Add files via upload

237 - prologistica/README.txt

@@ -0,0 +1,43 @@
+This is the source code of the submission "Weighted Ensemble of Statistical Models" to the M4 forecasting competition by ProLogistica Soft.
+
+
+1. REQUIREMENTS
+
+This software was developed and tested on Windows 7 Professional (64bit) in R 3.5.0, with the following packages:
+  M4comp2018 0.1.0 (https://github.com/carlanetto/M4comp2018)
+  trend 1.1.0
+  pracma 2.1.4
+  forecast 8.2 (version is important here, expect different results on forecast 8.3)
+  sets 1.0-18
+  dplyr 0.7.4
+  pbapply 1.3-4
+  robustbase 0.93-0
+  abind 1.4-5
+  parallel 3.5.0
+  stats 3.5.0
+
+We would also like to credit the forecTheta 2.2 package. We use a slightly modified version of forecTheta::otm function, so the package itself is not required.
+
+For trouble-free usage 12 GB RAM is strongly recommended.
+
+
+2. QUICK GUIDE
+
+Set a path to the "tester" directory (see the comment at the top of M4tools.R).
+
+For the forecasting demo on a small subset of M4 data set, see M4/M4demo.R
+For the step-by-step example experiment setup on M4 Yearly data, go to M4/tests/M4test.yearly.R
+The code to fully replicate our solution (for the entirety of M4 data) resides in M4/final/M4final.all.R
+
+For the details of our method, refer to M4-Method-Description-ProLogistica_Soft.pdf
+
+
+NOTE: Due to a minor mistake we made when preparing the final submission, a small group of series got assigned wrong forecasts. This means there will be a tiny discrepancy between our submission and the result calculated by the attached source code (where the problem was fixed). This concerns roughly 250 series from Daily category.
+
+---------------------------------------------
+
+ProLogistica Soft Sp. z. o. o.
+ul. Ostrowskiego 9
+53-238 Wrocław, Poland
+[email protected]
+www.prologistica.pl