วีดีโอ

also when predicting Future values you need to time shift your Target column backwards so say you have daily data and you want to predict tomorrows closing price you need to move the target column back one so that the input features are aligned with the future values to be predicted it will make a nan at the end of the dataframe but just remove the first 3 rows and last 3 rows of any dataset anyways lmaooo

the SEC edgar website has the all of the publicly listed us and international companies that file with them and all of their financials in a bulk data download setup its like 27Gigs of xml or json i forgot what but you can extract to csv or parquet and do some ML on that to get a more general ai system. Alternatively you can pretrain on all of the stock market financial data and then do some clustering based on log returns or some other metric to rediscover market segments and then to fine turning for each sector. Like one model that has seen the entire market and then a sector specific one and maybe do a weighting for them. Big data is your friend here too even if the predictions are only like 56% accurate if you do 5000 company predictions and only take the 90% confidence threshold the accuracy could be higher like 65% or something meaningful.

For something as high-dimensional, it would be cool to also see UMAP and tSNE plots of _all_ the normalized data.

Is it me or he does kinda look like mark ruffalo

Bro -98% hit me right in the soul … it’s as if I wrote it myself 😂😂😂. Funny how I understand the coding aspect but not the trading logic

I worked at GS for almost 2 years up until last December. I remember when this project was a baby. For a financial company, they take software engineering VERY seriously and they understand it's value. If you ever have an opportunity to work there, I recommend you take it.

Thanks.

what are you trying to achieve ? for me it is just code you compiled .am I beginner or you just show us how to run code nothing more ?

As DOW 30 is a price-weighted index, it can be more influenced by macroeconomic factors. Thus, our model should incorporate more macroeconomic-related features for better prediction power. Overall, I love your content. 🤩

I achieved a Annual return of 0.268% testing my quant script on the DOW 30 tickers, outperforming it by around 4% since it performs at about 0.226% using the same data and timeframe. However I didn't save the run! I am struggling to recreate the state conditions and code variables I chose during that performance. If you want to work together on this?

Hey! Click bait. First of all. The gs-quant is not open source. You need to authenticate to use it

hello. May i have the code please?

Open source my smooth hairless ass. Anyone up for a class action lawsuit for false advertising?

I talked to a developer on this project in my interview for Goldman it’s actually really interesting

Thanks for the nice video! IR stands for interest rate in the industry.

it works omg thank you so much

thanks!

They are just bragging about how far they have got in quantative trading, at the same method as we got Phi from Microsoft.

Always curious about the math ... found a seir model, creepy!

My Friend, Many Thanks for the great video of portfolio optimization using the Black Litterman Model. Can we perform the backtesting in this model like the one shown in your video of Mean Variance Portfolio Optimization, if yes can you please tell how? Thanks again.

Very interesting 😊

st louis fred data

Is it necessary an Account to have access to the Notebook featured ?

wow i didn't knew that

Nice , Can we have the google colab of this ?

haha! everyone complaining, 'useless without an account' WTF! "it's a big club and you ain't in it!" ...what do you think, there's no admission fees?

I really appreciate your work, I hope in future you will do a video for beginners on how to setup a deep learning model for trading from 0.

The FRED database is the Federal Reserve database. Its maintained in St. Louis if I am not mistaken.

IR stands for interest rates

You have to use midi_files in place of midi_file with pretty_midi. At 3:57

I have always had a question about mlagents: they randomly select actions at the beginning of training. Can we incorporate human intervention into the training process of mlagents to make them train faster? Is there a corresponding method in mlagents? Looking forward to your answer.

how to get a GS account? If not, nothing can be run

Gracias amigo :)

There are hidden markets at AV, but very difficult to find the mkt/symbols for...

So useless. can't even read the documentation without the GS account

Unfortunately you need a GS dev account seems , otherwise you can't use nothing

Random question here, not sure if you’ve thought of it. Why not try list the different adjustable parameters for the model and make use of Bayesian Optimization in order to try and squeeze the most possible performance out of the model ?

Pls do videos about kan model wave.kan chronosT5 bimamba &tiny time from ibm❤❤

hi,I got error at this part: while True: obs = obs[np.newaxis, ...] action, _states = actor_critic.predict(obs) obs, rewards, done, info = env.step(action) if done: print("info", info) break ----> 2 obs = obs[np.newaxis, ...] 3 action, _states = actor_critic.predict(obs) 4 obs, rewards, done, info = env.step(action) 5 if done: TypeError: tuple indices must be integers or slices, not tuple

I've had some good trades using a random forest classifier that just predicts if the change from todays close to tomorrows close will be a positive change or a negative change its only 62% accurate but if you're willing to give up some recall and only make guess 1 in a 1000 times you can boost the accuracy only betting when the model is very certain of the outcome. The downside is that it has no perspective on the amount of change so in back testing there are some volatility conditions i check to make sure the stock is moving and its not just getting a bunch of 0.01% moves right.

pretty useless, you need a password for most of the stuff.

how much does it cost? 👀

Love this men good code also

would be interesting to see how to get feature importance in the finrl ensemble model

Thank you!!!

Now how do you recommend those movies to users? I mean Each user has a different taste, how do you make recommendations like that, otherwise to every user that rated CATCHER IN THE RIE movie with a 9 lets say, will have the same recommendations?

After changing theme, the toolbar is gone. Do you know how to have toolbar with new theme?

Hints: 1. Big dataset. 2. Custom data preprocessing (standarized with lookback: close, change, d_hl)(No technical indicators derived from ohlcv). 3. Custom environment that you fully understand(account balance/positions value + data from 2. as observation/state space) 4. Custom model architecture that you choose. 5. Custom training loop with spawning in random place in your dataset for x steps. 6. Periodically testing on unseen data until profitable. 7. Long training.

It's open source, just read the code and modify it. If you don't understand it, maybe hire someone to write manual and additional funtions for you or don't put your money into it.