In finance, we spend so much time doing analysis and working with data. What’s the most important part of the process? Clearly, sharing your findings. If you are unable to communicate what you’ve done, and your audience can’t understand it, the value of your research is reduced. One way to share your research is through tables of numbers, but these can be difficult to decipher. A key part of making these numbers and tables more accessible is through data visualisation.
In the past few weeks, we’ve seen how important data visualisation is with the unfolding coronavirus crisis, to communicate what’s happening with the public, phrases such as “flattening the curve” have become very common. One example of particularly the effective data visualisations about the coronavirus has been the work of the FT. I’ve tried to mimic some of the charts on a Jupyter notebook, with some help from Ewan Kirk who’s coded up an interactive dashboard for coronavirus data.
I recently attended a online webinar on data visualisation which got me thinking. The presenter, Amina Salima, talked about many unusual types of data visualisation and tried to get us thinking creatively. She gave many examples of work in the area such as the work of Mona Chalabi, who focuses on creating original data visualisations, that are drawn by hand (examples here).
Part of the webinar, was dedicated to create our own data visualisation of how we had spent the day, using a tree like representation, with colour codes, and line styles. Each branch represented an hour of the day. A surprising amount of information could be displayed in a relatively concise way. An alternative way to represent the data could be through a table, however, such an approach would not be as engaging with others or memorable. Whilst, I would hardly call myself particularly artistic it did get me thinking more how I could approach data visualisation better in what I do with financial data.
In finance, we tend to use line charts and bar charts. However, there are many more original and novel data visualisations, we can think use. Instead of using a correlation heatmap, we can use network plots to illustrate the relationships between financial assets. We can use different colours and line widths to make visualisation clearer. We can use animation to display a lot of information, a way of showing changing market environments with many variables. There are so many tools for creating visualisations in Python, ranging from Matplotlib to Plotly, and my own chartpy library which acts as a wrapper for many of these different back ends. There’s a great gallery of interactive dash/Plotly dashboards that help users explore many different datasets. We can also use Jupyter notebooks to make it easy for the audience to dig down deeper into our results and visualisations, opening the black box of PDFs. For those wanting to create very flexible visualisations there’s D3.js although the learning curve is steep. We shouldn’t just limit ourselves to Excel when it comes to charts.
It isn’t easy to create a memorable data visualisation, but it is one of the most important parts of data science, even if often it’s considered an afterthought. Over the coming months, it’s an area I’ll be looking at more.