Chantal Blom graduation research – part 3 research report

In this last part of a 3-part series, Chantal Blom talks about her Data Science graduation project at Business Data Challengers.

About 3 months ago I told you about the data science graduation project I worked on together with Business Data Challengers and the Dutch Handball Association (NHV).

In this report I will briefly describe the three phases of the project.

1. Data preparation

In this phase we have chosen six physical tests from the data on the basis of which we want to compare talents. It quickly became clear that a certain test (the T-test) had only been done by some of the talent.

We have filled in these missing values. We first clustered the data with the other physical tests. We then used linear regression to estimate the T-test value of the talents using the other physical tests and the cluster variable.

The output of this phase was a data table with 141 talents and 6 physical test variables.

2. Data analysis

The second phase had the data table from the first phase as input. The aim of this second phase was to find the best model with which to compare talents. We determined the best model in the way I described in the previous post with k-Nearest Neighbors regression.

The only difference is that instead of looking at a certain number (k) of comparable talents, we can also look at the talents within a distance of a certain talent. In this way we have determined the best distance for comparison for the NHV.

3. Data visualization

Spider diagram Chantal Blom Data visualization talents

For example, with the best distance we can compare a talent with other talents in the same group. One way to visualize different talents is the spider chart. An example can be seen in the image. The different axes represent different physical tests. The pink talent is the talent we wanted to compare with the rest of the group.

Three talents appeared to fall within the best distance. A coach can now see how the talent differs from other talents in the group. The image shows that the talent differs mainly on the vertical jump and on the T-test. The coach may decide to pay more attention to these aspects during the training.

Conclusion

I learned a lot during this project and that is why I would like to thank BDC for this great opportunity. In the meantime, I have presented the project at the university in Utrecht and it is now time for a holiday!

In September we will visit the NHV in Papendal to present the findings.

Chantal Blom