreply.github.com> Cc: karenjh2 <khig@earthlink.net>; Author <author@noreply.github.com> Subject: Re: [DidierRLopes/GamestonkTerminal] LSTM neural network predict for dates at beginning of series - 2006 (#62) Hey, Thank you for the feedback :) During development I had some issues where the dataframe didn't have the order I wanted. Meaning that if selected a starting date of 2006, it would have used data from 2021 until 2006, and predict from future to past... I've done several tests around this, and thought I had fixed it by sorting these datetime index. Probably you did some sequence of commands that allowed to expose this again. Do you remember what you did to trigger this? Thanks! — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <https://github.com/DidierRLopes/GamestonkTerminal/issues/62#issuecomment-787485267> , or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALFQFQF6YIJI4KHXEY54AK3TBJ2KBANCNFSM4YLHUBNA> . <https://github.com/notifications/beacon/ALFQFQDCZOWDAMRDF4RZZYTTBJ2KBA5CNFSM4YLHUBNKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOF3YBEUY.gif>

[DidierRLopes]

You should have not needed to sort it, unless you tried to changed the code. Because this should be coded.

Why did you find initial menu hard to work with? Do you have any improvements suggestions? Thanks!

[karenjh2]

I ran neural_networks.py, and did not use gameston_terminal.py. So I had to get the df_stock data and sort it. I found the menu hard to use. For me, it was difficult to get to where I wanted to go. Suggestions. A dashboard. The heading would be where the user enters the ticker, interval etc. and that would always be visible. The next level menus would be under that. Also there would be a way to add shortcuts to get directly to one or more favorite functions. Another suggestion is to be able to upload a list of tickers and obtain results for all of them in one go. For example, the LSTM function would return a df, or csv file, with those tickers, the predicted stock prices, the last adjusted close price, the percent increase over the last adjusted close price. One could then sort by percent increase descending to get the stocks with the largest predicted increase in price. Karen Higgins From: DidierRLopes <notifications@github.com> Sent: Monday, March 1, 2021 3:29 AM To: DidierRLopes/GamestonkTerminal <GamestonkTerminal@noreply.github.com> Cc: karenjh2 <khig@earthlink.net>; Author <author@noreply.github.com> Subject: Re: [DidierRLopes/GamestonkTerminal] LSTM neural network predict for dates at beginning of series - 2006 (#62) You should have not needed to sort it, unless you tried to changed the code. Because this should be coded. Why did you find initial menu hard to work with? Do you have any improvements suggestions? Thanks! — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <https://github.com/DidierRLopes/GamestonkTerminal/issues/62#issuecomment-787487981> , or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALFQFQAGQXC2TBWQMX6ZO6DTBJ4PTANCNFSM4YLHUBNA> . <https://github.com/notifications/beacon/ALFQFQFRCNG5H5V2A2WPKMTTBJ4PTA5CNFSM4YLHUBNKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOF3YBZ3I.gif>

[DidierRLopes]

I ran neural_networks.py, and did not use gameston_terminal.py. So I had to get the df_stock data and sort it. I found the menu hard to use. For me, it was difficult to get to where I wanted to go. Suggestions. A dashboard. The heading would be where the user enters the ticker, interval etc. and that would always be visible. The next level menus would be under that. Also there would be a way to add shortcuts to get directly to one or more favorite functions. Another suggestion is to be able to upload a list of tickers and obtain results for all of them in one go. For example, the LSTM function would return a df, or csv file, with those tickers, the predicted stock prices, the last adjusted close price, the percent increase over the last adjusted close price. One could then sort by percent increase descending to get the stocks with the largest predicted increase in price. Karen Higgins From: DidierRLopes notifications@github.com Sent: Monday, March 1, 2021 3:29 AM To: DidierRLopes/GamestonkTerminal GamestonkTerminal@noreply.github.com Cc: karenjh2 khig@earthlink.net; Author author@noreply.github.com Subject: Re: [DidierRLopes/GamestonkTerminal] LSTM neural network predict for dates at beginning of series - 2006 (#62) You should have not needed to sort it, unless you tried to changed the code. Because this should be coded. Why did you find initial menu hard to work with? Do you have any improvements suggestions? Thanks! — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#62 (comment)> , or unsubscribe https://github.com/notifications/unsubscribe-auth/ALFQFQAGQXC2TBWQMX6ZO6DTBJ4PTANCNFSM4YLHUBNA . https://github.com/notifications/beacon/ALFQFQFRCNG5H5V2A2WPKMTTBJ4PTA5CNFSM4YLHUBNKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOF3YBZ3I.gif

Got you, that makes sense then.

Really? I thought that the menu organization was quite good, thanks for the feedback. Let me know how do you think we could structure it better.

Such dashboard that you mention already exists. If you load a stock like: load -t TLSA -s 2020-06-04 you are able to go see Daily Stock: TSLA (2020-06-04) on top of the predictions menu.

Agreed with your suggestion of improvement, and we'll work on it in the near future :)

Thanks!

[DidierRLopes]

I've added some of the ideas into the ROADMAP file. Feel free to create a PR against the Roadmap if you feel something from here is missing! Will close this for now