Reputation
13,312
Top tag
Next privilege 15,000 Rep.
Protect questions
Badges
1 19 58
Newest
 Nice Answer
Impact
~357k people reached

4h
revised Sales forecasting with non-stationary data
added 8 characters in body; edited tags; edited title
5h
answered Sales forecasting with non-stationary data
1d
comment DBSCAN: What is a Core Point?
The picture is the same as on Wikipedia. There, the figure caption says that MinPts = 3. Could your MinPts = 4 simply be an error?
2d
comment Which forecasting method should be selected in case of contradictory results from different accuracy measures?
+1. Note that the expected absolute percentage error (APE) is minimized by the (-1)-median of the future distribution, see p. 752 in the published version of Gneiting's paper. The MASE actually doesn't target anything exotic at all. It's a scaled version of the MAE - scaled by the in-sample error of the random walk forecast, which is a scaling factor that is independent of the forecast and the future distribution. Thus, the expected MASE is minimized by the same quantity as the expected MAE, namely the median.
2d
revised Which forecasting method should be selected in case of contradictory results from different accuracy measures?
edited tags
2d
answered Forecast accuracy metric that involves prediction intervals
2d
comment Forecast accuracy metric that involves prediction intervals
@Aksakal: the Mean Absolute Scaled Error (MASE) is a point forecast accuracy measure. It does not evaluate intervals.
2d
revised Multinomial Count Models
edited tags
2d
answered Multinomial Count Models
2d
comment Forecast function for a MA(2) time series
Hi, and welcome to CV! I'd recommend that you add the self-study tag to your question, take a look at our guidance on self-study questions and edit accordingly.
Feb
5
comment What can we say about population mean from a sample size of 1?
@amoeba: quite correct, but the OP did not specify any restrictions on the distribution.
Feb
4
comment When (if ever) is a frequentist approach substantively better than a Bayesian?
Related: Is there more to probability than Bayesianism?
Feb
4
comment How to find “theoretically best” model?
Whether a predictor $X$ is useful will always depend on the model - after all, your model might already include some $Z$ that is highly correlated with $X$, so that having both $X$ and $Z$ will be worse than having $Z$ alone, even if $X$ may be the actual driver. And of course, if you look long enough, you will always find something that looks useful, and even something that will improve predictions on the holdout sample - but that may be useless in "true" forecasting ("overfitting on the holdout sample").
Feb
4
answered How to find “theoretically best” model?
Feb
4
comment Time series forecasting with many predictors
One problem may be that lags are correlated - after all, this is why we include them in the model (and why, say, ARIMA models with large lags tend to do badly). But Random Forests are specifically good at working with correlated predictors, because of the randomized predictor preselection they employ. So it may well be worth a try.
Feb
4
comment Time series forecasting with many predictors
Yes, information criteria are also possible for lag selection. (Or, you could simply feed many lags to the Random Forest and let it figure out by itself which ones are relevant, see the importance entry that randomForest() returns. This may or may not work well.)
Feb
4
answered Time series forecasting with many predictors
Feb
4
answered Algorithm to determine a point in time series data, after which probability of increase in value is very low
Feb
4
answered Interpretation of level, trend and seasonal indices in Holt-Winters exponential smoothing
Feb
3
revised How to deal with data in which users_ids belong to more than one category (Multilevel) using Random Forest?
edited tags