BCG Gamma

• A Data Scientist works with data to deliver value to a business.
• Data Science can sometimes overlap with domains such as Data Analysts, Data Engineers and DevOps.
• How much the Data Scientists role overlaps with these other roles depends on the company.
• A Data Scientist is generally focused on modelling data to be able to predict an outcome accurately.
• Core skills Data Scientists use statistics, mathematics, programming and communication.

1. Business understanding & problem framing: What is the context of this problem and why are they trying to solve it?
2. Exploratory data analysis & cleaning: What data are we working with, what does it look like, and how can we improve it?
3. Feature engineering: Can we enrich this dataset using our expertise or third-party information?
4. Modeling and evaluation: Can we use this dataset to accurately make predictions? If so, are they reliable?
5. Insights & Recommendations: How to communicate the value of these predictions by explaining them in a way that matters to the business?

• Exploratory data analysis (EDA) is a technique used by a Data Scientist to gain a holistic understanding of the data that they are working with.
• It is mainly based on using statistical techniques (such as descriptive statistics) and visualizations to gain a deeper understanding of the statistical properties that the data holds.

• Feature engineering refers to the addition, deletion, combination, and mutation of your data set to improve machine learning model training, leading to better performance and greater accuracy.
• In the context of this task, feature engineering refers to the engineering of the data to create new columns that will help us predict more accurately.
• Effective feature engineering is based on domain knowledge of the business problem and the available data sources.

• When you are trying to predict an outcome, the result that you’re trying to predict can either be:
• A continuous number, e.g. an employee salary or a discrete value, e.g. a job title

In our example, we are trying to predict whether or not a client will churn, so it will only ever be (True/False, 1/0, Yes/No, etc…).