Home Math for a better future

Chi siamo La nostra identità

Soluzioni La matematica per evolvere

Casi studio Esempi di storie di successo

Education La formazione su misura

Contattaci Il filo diretto con gli specialisti

Home

Chi siamo

Soluzioni

Casi studio

Education

Contattaci

Magazine Unisciti a noi

case study

Media & Publishing

z = −1 + i nα = θ + 2πk f(x) = 3x − 2 z = −1 + i −x + 2y + z + t = −1 2x + 2y + z + 4t = 0 C = {z = x + iy x, y ∈ IR} [x0 −r, x0 + r] (cosnα + isinnα) = (cosα + isinα)n

Media & Publishing

Support to the planning of a television schedule

An important Italian radio and television broadcaster has entrusted us with the realization of a decision support tool that helps out the channel managers in planning the schedules.

z = −1 + i nα = θ + 2πk f(x) = 3x − 2 z = −1 + i −x + 2y + z + t = −1 2x + 2y + z + 4t = 0 C = {z = x + iy x, y ∈ IR} [x0 −r, x0 + r] (cosnα + isinnα) = (cosα + isinα)n

During the creation of the schedule, the channel manager has to manage multiple factors in order to decide the best placement of the content in it, in order to maximize audience ratings.
They must take into account all the scheduling constraints, the broadcasting rights available for each content, the target audience of the various time slots etc…
The tool we have developed supports the channel manager, producing an ordered list of titles that considers all the constraints. This content ranking is then used as input in an optimization model that, based on the expected performance and the different airing criteria, provides the optimal schedule for a time period of interest as a result.

From hours to minutes

The time spent on schedule creation and planning has been drastically reduced

Efficiency on rights

Thanks to the optimization algorithm, it is possible to make the best use of the rights on the available content

High profiling

Content not yet included in the schedule is automatically classified by theme and affinity with the broadcaster's networks

Find more on this case study

Deepening case study

Machine learning for share prediction

In order to carry out this project, we developed two machine learning models that predict content performance: a classifier and a regressor. The ranking was obtained by ordering content from best to worst based on the performance predicted by one of the developed models. From the user's perspective, the ranking can be modified to account for any additional constraints, such as the close expiration of the content's rights. All models work on an up-to-date database that develops an automated pipeline able to download the latest version of data from the broadcaster's infrastructure every night.

Read another case study from the same industry

focus