1 of 24

INTRODUCTION TO �DATA SCIENCE

– FARDINA FATHMIUL ALAM

(fardina@umd.edu)

CMSC 320

LECTURE: 01

Lecture 1 Part 1

2 of 24

Today we will cover

Chapter 1: https://ffalam.github.io/CMSC320TextBook/chapter1/Chapter_1_0.html

What is Data Science?
What is Big Data and Its Key Features?
Overview of Data Science Life Cycle.
Career in Data Science.

Remember, slides are just the summary, you must have to follow the textbook.

PART 2: Course Logistics and What to Expect.

3 of 24

Let’s start by understanding what is data science?

4 of 24

Need availability of DATA!!!

5 of 24

DATA SCIENCE

Data science is all about using data to solve problems.

Decision making

Which email is spam and which is not?

Product recommendation

Which movie to watch?

Predicting the outcome

Who will be the next President of the USA?
Many more ………….

In simple terms: Using data to draw an inference or predict an outcome. Such information can help us to make better decisions.

6 of 24

WHAT IS DATA SCIENCE?

Data science is an interdisciplinary field focused on discovering patterns and describing relationships using data.

7 of 24

WHAT IS THIS COURSE ABOUT?

This semester we will learn to take raw data and turn it into insights about the world or predictions about the future.

8 of 24

AI: any technique that enables computers to mimic human intelligence, e.g., using decision trees, rules, logic, ML

ML: a subset of AI that uses statistical techniques that enable machines to use experience to improve at tasks

Deep Learning: a subset of ML that uses vast amounts of data and multilayer neural nets to enable a machine to train itself to perform tasks

Statistics: science of collecting and analyzing numerical data, either to describe properties of a dataset or to make inferences based on a subset of data

Data Science: an applied branch of statistics that uses computer science techniques to manage, analyze, visualize and discover patterns in data

9 of 24

V’s of Big Data

Volume: Massive amounts of data

requires tools like Spark, Hadoop, and cloud platforms (AWS/GCP/Azure).

Velocity: Data generated and processed rapidly (real-time).

e.g., 900M FB photos/day, 500M tweets/day, 3.5B Google searches/day.

Variety: Data comes in many forms;

tables, images, video, audio, text, sensors.

Additional:

Veracity: Data quality can be messy, noisy, incomplete, or uncertain.
Value: Data must be useful and actionable.

10 of 24

EXAMPLE: ANALYZE PATIENT’S MEDICAL RECORD, PREDICT DISEASE OUTBREAK

Raw Data: Medical records, patient demographics, lab results.

Applies Techniques and Models: Analyze patient data to identify risk factors, predict disease progression, and recommend personalized treatments.

Insights and Predictions: Identifies patterns in patient data to

predict disease outbreaks,
optimize treatment plans,
provide insights for medical research.

11 of 24

THE DATA LIFECYCLE

Data collection

Exploratory analysis

Data viz

Analysis, hypothesis testing, & ML

Insight & Policy Decision

Data processing

12 of 24

REMEMBER: DATA SCIENCE IS NOT A STRICTLY ONE-WAY LINEAR PROCESS; IT'S DYNAMIC, ITERATIVE, AND ADAPTIVE

Step ‘x’

Step ‘y’

If need to revisit previous steps due to new insights or challenges that arise during later stage. Example: Later realize you need to collect more data. This allows for constant refinement and improvement as new information emerges and insights evolve.

learning and building as we go

13 of 24

BEFORE THAT: DEFINE PROBLEM STATEMENT

What problem are you going to solve?

Why do we need a well-defined problem statement?

A problem well defined is a problem half-solved. — Charles Kettering

Example: “I want to increase the profit” - is it a well defined problem statement?

How much to increase the average profit/ revenue such as 20% or 30% ?

What is the average time frame to increase the revenue?

14 of 24

COLLECT DATA

Data collection is a systematic approach to gather relevant information from a variety of sources.

Gathered from external sources
Gathered from existing company databases
Gathered by tools created by you

2 types of data collections methods:

Primary Data Collection
Secondary Data Collections

15 of 24

PRIMARY vs SECONDARY DATA COLLECTIONS

Situation: Some unique problem and no related research is done on the subject.

Solution: Collect new data → Primary data collection.

Situation: Some problem and the data is readily available or collected by someone else.

Solution: Use the data → Secondary data collection.

16 of 24

2. DATA PROCESSING

Clean or scrub data to ensure the data quality

Important: Do sanity check on data.
Why ? Bad quality may lead to unexpected results or misleading information.

Deal with duplicates
Formatting
Weird outliers
Mistakes

17 of 24

3. EXPLORE DATA: FIGURE OUT WHAT YOU HAVE

You’ll be sitting on like a terrabyte of raw data

What is there?
Are there any interesting correlations?
Do you have everything you need?

Do Exploratory Data Analysis (EDA) : Descriptive statistics and Visualizations

Extract useful insights from the data, understanding patterns, and setting the stage for effective model building and decision-making

18 of 24

4. BUILD A MODEL

Steps to the solution.

Example: build a machine learning model that predicts a student's test score based on their study hours.

Train the model → using the historical data,
Once it's trained, you can input a student's study hours to get a predicted test score.

19 of 24

5. INTERPRETATION

Deriving Insights and make policy decisions if needed
Present the results from your analysis to the stakeholders.
Convince People: Explain the specific conclusion and critical findings, probably in understandable manner.

The last step is getting a bunch of non-technical people to understand what your magical model is doing and why it’s right and they should listen to you.

Translate these findings into actionable insights.

20 of 24

Career in Data Science

High Demand
Job Growth
High Salary Potential:
Versatility: Applies to various industries (e.g., healthcare, finance, marketing, tech)

21 of 24

CAREER IN DATA SCIENCE

22 of 24