jays-code-collection / hmms_stock_market Goto Github PK

when i use this model, i got the accuracy about 50%( 31 true in 60 prediction). I want to ask about the accuracy about your test.
thanks~(maybe there are some wrong on my environment so that i can't use the code directly)

Hi, many thanks for this.

When running the code it fails always on the 2/5

Predicting future Close prices from 2023-01-18 00:00:00 to 2023-01-22 00:00:00
 40%|██████████████████████████████████                                                   | 2/5 [00:03<00:05,  1.97s/it]

seems that total_data contains NaN values and the score method don't like it. Any idea on how to fix this?

I have switched to downloading the data using yfinance, all the columns have no NaNs so it's not the root of the issue. Is there something I am missing that may cause the ValueError?

Training data period is from 2018-01-02 00:00:00 to 2021-05-18 00:00:00
2023-01-18 10:04:46,223 __main__     INFO     >>> Extracting Features
2023-01-18 10:04:46,223 __main__     INFO     Features extraction Completed <<<
Predicting Close prices from 2021-05-19 00:00:00 to 2023-01-17 00:00:00
100%|█████████████████████████████████████████████████████████████████████████████████| 419/419 [09:17<00:00,  1.33s/it]
All predictions saved. The Mean Squared Error for the 419 days considered is: 6.657134688795405
Predicting future Close prices from 2023-01-18 00:00:00 to 2023-01-22 00:00:00
 40%|██████████████████████████████████                                                   | 2/5 [00:03<00:05,  1.97s/it]
Traceback (most recent call last):
  File "/home/originn/HMMs_Stock_Market/src/stock_analysis.py", line 470, in <module>
    main()
  File "/home/originn/HMMs_Stock_Market/src/stock_analysis.py", line 465, in main
    use_stock_predictor(company_name, start, end, future, metrics, plot, out_dir)
  File "/home/originn/HMMs_Stock_Market/src/stock_analysis.py", line 354, in use_stock_predictor
    future_pred_close = stock_predictor.predict_close_prices_for_future()
  File "/home/originn/HMMs_Stock_Market/src/stock_analysis.py", line 248, in predict_close_prices_for_future
    predicted_close_prices.append(self.predict_close_price_fut_days(day_index))
  File "/home/originn/HMMs_Stock_Market/src/stock_analysis.py", line 223, in predict_close_price_fut_days
    ) = self._get_most_probable_outcome(day_index)
  File "/home/originn/HMMs_Stock_Market/src/stock_analysis.py", line 149, in _get_most_probable_outcome
    outcome_score.append(self.hmm.score(total_data))
  File "/home/originn/HMMs_Stock_Market/venv/lib/python3.10/site-packages/hmmlearn/base.py", line 259, in score
    return self._score(X, lengths, compute_posteriors=False)[0]
  File "/home/originn/HMMs_Stock_Market/venv/lib/python3.10/site-packages/hmmlearn/base.py", line 272, in _score
    X = check_array(X)
  File "/home/originn/HMMs_Stock_Market/venv/lib/python3.10/site-packages/sklearn/utils/validation.py", line 919, in check_array
    _assert_all_finite(
  File "/home/originn/HMMs_Stock_Market/venv/lib/python3.10/site-packages/sklearn/utils/validation.py", line 161, in _assert_all_finite
    raise ValueError(msg_err)
ValueError: Input contains NaN.

Shouldn't in def predict_close_price_fut_days(self, day_index) the self.days be incemented?

Pre-installation steps are not given.

The fix is as follows. Add to your import:

import yfinance as yf

then in the _split_train_test_data method under the try: change as follows:

yf.pdr_override()
used_data = data.get_data_yahoo(self.company, start=self.start_date, end=self.end_date)

Hello,

I'm trying to import stock data from a csv or xls file, avoiding yahoo, without success because I'm facing many errors and I'm almost new to Python code... In this case this is the example file, "stock.csv/xls". If it's possible could you please give me a tips to import stock data from local file? Thanks a lot in advance and sorry for the stupid question...
p.s plz don't blame me for my question, if it's not possible feel free to close the "issues".
Jean
stock.csv
stock.xls

previous_data_start_index = max(0, day_index - self.n_latency_days) previous_data_end_index = max(0, day_index - 1) previous_data = self.test_data.iloc[previous_data_end_index: previous_data_start_index]
previous_data_start_index is smaller than previous_data_end_index , so test_data.iloc[previous_data_end_index: previous_data_start_index] will be empty?

Running the following:

python stock_analysis.py -n AAPL -s 2018-01-01 -e 2020-12-15 -o ./appl_test/ -p True -f 8 -m True

Results in:

Graph shouldn't be showing "actual_close" for dates that haven't occurred yet, right?

I figured there had to be something crazy going on for the lines to be matching so close (like a bug/mistake). So I just ran the example for myself to see what the graph would look like.

I've yet to dig into the code. I've re-read the README a few times to see if I missed something.

Something else, I have no idea what the following means (I'm a complete newb to this):

-m stands for metrics, and determines whether or not to predict for the historical data, or just for the future days, if True all of the historical data in the test set will be predicted and the Mean Squared Error will be returned

I don't even know what to ask. Maybe my question is, what is there to predict in historical data?

Depend on which python version you are using you can use f-strings.

https://realpython.com/python-f-strings/#f-strings-a-new-and-improved-way-to-format-strings-in-python

You can use
save_dir = f"{out_dir}/{stock_name}_results_plot.png"
instead of this
save_dir = out_dir + '/' + str(stock_name) + 'results_plot' + '.png'

or
out_name = f"{out_dir}/{company_name}_HMM_Prediction_{round(mse, 6)}.csv"
instead of
out_name = out_dir + '/' + str(company_name) + '_HMM_Prediction_' + str(round(mse, 6)) + '.csv'

Simply looks cleaner.