1) The document discusses time series predictions using both statistical and neural network approaches.
2) With statistical models, the time series data must be made stationary by removing trends and seasonal variations in order to find the best fitting linear regression line for predictions.
3) Neural networks can directly model trends and seasons in time series data without prior data preprocessing, resulting in potentially more accurate predictions compared to statistical models.
1. 2018 IBM Systems
Technical University
August / 2018
São Paulo
Time Series Predictions
With Tensorflow
Paulo Queiroz
Systems Consultant
IBM Systems Lab Services