In time series table schema each observation is stored in a table as a separate record (line):
Field name (column name) | Description | Examples |
---|---|---|
series_id* | Time series identifier - a unique name that identifies a time series | “Y1” |
timestamp* | Any representation of the period to which the observation relates. We recommend the use of the ISO 8601 standard | "1997" in case of yearly data, "1997-01-20" in case of daily data, "1997-11" in case of monthly data, "1997-W03" in case of weekly data, "2018-Q2" in case of quarterly data |
value | The value observed | 1000 |
*the key (the unique value that should not duplicated) for this table schema is series_id, timestamp. In other words, we cannot have two (or more) records in a table relating to the same time series and the same period of observation (timestamp)
series_id value timestamp
1 Y1 3103.96 1984
2 Y1 3360.27 1985
3 Y1 3807.63 1986
4 Y1 4387.88 1987
5 Y1 4936.99 1988
6 Y1 5379.75 1989
7 Y1 6158.68 1990
8 Y1 6876.58 1991
9 Y2 5389.80 1984
10 Y2 5384.40 1985
Forecast Table Schema is needed to store forecasting results.
Each table line corresponds to all the forecasting results obtained for a given time series series using a given method for a given horizon for a given origin:
Field name (column name) | Description | Examples |
---|---|---|
series_id* | Time series ID for which the forecast was calculated | "Y1" |
timestamp* | Any representation of the period to which the observation relates. We recommend the use of the ISO 8601 standard | "1997" in case of yearly data, "1997-01-20" in case of daily data, "1997-11" in case of monthly data, "1997-W03" in case of weekly data, "2018-Q2" in case of quarterly data |
origin_timestamp* | Origin of the forecast (provided in a timestamp format) | "1996-12-29" |
horizon* | Forecast horizon | 3 |
method* | Method identifier - a unique name that identifies a method by which the forecasting result was calculated | "ARIMA" |
forecast | Point forecast | 234 |
lo95 | The lower limit for the 95% prediction interval | 178 |
hi95 | The upper limit for the 95% prediction interval | 273 |
*the key (the unique value that should not duplicated) for this table schema is series_id, method, timestamp, origin_timestamp, horizon.
series_id method timestamp origin_timestamp forecast horizon lo90 hi90
1 Y1 A 1989 1988 5406.43 1 5183.349 5629.511
2 Y1 A 1990 1988 5875.96 2 5652.879 6099.041
3 Y1 A 1991 1988 6345.48 3 6122.399 6568.561
4 Y1 B 1989 1988 5473.87 1 5250.789 5696.951
5 Y1 B 1990 1988 6010.43 2 5787.349 6233.511
6 Y1 B 1991 1988 6546.63 3 6323.549 6769.711
7 Y1 C 1989 1988 5406.43 1 5183.349 5629.511
8 Y1 C 1990 1988 5875.96 2 5652.879 6099.041
9 Y1 C 1991 1988 6345.48 3 6122.399 6568.561
10 Y2 A 1989 1988 4142.60 1 3919.519 4365.681