"My placement was with Sky Betting and Gaming in Leeds where I was tasked with modelling changes in customer behaviour. On my first morning, I was welcomed by my line manager, introduced to the team and given a desk and Macbook. I was on a team of six people, and they were extremely helpful at getting me settled-in and introduced to key members of staff from other teams. It’s always good to get on well with the tech support people!"
"The exact requirements for the project were not defined at this stage, so I spent some time talking to managers from the departments who would ultimately use the model. From them, I was able to determine what they would like and ask them about their understanding of the behaviour of their customers. I produced an initial specification and timetable and agreed them with my manager. As is so often the case, I found I didn’t keep to the timetable, but by producing it I was at least able to recognise when, and by how far, I was behind. I found it useful on a weekly basis to review how I was progressing, and think about what I could do to get back on-track. I had weekly meetings with my manager, and kept in contact with my academic supervisor, so there was no likelihood of getting too far behind.
"The team turned out to be quite a sociable group so there was no shortage of opportunities to explore the pubs of Leeds. My manager treated us to a night at the races, where I displayed my lack of judgement of horses. We had a corporate day where the company provided buses to take us to an airfield where they had set-up a fairground with free food and drink, and a very interesting guest speaker from Wired magazine.
"The environment in which I was worked was a combination of Hadoop and RStudio. Online data is stored in an Informix database before being uploaded to Hadoop overnight. I extracted the data and did most of the preprocessing using SQL in Impala, one of the Cloudera tools for Hadoop. Once the data was ready for building models, it was exported to a CSV file for importing into R.
"In the first few weeks of the project, I spent some time doing a literature search of studies relating to gaming behaviour. I carried out an exploratory data analysis(EDA) to confirm the findings from the literature and produced a presentation which I delivered to the team. The EDA built confidence in the features I was proposing to use, and I went into the modelling phase quite confident that the models would work well. This was where I hit a bit of a setback.
"My first models performed very poorly. I tried a number of changes to the data and saw little improvement. I checked the code producing my training data, and I couldn’t see any problems with it. As time passed and I was making little progress I became concerned about not being able to demonstrate methods and report results, and the impact this would have on my final report. This was where one of the benefits of having an academic supervisor was shown. I spoke to my supervisor, and he helped me come up with a strategy to move forward and gave me some much-needed encouragement. I also spoke to my colleagues about the issues I was having and they had some very helpful suggestions. I would strongly recommend anyone in a similar position to take advantage of the experience of their colleagues and supervisor to avoid losing sleep.
"Eventually my results improved and I had enough material for a decent report. Knowing that I was going to have to write a report in pdf, I had used one of my previous assignments to teach myself LaTeX which was time well-spent. I was even lucky enough to find a blog on www.sharelatex.com on the subject of laying out a dissertation style report using a master document. Despite being occasionally infuriating, I love LaTeX. Laying out plots and figures is really easy, and the final report looks great. You just need to remember to compile and save it frequently. I was pretty happy with the final report and the viva went well as far as I can tell.
"To sum up, my placement, while being a lot of fun, also had its challenging moments, which is how it should be considering it’s a Masters programme. I learned a lot from my colleagues, and I learned a lot about how Data Science is done in the commercial world. Sky Bet were great to work with, and Leeds is a wonderful city."