Microsoft Azure Machine Learning Studio

Editor's Notes

• #2: Sign into your Azure Machine Learning Studio account at ‘studio.azureml.net’. And download the two titanic files in the email. While you do this, I am going to briefly discuss some ML basics. I think I can get through some of these while you download and log in. You will have the slides if you need them.
• #3: Thank you for letting me show you what I am doing with ML at my internship and I hope it helps you see the possibilities for ML in your career. What is machine learning?
• #4: The basic premise of machine learning is to build algorithms that can receive input data and use statistical analysis to predict an output while updating outputs as new data becomes available.
• #5: A narrow definition in the economics world is that ML is a field in which algorithms are developed to be applied to large data sets. Primarily in one of three key areas.
• #6: To remind you of what Mike shared last week, these three categories are under the umbrella of either supervised or unsupervised learning. The majority of practical machine learning uses supervised learning. Supervised learning is where you have input variables (x) and an output variable (Y) and you use an algorithm to learn the mapping function from the input to the output. Supervised learning problems can be further grouped into regression and classification problems. Unsupervised learning is used to find associations which are there, but have not been discovered yet. A good example of unsupervised learning is the music recommendations on streaming services. Rachel likes country music and sometimes blues shows up on her Pandora. Why? Because the song has more in common with some aspects of country music than blues, or vice versa.
• #7: In economics, specifically econometrics, ML is used to assist in the prediction of an outcome. What kind of a model would we use to predict an outcome? Like, say a sock market crash… using a binary outcome variable? Logit/Probit is correct! The difference in ML and econometric models is that the variable are not explanatory. Also, the ML algorithm tests models in order to find the best model. In traditional economics, the researcher chooses the model to use and tests it against a few other models. Because of this model specification and explanatory reasons, econometric models are often simpler at the cost of more assumptions. ML allows the models to be more complex models with fewer assumptions. We are going to jump into azure now, which you all should be
• #8: ML is disrupting the econometrics in model discovery as shown earlier, but also in many other aspects. We all remember Dr. Sheron's problem set where we had to select among the variables, right? What did we have like 5 variables on housing? What if we had 500 variables and we were tasked with a report to the housing department? ML would allow us to find the best variable for the model. Variable discovery and model selection are reasons why some leading economist using ML suggest that in the future all professors will have a virtual TA finding the best model. In fact a Stanford economist suggest this will lead to a regularization and systematic model selection process which will become a standard part of empirical practice in economics. There are also studies now where they are using models to look at two different policy choices, essentially allowing economics and policy maker to experiment before they implement. In academics all of these will probably lag, but in business it will happen fast because it will provide a competitive advantage to the firm that can get this working. Translation to you as potential students is that your business will be working in machine learning and computer scientist do not know how to apply it to economics or statistics. You do! This field is so young.
• #9: Building on the disruption, ML is an opportunity for you to distinguish yourself from other possible candidates in the job world. In fact, at the recent Seminole futures I met with a company who reviewed my resume and said they were going to have to talk with their boss about creating a position for me because one didn’t exist, but they knew ML was something they are using in a different capacity and they are looking to expand on it. The field is young, we have the skillset in terms of knowing the models and the economics side. ML allow us to supplement our econ skills with technical skills to provide value to a company in many different areas. In short it is an opportunity for you to clarify this new, emerging blend of ML and economics in your career. It is a opportunity for you to provide something no other candidate can to your potential employer, someone who sees through the clutter and can focus on the future of economics as we know it.