Data Science Framework

Throughout this chapter we focus on Sierra Investments, a small wealth management firm. Sierra is evaluating whether it should invest in and buy certain loans. To make this decision, Sierra compares the interest revenue from the loans to the costs of acquiring loans, including costs of late payments, defaults, and collections activities. Paige Baumann, the management accountant at Sierra Investments, is interested in exploring how she might use data science to help managers make more effective investment decisions by identifying the payoffs and risks associated with different investment choices.^*

Concepts in Action

Carnival Uses Big Data and Machine Learning to Sail Toward Greater Profitability^*

rawpixel/123rf.com

Carnival, the largest cruise operator in the world, leverages big data, machine learning, and artificial intelligence (AI) to cut costs and drive revenue to create “smart cities at sea.” Its goal: to operate more efficiently while increasing sustainability.

Carnival uses data from thousands of sensors on board its ships and machine learning to predict water consumption on cruise ships. This allows Carnival to optimize how much water it should carry for a specific route with a specific set of guests. By carrying less water, each ship saves as much as $200,000 each year in fuel costs while also reducing carbon emissions.

Carnival also uses data science to reduce waste. Cruise ships generate several types of waste while ferrying passengers: food waste, meal servicing waste, cardboard containers, cleaning waste, and engine waste. Big data analytics is helping Carnival to reduce and recycle waste and cut costs.

Beyond cost reduction, Carnival is also using big data and artificial intelligence (AI) for price optimization and personalization. The company’s data science team regularly analyzes internal and external data from past consumers and vacationers to help match customers to the right cruise experience, whether that is a budget family cruise or a luxury voyage. Similarly, Carnival uses wearable devices and thousands of sensors to make personalized recommendations to customers for onboard activities. Delighting customers helps drive higher revenues.

As she contemplates next steps, Paige goes through the data science framework, a seven-step decision-making process for applying machine learning techniques in business situations:

Gain a business understanding of the problem
Obtain and explore relevant data
Prepare the data
Build a model
Evaluate the model
Visualize and communicate insights
Deploy the model

We discuss each of these steps in subsequent sections. As you go through these steps, keep the “big picture” in mind. Data Science relies on backtesting and feedback to choose among models. To do this, data scientists subdivide their data into a training sample (used in Step 4 to build a model), a cross-validation sample (used in step 5 to evaluate and choose among models), and a holdout sample (used in step 5 to test how the model would perform on completely new data). Backtesting helps the data scientist to decide how detailed a model to build. Too simple a model may miss important features for making good predictions; too detailed a model may introduce irrelevant features that weaken predictions.