0
$\begingroup$

Suppose I have a dataset where I want to give input of up-to 30 lines (can be variable with min 5 available) these lines are timestamps of every minute and want my model to predict output of next 5 lines(the output for training is also available) which model should be used or possible and how should I fit my data for the model?

Basically if you still don't get it I want to get my model to predict the value for a column for the next 5 mins from the input:

Example Input:

             Datetime       SCP      CPS%   ExtP1   ExtP2    ComFut
  2022-01-07 11:16:00  37890.75      3.55  439.05  414.75    104.30
  2022-01-07 11:17:00  37892.50      5.53  441.75  411.10    109.30
  2022-01-07 11:18:00  37919.95      5.37  426.80  425.00    112.75
  2022-01-07 11:19:00  37948.50      5.82  411.50  437.35    114.45
  2022-01-07 11:20:00  37928.40      5.48  421.90  428.40    108.90
  2022-01-07 11:21:00  37908.15      6.14  434.05  417.10    108.30
  2022-01-07 11:22:00  37916.25      5.49  428.80  422.75    113.80
  2022-01-07 11:23:00  37909.85      6.19  433.85  418.40    114.85
  2022-01-07 11:24:00  37893.55      4.50  439.35  414.00    104.50
  2022-01-07 11:25:00  37891.10      5.14  440.95  410.50    108.40
  2022-01-07 11:26:00  37899.80      7.81  441.80  409.60    117.00
  2022-01-07 11:27:00  37915.85      7.85  432.80  417.15    124.10
  2022-01-07 11:28:00  37934.45      7.58  420.35  425.20    120.55
  2022-01-07 11:29:00  37919.75      4.76  424.95  425.40    111.60
  2022-01-07 11:30:00  37920.05      5.95  426.45  422.55    112.50
  2022-01-07 11:31:00  37924.50      8.46  427.35  418.50    121.50
  2022-01-07 11:32:00  37937.50      9.03  422.60  425.10    122.85
  2022-01-07 11:33:00  37946.80      5.50  410.35  435.75    120.75
  .
  .
  .
  2022-01-07 11:42:00  37886.25      4.35  440.00  407.90    103.85
  2022-01-07 11:43:00  37887.15      5.18  441.75  407.15    108.85
  2022-01-07 11:44:00  37890.80      6.70  442.75  405.75    115.45
  2022-01-07 11:45:00  37895.75      5.17  437.15  411.40    106.25

Example Output:

             Datetime       SCP
  2022-01-07 11:46:00  37793.10
  2022-01-07 11:47:00  37787.45
  2022-01-07 11:48:00  37799.25
  2022-01-07 11:49:00  37765.70
  2022-01-07 11:50:00  37718.10

I have problem with using the variable number of inputs fitting to the models I've tried using many ML models but its to no avail I get the error saying the shape don't match for the Input and output. What should I do and use in this case and what models are better for my case? And how should I approach variable number of input and output?

$\endgroup$
1
  • $\begingroup$ Classical starting point for extrapolation would be ARIMA. Maybe this already suits your problem. $\endgroup$
    – Broele
    Jun 9, 2023 at 14:33

1 Answer 1

0
$\begingroup$

This seems like a time series problem. So I suggest you look into statistical methods like ARIMA. You could also try using the prophet library.

Otherwise, you could also try deep learning approaches using LSTM models.

The choice of model would also depend significantly on whether your data shows trend and seasonality, whether you want light or complex models.

$\endgroup$

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.