6 Appendix E ACCURACY OF FORECAST INFORMATION FOR TRAIN AND FREIGHT WAGON MOVEMENTS, Regulation (EU) 2026/253

In forcesince 6 Feb 2026
(1)

Since the train forecast information referred to in point 2.6.4 and the wagon forecast information referred to in point 3.2.3 refer to the predicted time of arrival of a train and of a freight wagon and its load, respectively, at an agreed reporting point, it will be used by stakeholders and customers involved in the rail transport service to plan subsequent operations or onward connections.

(2)

Since both the train forecast information referred to in point 2.6.4 and the freight wagon forecast information referred to in point 3.2.3 are an estimate that derive from the time originally planned in the working timetable +/- the deviation from the working timetable that occurred during the train operation, the originator of such a forecast shall:

(a)

carry out an ex-post measurement of the accuracy of its forecast in accordance with Article 10 in order to create confidence in the data to be reused by the recipient for the planning of subsequent operation;

(b)

ensure that the recipient of the forecast referred to in point (a) is informed of the expected accuracy of the forecast based on earlier train runs.

(3)

As an alternative to linear time-shifting methods commonly applied by some telematics stakeholders for the purposes of point (2) (a), others may apply different methods or use internal algorithm-based machine learning methods that may be integrated into an artificial intelligence to achieve more accurate and intelligent insights.

(4)

For the purpose of point (2), recurrent quality assurance checks on the accuracy of forecast information are laid down as follows:

(a)

absolute error:

The ‘absolute error’ is calculated as the absolute value of the difference between the actual date and time of arrival at a specified reporting point and the forecast date and time of arrival at that reporting point;

(b)

relative error:

To put the ‘absolute error’ at a specified reporting point in perspective with the duration of the remaining train journey, a ‘relative error’ is calculated as the ratio or percentage obtained by dividing the ‘absolute error’ by the ‘duration of the remaining train journey’;

The ‘duration of the remaining train journey’ is calculated as the difference between the actual date and time of the arrival of a train at the reporting point and the time when the corresponding forecast information has been provided;

(c)

accuracy:

The ‘accuracy’ are calculated as the complement of the ‘relative error’ to 100 %.

(5)

Where a data holder, that is to say the originator of the data, performs recurrent quality assurance checks as referred to in point (3), it shall assess the accuracy of the data as a stochastic probability of the statistical values of the forecast information at a specified reporting point for a group composed of previous rail transport services.

(6)

A data holder shall perform quality assurance checks on the accuracy of train forecast information only once that train has arrived, or after that freight wagon and its load have been handed over or delivered at, the reporting point that is subject to the forecast. Quality assurance checks shall be based only on historical data.

(7)

Indicators resulting from the measurements referred to in point (3) shall be available through the common Union web UI referred to in Article 5.