lstm for predict

Machine Learning and Modeling
,
1

I have a data set that wants to predict s2 column by LSTM?
how can do it?

r     date       time        c   s1      s2
1   2020-06-01	00:00:00	42	74.5	73.0
2	2020-06-01	00:15:00	46	73.0	74.0
3	2020-06-01	00:30:00	50	75.5	76.0
4	2020-06-01	00:45:00	34	77.0	80.0
5	2020-06-01	01:00:00	29	81.0	80.0
6	2020-06-01	01:15:00	18	76.5	78.0
7	2020-06-01	01:30:00	22	79.5	77.5
8	2020-06-01	01:45:00	26	71.0	75.0
9	2020-06-01	02:00:00	16	76.5	81.5
10	2020-06-01	02:15:00	18	83.5	76.0
11	2020-06-01	02:30:00	8	71.0	74.0
12	2020-06-01	02:45:00	7	80.0	79.0
13	2020-06-01	03:00:00	11	71.0	74.0
14	2020-06-01	03:15:00	12	76.5	76.0
15	2020-06-01	03:30:00	9	70.0	75.0
16	2020-06-01	03:45:00	6	79.5	77.5
17	2020-06-01	04:00:00	8	77.5	80.0
18	2020-06-01	04:15:00	8	82.5	79.0
19	2020-06-01	04:30:00	13	76.0	81.0
20	2020-06-01	04:45:00	12	82.5	83.5
21	2020-06-01	05:00:00	12	86.0	82.5
22	2020-06-01	05:15:00	6	75.0	82.0
23	2020-06-01	05:30:00	25	84.0	82.0
24	2020-06-01	05:45:00	30	81.5	81.0
25	2020-06-01	06:00:00	38	81.0	78.0
26	2020-06-01	06:15:00	65	77.0	74.0
27	2020-06-01	06:30:00	65	73.0	72.0
28	2020-06-01	06:45:00	106	72.0	72.0
29	2020-06-01	07:00:00	96	71.0	72.0
30	2020-06-01	07:15:00	103	73.0	73.0
31	2020-06-01	07:30:00	111	73.0	72.0
32	2020-06-01	07:45:00	89	70.0	71.0
33	2020-06-01	08:00:00	150	71.0	70.0
34	2020-06-01	08:15:00	125	69.0	68.0
35	2020-06-01	08:30:00	171	67.0	69.0
36	2020-06-01	08:45:00	125	72.0	70.0
37	2020-06-01	09:00:00	121	67.0	66.0
38	2020-06-01	09:15:00	154	65.0	66.0
39	2020-06-01	09:30:00	149	67.0	68.0
40	2020-06-01	09:45:00	158	69.0	68.0
41	2020-06-01	10:00:00	167	67.0	68.0
42	2020-06-01	10:15:00	141	68.0	65.0
43	2020-06-01	10:30:00	138	62.5	65.0
44	2020-06-01	10:45:00	142	69.0	68.0
45	2020-06-01	11:00:00	120	67.0	68.0
46	2020-06-01	11:15:00	127	69.0	69.0
47	2020-06-01	11:30:00	160	69.0	69.0
48	2020-06-01	11:45:00	155	70.0	67.0
49	2020-06-01	12:00:00	133	66.0	68.0
50	2020-06-01	12:15:00	134	70.0	70.0
2

Rob Hyndman has a free ebook that you may find helpful:

Forecasting: Principles and Practice
Forecasting: Principles and Practice (3rd ed)

See Chapter 7 for TSLM examples.

3

Why would you suggest this ebook, since this chapter you referred to doesnt include examples for his actual question?

4

imo the best ressources to learn keras R are:

https://blogs.rstudio.com/ai/#Time_Series

5

Sorry, due to a combination of bad eyes and methodological bias I confused LSTM with TSLM.

Thanks for the link, looks like an interesting technique where a more unsupervised approach is required.

6

Thanks
What is best way for modeling and predict this dataset? Time series or LSTM?

7

I am by far not an expert for modeling, but for your dataset it may not need such a complex model. It seems that s1 column is slightly decreasing as time goes on and c initially decreases as well to a certain time point and then increases. The date column is irrelevant since its not changing at all. With a simple linear model and time, c and s1 as your independent variables, you should get good predictions.

8

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.

If you have a query related to it or one of the replies, start a new topic and refer back with a link.