Data Scientist

When I tell people that I interned as a Data Scientist over the summer, the first question I get asked is, "Can you tell me who's going to win the 52nd Super Bowl?"

Data scientists appear to be wizards who whip out their crystal ball (MacBook Pro), chant some mumbo-jumbo (machine learning, big data, and automation) and make accurate predictions of what the future will hold. But, I would like you to know that data science is not magic and that I'm no magician.

A data scientist draws information, or business insight, from raw data. One should develop a deep understanding of statistics and computer science as well as gain industry knowledge to conduct such work.

There are three kinds of responsibilities that sum up the bread and butter of data scientists:

• Data Structuring
• Data Analytics 
• Data Visualization

Allow me to share the ingredients that make up the "magic":


Data Structuring

Structuring the data generally takes the most time, it involves:

• Tracking appropriate metrics for the business
• Storing data on suitable platforms
• Correcting data anomalies 

Here's an example: I want to predict sales and supply levels of a dying lemonade stand to make it profitable. What kind of information do you think I should track? Buying habits of existing customers, such as what flavors of lemonade are most popular, can be an important factor I would track. As a data scientist, I would first consider all possible metrics before picking the one that makes the magic happen. Next, I would choose an efficient platform to store this data (for example, google sheets). Finally, to make sure the data is consistent, I would verify that there aren't any missing values or incorrect entries.


Data Analytics

Analytics is the glory work we love to talk about, it entails:

• Developing or applying algorithms (fancy word for a set of rules) 
• Extracting underlying patterns from the model
• Using theoretical knowledge to predict the future state 

The solution to any data science problem generally begins with asking an interesting question. For example, how many glasses will the lemonade stand sell next month? Next, I would take the collected data (for example, transactions of the past six months) and apply various statistical models. To choose the model with the highest accuracy, I would predict the current month's sales using the past five months transactions. Finally, I can now apply the chosen model to predict the sales for the following month.


Data Visualization

We use visualizations to share our findings with the outside world, it includes:

• Creating charts and infographics that summarize data
• Programming interactive elements into visualizations that enable customization
• Automating dashboards to help decision making

Explaining the structure, analysis and the rest of the mumbo-jumbo in plain English and through visuals is the last and most crucial part of my job. Like the customer that avoids reading what goes into the lemonade, my audience doesn't want to know if I applied the Bayes Classifier or the Random Forrest algorithm. The visualization should always pass the 60-second test: The reader should be able to understand and act on the trends in the data.


Readability
Flesch Reading Ease: 57
Flesh-Kincaid Grade Level: 10.2
Passive Sentences: 0%