Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Stock Price Prediction with LSTMs

miguelgfierro.com

Post reply on HN
Print view View original

11–20 of 20 posts

Re: Stock Price Prediction with LSTMs

#11

Post by highd »

This model is showing predictions one day into the future. The "test set" plot is all predictions made with data from 1 day ago. The input sequences have size 1, so no recurrence is happening (see to_1dimension in https://github.com/miguelgfierro/sciblog_support/blob/master...). If you predict that tomorrow will have the same price as today you'll get better plots under the same operating conditions.

In[2]: TIME_AHEAD = 1

Train set has ~1e-6 MSE, Test set has ~0.8 (0.94^2).

EDIT: I should say this person is probably learning, and a lot of this is honest mistake.

Re: Stock Price Prediction with LSTMs

#12

Post by m3kw9 »

m3kw9 wrote: post #7

The model could be trained in such a way that it will only work with historical data and pass only back tests accurately. This is because you are training against the “answer” which is already known.

Training with trends and patterns is like a technical trader looking for patterns. When things go right they attribute to the pattern, but when it’s wrong, they attribute it to unforeseen events. The unforeseen events is something if can be predicted, you will be rich lol

Re: Stock Price Prediction with LSTMs

#13

Post by claytonjy »

m3kw9 wrote: post #7

The model could be trained in such a way that it will only work with historical data and pass only back tests accurately. This is because you are training against the “answer” which is already known.

it is; by scaling his single variable before making a test-train split, he's allowed his model to know how the future is distributed.

Re: Stock Price Prediction with LSTMs

#15

Post by bcheung »

Cool to see that you created a notebook and are publishing this like a paper. I was thinking of doing some projects and this looks like a good format to follow.

However, this looks to be grossly overfitting. You can't just randomly drop out samples in a time series and use those for the test dataset. You need to cut out larger contiguous sequential time ranges and reserve those for your test set. Probably a single contiguous time window.

Anyone can predict with a high degree of accuracy what stock price is given the last 5 days and the next 5 days.

Predicting a few randomly dropped out pixels in that graph is really easy due to the nature of time series. You can just interpolate it.

Re: Stock Price Prediction with LSTMs

#16

Post by highd »

bcheung wrote: post #15

Cool to see that you created a notebook and are publishing this like a paper. I was thinking of doing some projects and this looks like a good format to follow. However, this looks to be grossly overfitting. You can't just randomly drop out samples in a time series and use those for the test dataset. You need to cut out larger contiguous sequential time ranges and reserve those for your test set. Probably a single co…

It's actually using just one day's data (one point) to predict the next day's data, let alone predicting 5 days.

Re: Stock Price Prediction with LSTMs

#17

Post by bcheung »

p1necone wrote: post #4

According to the graph this seems to basically perfectly predict future stock price. I don't believe it's possible to be that accurate at all, what am I missing here?

It's because the test dataset involves randomly choosing samples spread throughout the time series.

Imagine dropping 10% of the pixels randomly spaced out in a historical graph and then try to fill them in. It's trivial because you can just average the previous and next sample and have an extremely high accuracy.

Re: Stock Price Prediction with LSTMs

#18

Post by claytonjy »

bcheung wrote: post #17
p1necone wrote: post #4

According to the graph this seems to basically perfectly predict future stock price. I don't believe it's possible to be that accurate at all, what am I missing here?

It's because the test dataset involves randomly choosing samples spread throughout the time series. Imagine dropping 10% of the pixels randomly spaced out in a historical graph and then try to fill them in. It's trivial because you can just average the previous and next sample and have an extremely high accuracy.

I don't think that's true, because of the `shuffle = False` argument to `train_test_split`; it's just chopping the tail off for the test set.

Re: Stock Price Prediction with LSTMs

#19

Post by turingcompeteme »

antirez wrote: post #10

Actual way this experiment should work: feed the recurrent NN with stock markets data from the latest 10 years but the latest year. Then display the prediction for the latest year and the actual data. I want to see if it will match like that...

Predicting for only the last year is good, but there are two even better ways:

- An expanding window where you train on the first year of data and predict the second. Then train on the first two years and predict the third, etc.

- A rolling window where you train on years 1 and 2 and predict 3. Then train on 2 and 3 and predict 4, etc.

You need to show that your predictions work for any time period, not just the past year.

Re: Stock Price Prediction with LSTMs

#20

Post by bduerst »

antirez wrote: post #10

Actual way this experiment should work: feed the recurrent NN with stock markets data from the latest 10 years but the latest year. Then display the prediction for the latest year and the actual data. I want to see if it will match like that...

You'll see things like the global recession recovery and other major outliers skewing your data.
Post reply on HN