What is data science?
data science seems really exciting but first, let us get our basics clear! What actually is data science? I’m not going to bore you with long lines of definition so here’s a short explanation:
Data Science is an amalgamation of Statistics, Computer Science, and specific domain knowledge.
Statistics and computer science are the generic fundamentals that can be perfected by studying and a little bit of practice. It is the domain knowledge that takes time, research, and effort to gain.
You don’t need to master each vertical but having a decent grip on all will help you in the long run.
Data Science is quite a big field in itself. It starts with simple data reporting activities to advanced predictive modeling using Artificial Intelligence. As you can observe by looking at the Data science spectrum below, the higher the complexity the higher its business value.
Data science is thrilling! Now, let’s look at the actual role of a data scientist.
What does a data scientist role look like?
Caution: These terms are losely used in the industry. The exact role can depend on the maturity of your organization in data initiatives.
The role of a data scientist is fairly expansive and will depend majorly on the type of project that you are working on. Here, we will discuss the general lifecycle of a data science project.
Understanding the problem statement – Seems really simple, right? Believe me, it isn’t. Understanding the problem statement will be the make-or-break situation for the complete duration of the project. At this stage, A team of data scientists and the concerned team go over the objectives and expected requirements of the project. It requires good communication skills, stakeholder management for this step. A good data scientist won’t hesitate to spend an ample amount of time on this step. Once the problem statement is clear, the data scientist can move on to the collection of data
Gathering Data – Once the requirements are obtained and the hypothesis formed, the data scientist then proceeds to mine the needed data. The source of the data can vary such as company data warehouse, web scraping, and so on
Data Cleaning – This is the most time-consuming process of the entire data science project. It may take up to 80% of your time. Here, the data scientist will be munging, manipulating, wrangling the data. The time and effort are worth it since the health of your data will reflect the health of your output model. During this stage, the data scientist deals with outliers, missing data values, correcting the data types, and many other operations. This is not the most exciting step but the most essential one
Exploratory Data Analysis (EDA) – It is basically the step where the data scientist gets the “feel” of the data. It is at this stage that you can analyze each feature or multiple features in the dataset and check how they behave. You may also analyze the relationship of features with other features. You can expect a lot of data visualization at this stage. Be ready to gain some crucial insights during this stage that will help you in other steps
Feature Engineering – Feature engineering is not so much of a step but an art. It is an iterative process, going one by one through all the features and applying operations to improve the performance of the model. For example, you can combine some of the strong features and try to improve the model. It will require a lot of trial and error
Model Building – Model building in itself is relatively a fast step but planning is important. Do you want a model with high accuracy or a model that can return the importance of features? You will need to think upon and select your strategy for model building and its evaluation
Deployment – Once you have built and evaluated your model, it is finally time to deploy it in the real world.
