Verify/Correct Understanding of Trues/Preds in Test for Batch

[arkemp11]

Good work! Thank you for sharing the code to reproduce your paper.

For batch_size=16 and seq_len=10 for preds and trues in exp_model.test and denormalizing the close price values for a stock where:

true_values = denorm_trues[15, :, 0]
pred_values = denorm_preds[15, :, 0]

For the tail batch_size in the input csv results in:

True Values: [77.65619  78.595825 79.35186  77.06217  78.23942  79.63267  79.27625
 78.97385  80.82072        nan]
Predicted Values: [78.42562  79.20458  79.50893  79.432274 79.89651  77.55834  77.40948
 77.96623  77.003975 78.728676]
Dates:[2024-02-16  2024-02-20  2024-02-21  2024-02-22 2024-02-23  2024-02-26  2024-02-27
 2024-02-28  2024-02-29 2024-03-01]

Is it correct to say that on 2024-02-16 when the stock price was 77.65619 the predicted price for 2024-03-01 is 78.728676?

Or is it that on 2024-03-01 the predicted price is 78.728676?

If not, how do I correct my understanding to get a future predicted price from trues and preds?

FYI, for this example I edited the data_loader for test to this:

        if self.set_type == 2:
            if self.seq_len == 10:
                start_idx = 760
                end_idx = 796
            elif self.seq_len == 20:
                start_idx = 740
                end_idx = 796
            elif self.seq_len == 40:
                start_idx = 700
                end_idx = 796
            elif self.seq_len == 60:
                start_idx = 660
                end_idx = 796

            border1s = [0, num_train-self.seq_len, start_idx]
            border2s = [num_train, num_train+num_vali, end_idx]
            border1 = border1s[self.set_type]
            border2 = border2s[self.set_type]

Thank you in advance.

[koa-fin]

I am not sure what you are trying to achieve.
The predicted values, ground truth values, noisy values, and input values can all be obtained here:

dva/exp/exp_model.py

Lines 175 to 178 in 3c0bbde

	preds = np.array(preds)
	trues = np.array(trues)
	noisy = np.array(noisy)
	input = np.array(input)