M3-Competition Benchmark Task

An M3-Competition Benchmark Task is a univariate forecasting task that is a Benchmark Task.

• See: Univariate Time-Series Dataset, M-Competition, Benchmark Task, Forecasting Task.

---

---

References

• http://www.forecasters.org/data/m3comp/m3ijf-c.dta

2011

• http://www.forecasters.org/data/m3comp/m3comp.htm
  • The data file contains the following information for each series:

   The first line which includes the following:
       The ID number of the series (from N0001 to N3003)
       The number of data points (observations) of each series (let us call this number n).
       The number of required forecasts
           6 for yearly data
           8 for quarterly data
           18 for monthly data
           8 for "others" 
       The type of series (e.g., YEARLY/MICRO or MONTHLY/INDUSTRY).
       The ID number of each type of series (e.g., there are 146 YEARLY/MICRO series so this ID number goes from 1 to 146, while there are 334 MONTHLY/INDUSTRY series making their ID go from 1 to 334).
   n observations of time series data (8 per line)
   The starting date of the data (Year and Period, if no information is available or the information is not relevant the value is 0.0).
   A short description (title) for the time series.

• http://forecasters.org/resources/time-series-data/m3-competition/
  • The 3003 time series of the M3-Competition are distributed as follows:

Types of Time Series Data
Interval 	Micro 	Industry 	Macro 	Finance 	Demog 	Other 	Total
Yearly 	146 	102 	83 	58 	245 	11 	645
Quarterly 	204 	83 	336 	76 	57 	0 	756
Monthly 	474 	334 	312 	145 	111 	52 	1428
Other 	4 	0 	0 	29 	0 	141 	174
Total 	828 	519 	731 	308 	413 	204 	3003

2000

• (Makridakis & Hibon, 2000) ⇒ Spyros Makridakis, and Michèle Hibon. (2000). “The M3-Competition: results, conclusions and implications.” In: International Journal of Forecasting, 16(4). doi:10.1016/S0169-2070(00)00057-1.
  • ABSTRACT: This paper describes the M3-Competition, the latest of the M-Competitions. It explains the reasons for conducting the competition and summarizes its results and conclusions. In addition, the paper compares such results/conclusions with those of the previous two M-Competitions as well as with those of other major empirical studies. Finally, the implications of these results and conclusions are considered, their consequences for both the theory and practice of forecasting are explored and directions for future research are contemplated.
  • AUTHOR KEYWORDS: Comparative methods — time series: univariate; Forecasting competitions; M-Competition; Forecasting methods, Forecasting accuracy.
  • QUOTE: The 3003 series of the M3-Competition were selected on a quota basis to include various types of time series data (micro, industry, macro, etc.) and different time intervals between successive observations (yearly, quarterly, etc.). In order to ensure that enough data were available to develop an adequate forecasting model it was decided to have a minimum number of observations for each type of data. This minimum was set as 14 observations for yearly series (the median length for the 645 yearly series is 19 observations), 16 for quarterly (the median length for the 756 quarterly series is 44 observations), 48 for monthly (the median length for the 1428 monthly series is 115 observations) and 60 for ‘other’ series (the median length for the 174 ‘other’ series is 63 observations). Table 1 shows the classification of the 3003 series according to the two major groupings described above. All the time series data are strictly positive; a test has been done on all the forecasted values: in the case of a negative value, it was substituted by zero. This avoids any problem in the various MAPE measures.

---