Results
We are proud to announce that the winner of the Computational Intelligence in Forecasting Competition 2016 is Slawek Smyl (Microsoft, USA) who competed with a method called LSTM Neural Network Applied To Deseasonalized Data. Congratulations!
The organizers would also like to express gratitude to all competitors for their work, honor and courage to face the independent comparison and for their thirst for truth and knowledge.
A table of detailed results:
| # | Method | SMAPE | SD# |
|---|---|---|---|
| 1 | LSTM-Deseasonalized | 0.105 +- 0.107 | 1 |
| 2 | LSTMs and ETS | 0.108 +- 0.116 | 2 |
| 3 | *ETS | 0.119 +- 0.142 | 9 |
| 4 | MLP | 0.121 +- 0.135 | 7 |
| 5 | REST | 0.124 +- 0.133 | 6 |
| 6 | *FRBE | 0.129 +- 0.162 | 13 |
| 7 | HEM | 0.130 +- 0.147 | 11 |
| 8 | *Avg | 0.131 +- 0.133 | 5 |
| 9 | *BaggedETS | 0.131 +- 0.176 | 17 |
| 10 | LSTM | 0.133 +- 0.155 | 12 |
| 11 | Fuzzy c-regression | 0.137 +- 0.127 | 4 |
| 12 | PB-GRNN | 0.145 +- 0.166 | 14 |
| 13 | PB-RF | 0.145 +- 0.166 | 15 |
| 14 | *ARIMA | 0.146 +- 0.218 | 21 |
| 15 | *RW | 0.146 +- 0.137 | 8 |
| 16 | *Theta | 0.148 +- 0.122 | 3 |
| 17 | PB-MLP | 0.149 +- 0.172 | 16 |
| 18 | TSFIS | 0.151 +- 0.147 | 10 |
| 19 | *Boot.EXPOS | 0.153 +- 0.206 | 20 |
| 20 | MTSFA | 0.165 +- 0.180 | 18 |
| 21 | FCDNN | 0.166 +- 0.194 | 19 |
| 22 | MSAKAF | 0.204 +- 0.225 | 22 |
| 23 | HFM | 0.224 +- 0.251 | 23 |
| 24 | CORN | 0.288 +- 0.263 | 24 |
*) Statistical methods evaluated as a benchmark.
Detailed results as provided by contestants may be obtained in the download section.
All the competitors are winners and deserve maximal respect!