Enhanced visualization for multivariate columns

[skmanzg]

I tried to fix the visualizing and saving process for convenient purposes.

Previous:

• Multivariate predicted columns are not visualized but the last column
• Since the name of the column is not available in the result, User had to figure out which columns is visualized
• Metrics are calculated for the last column only from random samples, which makes it less practical
• The user had to choose various samples from saved npy files and could not know which one was used for initial visualization.

Now:

• Multivariate predicted columns are visualized naturally not to mention the case of univariate prediction.
• The column's name appears in the result pdf to make it easy to interpret.
• Metrics are calculated for each column and recorded for each when visualization. The text of 'mse, rmse, mae' appears for practical purposes instead of using dtw.
• The user can use core_true.npy and core_pred.npy to reproduce visualized result while pred.npy and true.npy files are saved separately.

Issue: there might need a test for unexpected bugs because of this code updated

• 'exp_long_term_forecasting.py' is the only applicable and that means, updated code might not work for the rest.
• self.args.use_dtw does not appear in the training process nor metrics.npy. Namely, the code below is disconnected.

        if self.args.use_dtw:
            dtw_list = []
            manhattan_distance = lambda x, y: np.abs(x - y)
            for i in range(preds.shape[0]):
                x = preds[i].reshape(-1,1)
                y = trues[i].reshape(-1,1)
                if i % 100 == 0:
                    print("calculating dtw iter:", i)
                d, _, _, _ = accelerated_dtw(x, y, dist=manhattan_distance)
                dtw_list.append(d)
            dtw = np.array(dtw_list).mean()
        else:
            dtw = 'not calculated'

I have tried to preserve the original code and features as much as possible when I update except for this part. I honestly do not know how to handle this.

[wuhaixu2016]

Hi, you definitely do a great job. Thanks for your contribution. However, it will take some time for us to check the committed code. We will merge this asap.