Simple Moving Average (SMA) is one of the basic forecasting methods. It doesn’t rely on time series decomposition, doesn’t have a seasonal component, and doesn’t include explanatory variables. Yet, in a supply chain context, SMA is sometimes a tough benchmark to beat. Why? First things first, SMA is simply the arithmetic mean of several recent […]

# extrapolation methods

# What about the training/test sets?

Another question my students sometimes ask is how to define the sizes for the training and test sets in a forecasting experiment. If you’ve done data mining or machine learning, you’re likely familiar with this concept. But when it comes to forecasting, there are a few nuances. Let’s discuss. First and foremost, in forecasting, the […]

# How to choose forecast horizon?

One of the questions my students sometimes ask is how to set the forecast horizon. The answer depends largely on the task at hand, but there are still some guidelines. First, the forecast horizon depends on data granularity. A “year ahead” forecast on monthly data means forecasting 12 steps ahead, while for daily data, it […]

# Straight line is just fine

Look at the image above. Which forecast seems more appropriate: the red straight line (1) or the purple wavy line (2)? Many demand planners might choose option 2, thinking it better captures the ups and downs. But, in many cases, the straight line is just fine. Here’s why. In a previous post on Structure vs. […]

# Are all forecasts wrong?

You’ve probably heard the phrase “all forecasts are wrong”, suggesting that the future is unpredictable and that no forecast will ever match the actual outcome. Well, this phrase is not entirely correct, and here’s why. When your favourite forecasting approach generates point forecasts, it usually provides a conditional mean. This means it’s giving you the […]

# Multistep loss functions: Geometric Trace MSE

While there is a lot to say about multistep losses, I’ve decided to write the final post on one of them and leave the topic alone for a while. Here it goes. Last time, we discussed MSEh and TMSE, and I mentioned that both of them impose shrinkage and have some advantages and disadvantages. One […]

# Multistep loss functions: Trace MSE

As we discussed last time, there are two possible strategies in forecasting: recursive and direct. The latter aligns with the estimation of a model using a so-called multistep loss function, such as Mean Squared Error for h-steps-ahead forecast (MSEh). But this is not the only loss function that can be efficiently used for model estimation. […]

# Recursive vs Direct Forecasting Strategy

Have you heard about the recursive vs direct forecasts? There’s literature about them in the areas of both ML and statistics. What’s so special about them? Here is a short post. The term “recursive” forecasting refers to the approach, when we produce one-step-ahead forecast first, then use it to produce two-steps-ahead, three-steps-ahead, and so on. […]

# What’s wrong with ARIMA?

Have you heard of ARIMA? It is one of the benchmark forecasting models used in different academic experiments, although it is not always popular among practitioners. But why? What’s wrong with ARIMA? ARIMA has been a standard forecasting model in statistics for ages. It gained popularity with the famous Box & Jenkins (1970) book and […]

# Why you should not use Holt-Winters method

Whenever I see results of an experiment that include Holt-Winters method, I shrug. You should not use it, and here is why. Holt-Winters was developed in 1960 by a student of Charles Holt, Peter Winters (Winters, 1960). He extended Holt’s exponential smoothing method (the method that introduced a trend component) to include a seasonal component. […]