Can i use autoencoder for predicting time series missing data? I have time series data set of current and voltage at a regular interval of time there are some missing value . can i use autoencoder to predict the missing value? 
 A: Intervention Detection can be used to predict/replace missing values. see http://docplayer.net/12080848-Outliers-level-shifts-and-variance-changes-in-time-series.html
A downvote probably motivated by a lack of understanding of outlier detection procedures motivated me to be somewhat more pedantic and to explain in detail with an example.
Consider data shown here where period 4 is known to be 1140.0 .
 . Now let's consider the case where the 4th value is missing and we want to estimate it for completeness sake. Let's change the 4th value from 1140 to 0. 
 . We then use AUTOBOX which senses an unusual value of 0.0 at time period 4 and develops a useful model combining arima structure and latent deterministic structure.  Here is the model  .
A listing of the identified anomalies is here  where -1090 is the estimated coefficient at period 4 ...thus the replacement value is 0 - (-1090) or 1090 . Recall that the original true value was 1140.
The advantage of using Intervention Detection is that if there is stochastic memory in the data it can play a useful role in predicting the unknown/unrecorded values. Of course if there are large amounts of missing values there can be consequences.

