Yap Jo-Yee

I was looking for a hands-on class with lots of instructor-student interaction, because I wanted to be able to practice in class, getting feedback as I went along. For me, feedback is a crucial part of learning as I improve. As a beginner, the elements of discussion with someone who knows the best practices is immensely valuable compared to the things that blog articles assume and don’t mention.
Program Taken:
Applied Data Science Bootcamp Graduate (Jul'21)
Example of my work:
Click here to view my work
View program details

I'm from an economics background and interested in everything to do with development and human behaviour. The main question I try to tackle in my work is, “How can we create human-centred solutions that improve the welfare of marginalised communities?”. 

Like many questions, data is integral to this one. We have to understand what indicators to measure, how to measure and collect the data, how to process it, how to model it, and how to turn the insights we get into action. I wanted to find better ways to do all of that, which led me to look out for data science courses.

I was looking for a hands-on class with lots of instructor-student interaction, because I wanted to be able to practice in class, getting feedback as I went along. For me, feedback is a crucial part of learning as I improve. As a beginner, the elements of discussion with someone who knows the best practices is immensely valuable compared to the things that blog articles assume and don’t mention.

A good teacher not only offers knowledge, but the ability to transfer that knowledge in application in an information-saturated world, and I'm grateful to have found that in Forward School.

One of my favourite moments during the course was when I managed to deploy a machine learning web app for my capstone project. I failed to deploy a web app on a previous assignment, so this was especially rewarding. I think I did a little dance-wiggle in my chair to celebrate. Furthermore, my team partner and I built a dashboard for crime in Malaysia within 2 weeks. She scraped data off news sites and visualised them with Python, and I put it all together using Django. There were many frustrations along the way, and the learning curve was steep. But it made success all the more sweeter.

My takeaway from all of this is, it’s always going to seem impossible at the start. Do it anyway.

My Capstone Project presentation during Demo Day! (Scan the QR to see more)
I was looking for a hands-on class with lots of instructor-student interaction, because I wanted to be able to practice in class, getting feedback as I went along. For me, feedback is a crucial part of learning as I improve. As a beginner, the elements of discussion with someone who knows the best practices is immensely valuable compared to the things that blog articles assume and don’t mention.

My capstone project was about financial inclusion - essentially, access to basic, formal financial products like a bank account, insurance etc. 

It is often the most deprived who cannot access these financial products. As a result, life is a lot harder for them. They might not be able to save and borrow effectively, and this makes wealth and income generation much tougher. The lack of insurance also means that when accidents happen, there is no financial cushion. These examples are only the tip of the iceberg.

The main idea behind the capstone project was to predict the future, not present, financial inclusion of a household, using data of the household’s current assets, and the physical condition of their home. The issue with many government policies is that by the time national surveys are completed, processed, and used to create policies, these policies are a few years too late, because they try to address situations in the past. At best, they’re less relevant. At worst, they aggravate the problem.

However, if we can use predictive modelling to address financial inclusion, policymakers can nip a growing problem in the bud, target better, and estimate the amount of resources that will be needed for the future. To my mind, that makes it a worthwhile project.

The project went through many iterations, simply because of data limitations. I definitely wouldn’t want anyone to think that this was the first idea, or that it was good-to-go right from the start. My instructor and I had lots of discussion over the project, and his input helped shape it tremendously. Even after having found a suitable dataset, it took some wrestling to nail the y-variable (what I want to predict), the x-variables (the predictors), and the model form.

The course gave me a view of the whole data pipeline, and helped me think about it in a more systematic way. More importantly, I’m more aware of how data can be used to solve problems, and which problems too. 

Machine learning is no longer such a black box to me. Consequently, where I used to see dead-ends in my projects at work, I now see possibilities. It’s also helped me consider a broader range of questions, now that there are more tools to solve them with.