236

Forecastability You are right that this is a question of forecastability. There have been a few articles on forecastability in the IIF's practitioner-oriented journal Foresight. (Full disclosure: I'm an Associate Editor.) The problem is that forecastability is already hard to assess in "simple" cases. A few examples Suppose you have a time series like ...


55

The answer from Stephan Kolassa is excellent, but I would like to add that there is also often an economic stop condition: When you are doing ML for a customer and not for fun, you should take a look at the amount of money the customer is willing to spend. If he pays your firm 5000€ and you spent a month on finding a model, you will loose money. Sounds ...


8

There is another way. Ask yourself - Who or what makes the best possible forecasts of this particular variable?" Does my machine learning algorithm produce better or worse results than the best forecasts? So, for example, if you had a large number of variables associated with different soccer teams and you were trying to forecast who would win, you ...


8

Here's a second idea based on stl. You could fit an stl decomposition to each series, and then compare the standard error of the remainder component to the mean of the original data ignoring any partial years. Series that are easy to forecast should have a small ratio of se(remainder) to mean(data). The reason I suggest ignoring partial years is that ...


8

This is a fairly common problem in forecasting. The traditional solution is to compute mean absolute percentage errors (MAPEs) on each item. The lower the MAPE, the more easily forecasted is the item. One problem with that is many series contain zero values and then MAPE is undefined. I proposed a solution in Hyndman and Koehler (IJF 2006) [Preprint ...


7

Parameters m and r, involved in calculation of approximate entropy (ApEn) of time series, are window (sequence) length and tolerance (filter value), correspondingly. In fact, in terms of m, r as well as N (number of data points), ApEn is defined as "natural logarithm of the relative prevalence of repetitive patterns of length m as compared with those of ...


5

You might be interested in ForeCA: Forecastable Component Analysis (disclaimer: I am the author). As the name suggests it is a dimension reduction / blind source separation (BSS) technique to find most forecastable signals from many multivariate - more or less stationary - time series. For your particular case of 20,000 time series it might not be the ...


2

Simulation can help you here. There won't be a one-size-fits-all answer. Specifically, I would recommend that you take your time series, model it and calculate your point forecasts. Then simulate future unconstrained demands that are consistent with your model, especially your model estimate of residual variance. (Note that this will almost certainly ...


1

The first thing to look at is what happened in months 10 and 11: where did the large increases come from? Has the product reached a plateau? If so, essentially forecast a flat line on the level of sales in months 11-17. Or is another increase of a similar magnitude likely? Or was this a one-time push that saturated the market, so that future sales will drop ...


1

Your approach won't work. To convince yourself imagine that you had only a single value and repeated it one thousand times to have a sample of one thousand. Would single value repeated one thousand times have any additional information value? Actually many software packages will throw errors when encountering such zero-variance data. You can also check this ...


Only top voted, non community-wiki answers of a minimum length are eligible