Please consider the sample dataset below.
In simple terms,
An sensor is defective and hence measurement has incorrect values since 2000 and we have the data for 10 years with both: measured and actual.
Now, we want to have the actual values from the algorithm.
We tried, XGBoost and CatBoost by creating another column named diff = measured- actual
and fed to the algorithm to identify the pattern. but not sure which algorithm is appropriate although suspecting **Neural network or Time series (ARIMA)** could work but not sure
because we have just 10 years data on monthly level
```
library(tidyverse)
train_data
Maybe the R package introduced in the linked article can be of help.
https://link.springer.com/article/10.1007/s10661-019-7586-x
Cheers,
Rainer
To build a regression, first you need to define your inputs(features) properly.
Since it is time series problem and you have only measured and actual data, I recommend to convert your data similar to the following format:
Inputs:
measured t
measured t-1
measured t-2
actual t-1
actual t-2
Target:
actual t
(You need to convert each row iteratively)
Then you can try some regression methods. Since your data has only 120 instances, I recommend to start with simple linear regression.
Good luck.
Hi Ian,
you can try TensorFlow (deep learning neural net tools - Google (free)) with your data and study the predictive capability of your network. Then you can compare with ARIMA results.
Cheers
- Badges
- Science topic
- Similar topics
- Statistics
- Time Series