Видео

Does this model work for "long term" predictions? For example, we are training with 2000 data points aprox. Can we predict the next 100 for example? Without using the real ones, which will adjust the predictions

Well done!

Hello...Can you provide an example of how to apply DQNs to the Taxi-V3 environment? I realize that such a small grid would not require a DQN, however I believe it should still be possible. Any assistance would be greatly appreciated!

can i get feature importance like in "fbprophet : regressor_coefficients(model)" in neural frophet

How is it possible that timestep is 30 and the LSTM layer is 64. Shouldn't it be less than timestep to actually encode it?

Hello, May I ask, if I want to predict sample by sample based on the previous state of the output and one exogenous input is that possible to do with ARIMAX or SARIMAX, if yes how can I train them to do so.

Hi, thanks for code about NeuralProphet.It's really good, than i see before.But it gives only one day next predict.Can we make it more days.

One can try also the LSTM model but then create the data sequences themselves. One can get rid of the outliers or the anomalies use the quantile method which squeezes the data points at the tail of distribution to lie within the range of data (standard deviation). One can then check the outliers from the boxplot.

Hi, should check the data stationarity first through adfuller test, then use the differencing to find parameter (d) and ACF and PACF for (p,q). one should use the holdout method to create train and test set, then validate the model against the test data. if the prediction agrees with the test set within the confidence interval (95%), then one can extrapolate to one month for example and see the predictions. Also, use different metrics to validate your model (mae, mse, corr, etc).

pretty handy and useful tools, especially the last part which integrates data customization with code simultaneously. For visualization I prefer Matplotlib, Seaborn and Plotly

Hi, one can plot the auto-correlation and partial-correlation to extract the parameters required for the ARIMA model (p,d,q). You can also use Grid-Search to find the optimal parameters for your score.

"promo sm" 😠

There is a fundamental issue in this code. Autoencoder output is (30,1) while the y train is just one value ( the 31st day closing value for every 30 days ). The autoencoder should not be trained with X_train, y_train. It should be X_train, X_train as AE tries to reconstruct the input, not any forecasting. Here you have mixed both.

In case my signal has N dimensions (86 in my case) instead of 1 (like in the video), how do I compute the MAE?

please do a tutorial of this with google collab. trying to do this in google collab. am having trouble

Firstly thanks. My question is that when input is 30*1 means 30 then how can be output 64 while in autoencoder we compress data then decode for example 30 to 15 to 10 then decode

Hi, I want to implement your code in my env, but not found We could talk, if you know how to fix my mistake? Thanks you.

Helping

what variables/model did u use to predict the closing price?

Nice.

While pip installing the neural prophet I'm getting errors

hello how are u where I can find this code history = model.fit(x_train, y_train, epochs-epochs, batch_size-batch_size, validation_split=0.1, callbacks = [EarlyStopping(monitor = 'val_loss', patience=3 ,???????

I like it better than Chat GPT. Wish it was more phone friendly.

Thank you, when I run this from my PyCharm the Chrome Browser is being opened and the conversation is run on the browser. What am I doing wrong?

This method is not working :(

Where is the code exactly? Thanks!

Thanks for the video!

Sehr Interessant

Sehr gut, so etwas habe ich lange gesucht und endlich gefunden. Danke!!

Super

I will include the YOLO model in my outdoor cam to spy the neighborhood

An interesting topic

It seems that only one dataframe was added, not 365, in the future prediction part code. Isn't 365 length inferred by adding n_forecast=365 in the model setting part of execution box 26?

how to show the actual result and prediction result ?

please do SAC too

Part 2

Sorry, but it looks like overfitting. Model result is very similar to last price:(

I have a question about the Xgboost algorithm. The question is how parallelization works in the Xgboost algorithm and explain me with an example.

Cool!

good

As always interesting:)

Thank you so much, your videos are really good and easy to understand 👍, please if have enough time try to cover the custom environments, it would be very helpful 👍🙏

Interessant!

Sehr gutes Video

it will be more useful if it was tweet analysis

Wheres' the csv file used in the project?

Interesting topic:)

Can you share your code? URL? Thanks

Nice