### Winning with Simple, not even Linear Time-Series Models

Today, we want to consider almost trivially simple models. If your dataset is small, the subsequent ideas might be useful.

Today, we want to consider almost trivially simple models. If your dataset is small, the subsequent ideas might be useful.

Let's make GARCH have varying coefficients to handle non-linear conditional variance.

While point forecasts are very popular, be aware of some unlucky pitfalls.

Point forecasts are good for making decisions. With probabilistic forecasts, you can also make the right ones.

Tree models are not just useful for point and mean forecasts.

Random Forests are flexible and powerful when it comes to tabular data. Do they also work for time-series forecasting? Let's find out.

One primary limitation of GARCH is the restriction to a single dimensional time-series. In reality, however, we are typically dealing with multiple time-series.

You have heard about integrated time-series data but what about cointegration?

Facebook Prophet is highly popular for time-series forecasting. Let me show you why I am not a big fan and what else you can use.

CUSUM is arguably the simplest algorithm for change point detection problems as in IoT or finance applications.

A quick Jupyter notebook about LSTMs and Copulas using tensorflow probability.

Who says that the GARCH conditional distribution needs to be Gaussian?

Yet another ARMA time-series model for non-normal data.

Bayesian inference with less pain and even less code.

Personal thoughts about a somewhat controversial paradigm.

How Gaussian Process regression can handle varying variance.