What is Data Science? | AI Pathshala

Definition

Data science is the field of getting useful knowledge from data, using statistics, programming and knowledge of the problem area. It covers the whole path: asking the question, collecting and cleaning data, analysing it, building models where needed, and communicating the result.

In simple words

A kirana shop owner looks at last year's sales and notices that cold drinks sell out on hot weekends, so they stock more. That's data science at its smallest. At a big company the data is huge and the tools are stronger, but the thinking is the same.

Why it matters

Most business decisions don't need a neural network. They need someone to get the data right and answer the question honestly. Data skills also form the base of ML work, because a model can only be as good as the data behind it.

How it works

Role	Main question	Typical tools
Data Analyst	What happened, and why?	SQL, Excel, Power BI, pandas
Data Scientist	What will happen, and what should we do?	Python, statistics, scikit-learn
Data Engineer	How does data reach everyone reliably?	SQL, Spark, Airflow
ML Engineer	How does a model run reliably for users?	Python, Docker, MLOps tools

The usual steps

Understand the question, and what decision depends on it.
Collect the data and check where it came from.
Clean it. This usually takes the most time.
Explore it with charts and summaries.
Analyse or model it.
Communicate the result clearly, including what you don't know.

In code

Real analysis often starts this simply: group the data and compare.

sales.py

import pandas as pd

sales = pd.DataFrame({
    "day": ["Sat", "Sun", "Mon", "Tue", "Sat", "Sun", "Mon", "Tue"],
    "weather": ["hot", "hot", "hot", "mild", "mild", "mild", "mild", "hot"],
    "cold_drinks": [90, 95, 40, 22, 35, 38, 20, 44],
})
sales["weekend"] = sales["day"].isin(["Sat", "Sun"])
print(sales.groupby(["weekend", "weather"])["cold_drinks"].mean())

Output

weekend  weather
False    hot        42.0
         mild       21.0
True     hot        92.5
         mild       36.5
Name: cold_drinks, dtype: float64

Common mistakes

Starting with a model before understanding the question.
Skipping cleaning. Wrong data gives confident, wrong answers.
Reporting only what supports the result you hoped for.

Check yourself

1. Who is mainly responsible for reliable data pipelines?

2. Which step usually takes the most time in data work?

Summary

Data science turns data into useful decisions.
It blends statistics, code and knowledge of the problem.
Good data work comes before good models.

Finished reading?