GRADUATION RESEARCH CHANTAL BLOM - PART 2

It's been a while since I last talked about the Data Science project I'm working on with BDC and the Dutch Handball Association (NHV).

We're already over halfway through the project, and the direction we're heading in terms of data analysis has become much clearer. In this post, I'll explain the idea and where we stand now.

At the NHV, data is collected from handball talents from various teams.
This is data from measurements such as high jump, throwing from different positions, a sprint test, etc.
The talent measurement results are now structured.

The idea now is to give scouts and coaches an opportunity to compare talents.
For example, it can be useful for trainers to see where similar talents differ
and where a talent could receive additional training.
We compare talents by determining the distance between talents in the feature space.
We then look at the k closest talents.

There are now two parameters that can make a comparison better or worse: the number of talents we look at and the measurements we include in the comparison.
To determine the best combination of these parameters, we use k-Nearest Neighbor regression.
We use some of the data to try out different combinations.

The video below explains how to determine the best value of k
using k-Nearest Neighbor regression.

The idea is that we use the algorithm to estimate the value for each measurement of each talent and
determine the difference between the estimated values and the actual values.
We then do this for different combinations of k and properties.

The model, the best value for k and the best combination of properties, can be used in different ways in an application to compare talents.

I will share more about this in a next article.

 

Chantal Blom

Business Data Challengers