This morning kicked off with a blog post from Andy telling us to create a fake dataset in Python using Chat GPT and to then use that data in Power BI to create a KPI Dashboard. None of the above pieces of software are my strong suit so safe to say I was in for a big learning curve today.
I first began to prep Chat GPT by feeding it small parts of the brief and asking it to create ideas for fake companies. I went back and forth with it for a little while until it suggested this: Space Tourism Agency: CelestialVoyages
I then prepped chat GPT with some headers I wanted and asked it to suggest me a few more. I then asked for a Python Script.
Next came the first big obstacle for the day; installing Python. I'll be honest, the process of installing Python lasted around an hour because there seem to be so many different versions / software that run it and most weren't co-operating with the code given by Chat GPT. Eventually Lydia (someone in my cohort) put me out of my misery by pointing me at a website called JupyterLite.
I encountered a couple of error messages which I fed back to Chat GPT which recommended valid solutions. Next I wanted to input some trends in my fake data.
I recommended a couple of trends I wanted and also asked Chat GPT for trends it thought I should implement. After this I told Chat GPT to implement them. This became my second obstacle for the day. Interestingly, there were some trends implemented that worked on the first or second attempt incredibly well but other Chat GPT solutions were flawed to say the least. Time was ticking, and no matter how much I tried to prep Chat GPT, I couldn't seem to get the results I wanted.
I myself am very rusty when it comes to using Python (which I haven't used in over 5 years) but I think for someone who is competent with the software, they may have had an easier time with this.
One particularly difficult issue I had was that each of my rows represented a space trip but included demographic data for a customer. The results of the output were below.
I reached out to the wider company who tried to help, but it seemed the solution was incredibly long and I simply didn't have the time.
So I moved on in spite of the glaring data quality issues and moved on to sketches. Since the brief was for a KPI dashboard, I decided to keep my sketch very simple and go for only key insights. I also went to chat GPT for a Space colour scheme but the results weren't ideal.
I went on to build in Power BI, another software that I'm not too familiar with. I had some quite ambitious plans that I just couldn't figure out how to achieve, despite asking around. In the end, I have quite a simple dashboard but I'm happy with the information and design.
In future, I would love to redo this brief over the span of a week where I'd have time to brush up on my Python and Power BI skillset.