Certified Data Analyst Professional

The big picture

Artificial Intelligence (AI) is the broad goal of making machines do things that normally need human intelligence. Machine Learning (ML) is the main way we reach that goal today: instead of writing rules by hand, we let a program learn patterns from data. Deep Learning (DL) is a powerful branch of ML that uses neural networks with many layers. Data Science is the wider craft of turning raw data into decisions — it uses statistics, programming, and ML together.

Three kinds of machine learning

• Supervised learning — you have labeled examples (input → known answer) and the model learns to predict the answer. E.g. predict a house price from its features.
• Unsupervised learning — no labels; the model finds structure on its own, such as grouping similar customers (clustering).
• Reinforcement learning — an agent learns by trial and reward, like a program learning to play a game.

The data-science workflow

Almost every project follows the same loop. Knowing it keeps you oriented no matter how complex the problem gets:

• 1. Define the question and success metric with stakeholders.
• 2. Collect and store the data (files, databases, APIs).
• 3. Clean and explore it (EDA) — understand what you actually have.
• 4. Prepare features and split into train/test sets.
• 5. Train models, evaluate, and tune them.
• 6. Deploy the chosen model and monitor it in production.

Where the tools fit

In this program you will use Python as the language, SQL to pull data from databases, NumPy/Pandas to explore it, Matplotlib/Plotly to visualize it, scikit-learn to build models, and Streamlit to ship them. Each later topic in this guide is one stop on that path.

Descriptive statistics: center and spread

First, summarize a column. The center is the mean (average) or median (middle value — robust to outliers). The spread is the standard deviation: small means values hug the mean; large means they scatter. Always look at both — the same mean can hide very different data.

Distributions & the normal curve

A distribution shows how often each value occurs. The famous normal (bell) curve appears everywhere. Its rule of thumb: about 68% of values fall within 1 standard deviation of the mean, 95% within 2, and 99.7% within 3. That is how you judge whether a value is 'unusual'.

Probability in one minute

• Probability measures how likely an event is, from 0 (never) to 1 (certain).
• Conditional probability P(A|B) — the chance of A given B already happened — is the heart of many models (e.g. spam filters).
• Independent events multiply: P(A and B) = P(A)·P(B) when one does not affect the other.

Inference: is this pattern real?

You rarely have all the data — only a sample. Inferential statistics lets you generalize from the sample to the population. A hypothesis test asks: 'if there were truly no effect, how surprising is what I saw?' The p-value answers it: a small p-value (< 0.05 by convention) means the result is unlikely to be pure chance.

Correlation ≠ causation

Correlation measures how two variables move together (from -1 to +1). It is essential for feature selection — but two things can correlate without one causing the other (ice-cream sales and drownings both rise in summer). Proving causation needs a controlled experiment, not just a correlation.

Formulas & references

A formula starts with =, and the magic is cell references: write the logic once and fill it down. The key distinction is relative (A1, shifts when copied) vs. absolute ($A$1, stays fixed) — mixing these up is the #1 Excel bug.

=SUM(B2:B100)            total
=AVERAGE(B2:B100)        mean
=IF(C2>1000,"High","Low") condition
=VLOOKUP(A2, Sheet2!A:D, 4, FALSE)  join by key
=XLOOKUP(A2, ids, prices)           modern lookup

Lookups join your tables

VLOOKUP / XLOOKUP pull a value from another table by matching a key — the spreadsheet version of a SQL JOIN. XLOOKUP is the modern replacement: it looks left or right and returns a clean 'not found' message.

PivotTables — summarize in seconds

A PivotTable groups and aggregates thousands of rows by drag-and-drop: put a category in Rows, a number in Values, and Excel builds the summary. It is the same idea as SQL's GROUP BY, without writing code — the single most valuable Excel skill for analysts.

Clean before you analyze

• Remove Duplicates, Text to Columns, and TRIM/CLEAN fix messy imports fast.
• Data Validation stops bad entries at the source (dropdown lists, ranges).
• Conditional Formatting turns a table into a heat map for instant patterns.

Tables, rows, columns

A relational database stores data in tables — a grid of rows (records) and columns (fields). Each table usually has a primary key that uniquely identifies a row, and tables relate to each other through those keys. MySQL is the engine you will use.

The anatomy of a query

Reading data is one statement whose clauses always run in a fixed logical order. Learn this skeleton and you can read almost any query.

SELECT   city, COUNT(*) AS deals, AVG(price) AS avg_price
FROM     sales
WHERE    year = 2025
GROUP BY city
HAVING   COUNT(*) > 10
ORDER BY avg_price DESC
LIMIT    5;

JOINs — combining tables

Data you need is usually spread across tables (customers, orders, products). A JOIN stitches them back together on a shared key. INNER JOIN keeps only matches; LEFT JOIN keeps every row from the left table even when the right has no match.

SELECT o.id, c.name, o.total
FROM   orders o
JOIN   customers c ON c.id = o.customer_id
WHERE  o.total > 1000;

Aggregation is where insight lives

• COUNT, SUM, AVG, MIN, MAX summarize many rows into one number.
• GROUP BY computes those summaries per category (per city, per month).
• This single pattern answers most business questions: 'what is the average/total X per Y?'

Power Query: get and clean data

Before visuals, Power Query connects to sources (Excel, SQL, web) and cleans them with recorded, repeatable steps — remove columns, split, pivot, merge. Because the steps are saved, a refresh re-runs the whole cleanup on new data automatically.

The data model & star schema

Power BI shines when you relate multiple tables instead of one giant flat sheet. The best practice is a star schema: a central fact table (transactions) linked to dimension tables (date, product, customer). Good relationships make every visual filter every other one.

DAX: measures that calculate on the fly

DAX is Power BI's formula language. A measure recomputes for whatever the user has filtered — so 'Total Sales' automatically means 'total sales for this region, this month' when they click a slicer. The key mental model is filter context: a measure's value depends on what's selected around it.

Total Sales = SUM(Sales[Amount])

Sales YoY % =
VAR ThisYear = [Total Sales]
VAR LastYear = CALCULATE([Total Sales],
                 SAMEPERIODLASTYEAR(Dates[Date]))
RETURN DIVIDE(ThisYear - LastYear, LastYear)

Design for the decision

• Lead with the headline KPI, then supporting detail — top-left is read first.
• Slicers let users self-serve; tooltips add depth without clutter.
• Publish to the Power BI Service to share and schedule automatic refreshes.

Tableau — visual-first analysis

Tableau is built around drag-and-drop exploration: drop a field on Rows or Columns and it draws the best chart instantly. It excels at fast visual discovery and polished, interactive dashboards, and it speaks the same grammar as Power BI — dimensions (categories) vs. measures (numbers).

Looker Studio — free and web-native

Google's Looker Studio connects natively to Google Sheets, BigQuery, Google Ads and Analytics, and shares like a Google Doc. It's the go-to for marketing and web reporting when the data already lives in the Google ecosystem.

Pick the right tool

• Power BI — deep modelling (DAX), Microsoft/enterprise stacks.
• Tableau — fast visual exploration and best-in-class chart aesthetics.
• Looker Studio — free, Google-native marketing/web dashboards.
• Power Automate / AI tools — schedule refreshes and trigger alerts on data.