Neural Networks for Time Series Input: Coping with Missing Data | by Sara Nóbrega | January, 2025
Part 3: Find out how a simple Keras sequence model can work
One of the common problems in time series analysis of missing data.
As we saw in Part 1, simple estimation techniques or regression-based models like linear regression and decision trees can take us a long way.
But what if we need to handle more subtle patterns and hold subtle dynamics for complex time series data?
In this article, we will explore how a A Neural Network (NN) can be used to calculate missing values.
The power of their NNs the ability to capture nonlinear patterns and interactions in data. Although NNs are often expensive, they can provide a very efficient way to fit missing time series data in situations where simple models fail.
We will work with the same data set as in Part 1 and Part 2, with 10% missing values, randomly introduced into the pseudo power generation dataset.
Don't miss it Part 1 of this thread:
