Chintan Shah – CSCC46 2020 Course Blog

How Network of Attitudes Affect Voting Behaviours

With the US election being the main news story for the past few months, I thought it would be a good time to see how network analysis could apply to the US presidential elections. Obviously there are many different ways network analysis can be applied to elections but I wanted to focus on how network of attitudes can affect voting outcomes. This was particularly more interesting to me because after seeing the results of the election, it is certain that the American people have very polarizing views on politics and it would be interesting to see how much that is influenced by networks of people, specifically how network of people’s attitudes towards the presidential candidates affect voting.

It has been known that attitudes have a big impact on a person’s personality and their social behaviours so it would be interesting to see how network of attitude attributes can have a strong impact on the decision of which presidential candidate to vote for. That is precisely what this research team tried to do as they used data from previous US presidential elections to confirm their hypotheses of how attitude networks can predict the elections and how it entirely depends on level of connectivity and how the central element of attitudes has the strongest impact which directly relates to material learnt in CSCC46 so far.

Highly connected attitude networks having stronger impact on voting decisions

This research team was able to see how highly connected attitude networks have a much stronger impact on voting decisions and this is shown in the image above. You start to see a relation between connectivity and impact on voting and how the nodes in the two network graphs represent different attitudes and the edges are the correlations between the two attitudes with thicker edges representing higher correlations. The nodes that are closely put together also represent highly connected attitude networks. With this data, the research team was also able to take it further and see which attribute node is central thus having the biggest impact on voting behaviours. This would immensely help presidential candidates see what they have to showcase the most because the majority of people are looking for just that.

With all that said, it would be interesting to see what attitude types affected the 2020 election the most and apply the same network analysis above on the voters of today. I also hope to see more social aspects to this where you can start to see how your friends affect your own political ideologies and how that can subsequently affect voting and your choice of presidential candidate. I hope more research into elections continues as it is always interesting to see analytics in various aspects of socially relevant behaviours.

Source: https://nature.com/articles/s41598-017-05048-y

Network Analysis on Hockey Players

With the National Hockey League season ending a couple weeks ago, it would be a good time to estimate how teammates influence scoring by using network analysis. I grew up watching hockey and I have always wondered how we could use standard, readily available metrics in such a way to accurately see a players ability to perform. The curiosity has only grown bigger as I have taken more practical courses for network analytics like CSCC46.

It is known that box scores in any sport can never tell the whole story of the game and so deeper analytics should always be done to truly know how the team and players performed. There has also always been a place for advanced analytics in hockey where they use standard metrics from the boxscores to calculate advanced analytics such as Corsi. However, it gets really interesting when you can use standard metrics from these boxscores and start to apply network analysis to it, directly relating to CSCC46 course material. This is precisely what Oppenheimer attempted to do, using standard metrics to calculate a single analytic that can give you a good understanding of the players performance on the team.

2014-15 New York Islanders Shot and Passing Network

Oppenheimer was able to calculate betweenness and make it meaningful for hockey by visualizing scoring as a directed network where the nodes represent players and the edges represent primary assists to and from each players’ goals. The above diagram depicts this but the edges reflect shot and passing instead of primary assists. This calculation will show us that if the player has a high betweenness, it means their production is really high as they score and assist frequently and their assists are sometimes distinct too but you also start to see if some players rely on other players for their scoring. Sometimes however, this analytic gets meaningless if the team is stacked in the sense that they have a lot of players who can produce. This is because, betweenness is a relative metric hence it is relative to the players’ team and so being on an offensive team, the player can be deemed less influential than they really are.

With all that said, I believe this is the start of applying network analysis to the game of hockey and there is a lot to explore in this field and further analyze a players production to see if they would be a good fit for another team. The really cool thing about betweenness is that it can also be applied to other sports in virtually the same way and you can play around with different types of edges and nodes. I hope to see more research in this as I would love to see the world of analytics in hockey and possibly alter how teams are built forever.

Source: https://towardsdatascience.com/whose-point-is-it-anyway-using-network-analysis-to-estimate-teammate-influence-in-hockey-scoring-e9fc97f26648