“Unlike a machine learning project, the output of a data science project is often a set of actionable insights, a set of insights that may cause you to do things differently. So, data science projects have a different workflow than machine learning projects”
– Andrew Ng’s “AI for Everyone” Coursera
In a continuation of my last post about the difference between buzzwords and keywords in the Data Science fields, I wanted to elaborate on the differences between a machine learning project and a data science project. I also wanted to talk about the implementation of such projects in businesses.
A data science project has a different set of outputs to aid a business in making decisions (therefore actionable insights) and extracting meaning. In contrast, a machine learning project focuses on building a model that can learn by itself using data.
Therefore, as Andrew Ng says, there is a different workflow and different key steps that need to be taken in each project.
A Machine Learning Project
- Collect relevant data
- Train model
- Deploy model
A Data Science Project
- Collect data
- Analyze data → Iterate many times to get a good insight
- Suggest hypotheses and actions
- Deploy actions
As you can see, just as machine learning and data science can overlap in ideas and processes, so do the projects related to each field. The steps and thinking involved are very similar but they look at issues from two different angles. A data science project tends to look at issues through a very mathematical and statistical lens. You’re finding patterns and seeing how that can apply in the future. A machine learning project tends to look at issues through an “intelligence” lens. How can a machine pick up knowledge and keep learning to optimize?
Applying This
Applying this to a business, we have to think of three main aspects: the problem, the approach, and the process.
What problem are you trying to solve? Is this an overly ambitious project? Many AI projects experience setbacks or failures along the way.
What approach are you looking for? Think about using an incremental implementation approach rather than transforming an entire business model. This comes back to the previous point about setbacks and failures. It would be easier and more efficient to focus on augmenting human capabilities rather than replacing humans. Think superhuman compared to robots. We want a superhuman.
The last aspect to consider is the process. We must understand which technologies
perform what types of tasks directly address the business needs, and develop plans to scale this company-wide or department-wide.
SANJITA SRINATH 
Leave a comment