Five Reasons Why Simple Models Are A Data Scientist’s Best Friend
Forbes
Published Aug. 10, 2021
Written by Sreekanth Mallikarjun, Chief Data Scientist at Reorg, a global provider of credit intelligence, data and analytics, and Adjunct at UVA’s School of Data Science.
Tech-enabled SaaS growth companies are among the most progressive companies when it comes to making use of data science. In a global survey conducted by McKinsey Analytics in 2020, only 16% of respondents said their companies had taken data science models involving deep-learning techniques beyond the piloting stage. At Reorg, I am fortunate to lead a team of data scientists and data engineers working on real-world problems every day.
In my experience dealing with large, ambitious data science projects, simplicity can be powerful. While it may be tempting to reach for sophisticated, cutting-edge methods first when faced with an intricate new problem, complexity has its drawbacks. A single, large, complex model could be overwhelming for data scientists to conceptualize and build and harder for project managers to manage. A black box model could lead to the disengagement of domain experts when it is no longer obvious where their expertise is being applied.
Over the years, I have learned that you do not necessarily need a complex model to solve a complex question. In this short article, I will lay out five reasons (R.A.I.S.E.) why simple models really are a data scientist’s best friend.
