0
$\begingroup$

I'm trying to create model for prediction multiple correlated time series features. Issue is that input dataset consists of a number of "projects" with different duration and different categorical data.

For example input

+---------+-------------+----------------+-----------------+-----------------+----+----+----+
| Project | Week_number | Duration_weeks | Proj_metadata_1 | Proj_metadata_2 | A  | B  | C  |
+---------+-------------+----------------+-----------------+-----------------+----+----+----+
|       1 |           1 |              3 |             121 |            1121 | 10 |  5 |  8 |
|       1 |           2 |              3 |             121 |            1121 | 12 |  2 |  5 |
|       1 |           3 |              3 |             121 |            1121 | 15 |  0 |  4 |
+---------+-------------+----------------+-----------------+-----------------+----+----+----+
|       2 |           1 |              6 |             121 |            5121 | 55 | 12 | 21 |
|       2 |           2 |              6 |             121 |            5121 | 35 |  8 | 18 |
|       2 |           3 |              6 |             121 |            5121 | 42 |  4 |  2 |
|       2 |           4 |              6 |             121 |            5121 | 15 |  1 |  1 |
|       2 |           5 |              6 |             121 |            5121 | 68 |  2 |  1 |
|       2 |           6 |              6 |             121 |            5121 | 27 |  6 |  0 |
+---------+-------------+----------------+-----------------+-----------------+----+----+----+

Need to predict 'A','B','C' timeseries for a new "project", knowing "Duration_weeks" and taking into account "Proj_metadata".

Googled a lot, tried bi-directional LSTM, concatenating with categorical embedded features, but not succeed yet.

I'm thinking how to solve properly issue with different length of input "projects" (5..1000weeks) and how correctly apply categorical project data. Are there any good practices?

Appreciate any help or advice on this topic.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.