Consider the following example dataset with 1-minute timestamps and 5 numerical columns:

Example dataset:

Timestamp            Column 1   Column 2   Column 3   Column 4   Column 5   Result
2023-06-20 09:30:00  10.5       50         0.75       100        8.2         NaN
2023-06-20 09:31:00  11.2       45         0.80       98         8.6         lower
2023-06-20 09:32:00  12.1       42         0.78       101        8.8         lower
2023-06-20 09:33:00  11.7       48         0.82       99         8.4         higher
2023-06-20 09:34:00  10.9       55         0.85       102        8.9         higher

In this dataset, each row represents a specific point in time, and the columns correspond to different measurements or variables recorded at that timestamp. Let's discuss the meaning and purpose of the "Result" column.

The "Result" column represents the comparison between the current value in Column 2 and the previous value. It indicates whether the current value is higher or lower than the previous value. The "Result" column is determined based on the following conditions:

  • If the value in Column 2 is higher than the previous value, the corresponding entry in the "Result" column is labeled as "higher".
  • If the value in Column 2 is lower than the previous value, the corresponding entry in the "Result" column is labeled as "lower".
  • For the first value in Column 2 (the earliest timestamp), there is no previous value to compare with, so the "Result" column contains a NaN (Not a Number) value or any suitable representation for missing values

I am seeking recommendations for packages that can effectively handle this task. Specifically, I would like to know if there are any recommended packages within Tensorflow or PyTorch that would be suitable for this prediction problem. Additionally, I would appreciate insights into the approach or methodology that could be employed to achieve accurate predictions.

Any suggestions, examples, or code snippets demonstrating the implementation using the recommended packages and approaches would be highly valuable. Thank you in advance for your assistance!


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.