Data Analysis

Pandas Data Analysis Project

Sonu Gupta — Sun, 30 Aug 2026 09:12:37 GMT

Introduction

This project was created to practice a complete data-analysis workflow using Python and Pandas.

I worked with a financial fraud detection dataset, focusing on data manipulation, structured processing, and exploring patterns within the data.

What I Did

Using Pandas and Python, I:

Processed and explored the dataset
Performed data manipulation and transformation
Analyzed patterns within the data
Used Python-based visualization to better understand the results

Tech Stack

Python · Pandas · Data Analysis

What I Learned

This project helped me strengthen my understanding of Pandas, dataset exploration, data transformation, and analytical workflows.

It also gave me more practice working with structured data before moving toward larger analytics projects.

Project

GitHub: https://github.com/mahii-17/pandas\_project

YouTube Analytics Dashboard: Turning Video Data into Actionable Insights

Sonu Gupta — Sun, 30 Aug 2026 09:04:33 GMT

How I built an interactive dashboard using Python, Pandas, NumPy, Plotly, and Streamlit to understand video performance.

Introduction

Creators generate a huge amount of data every time they publish a video.

Views, likes, comments, watch time, engagement, and other performance indicators can all tell a story. But looking at these numbers individually doesn't always make it easy to understand what is actually happening.

For this project, I built a YouTube Analytics Dashboard using Python to turn video-performance data into an interactive and easier-to-understand analytical experience.

The dashboard tracks 7+ performance KPIs, analyzes 30-day video performance, and uses percentile-based comparisons to help evaluate how an individual video performs relative to typical channel performance.

The Problem

YouTube provides a lot of information about content performance.

The challenge is not simply having access to the numbers.

The real challenge is answering questions such as:

How is a video performing?
Which performance metrics deserve attention?
Is a video's performance typical for the channel?
How can multiple performance indicators be viewed together?
Can raw analytics data be turned into something easier to interpret?

I wanted to build a dashboard that could bring these pieces together in one place.
Project Goal

The main goal of this project was to create an interactive dashboard that could:

Track important YouTube performance metrics

Analyze recent video performance

Compare individual videos with typical channel performance

Present analytical results through interactive visualizations

Instead of looking at raw rows of data, the dashboard provides a more structured way to explore video performance.

Dataset & Analysis

The project works with YouTube video-performance data and focuses on analyzing recent performance over a 30-day period.

The analysis combines multiple metrics so that performance can be viewed from different perspectives rather than relying on a single number.

This is particularly useful because a video with high views does not necessarily tell the complete story of its performance.

1. Data Preparation with Pandas

I used Pandas to work with the dataset and prepare the information for analysis.

This involved processing the available video data and structuring it in a way that could be used effectively by the dashboard.

Pandas was also used as part of the feature-engineering and analytical workflow.

2. Feature Engineering

Raw metrics are useful, but derived metrics can provide additional analytical context.

For this project, I used feature engineering to create a more meaningful way of analyzing video performance.

This allowed the dashboard to move beyond simply displaying raw values and instead support comparisons between different performance indicators.

3. KPI Tracking

One of the main parts of the dashboard is its KPI layer.

The dashboard tracks 7+ YouTube performance KPIs, giving users a quick overview of how the channel or videos are performing.

The idea was to make important information visible immediately rather than requiring users to inspect individual records.

4. 30-Day Performance Analysis

I focused on 30-day video performance to understand how videos were performing over a recent period.

This provides a more useful analytical window than looking at isolated values.

The dashboard can therefore be used to explore recent performance patterns and identify how videos are behaving within the selected period.

5. Percentile-Based Comparison

One of the more interesting parts of the project was comparing individual videos with typical channel performance using percentile-based insights.

Instead of asking only:

"How many views did this video receive?"

the analysis can ask:

"How does this video's performance compare with what is typical for the channel?"

This adds context to the raw metric.

A number becomes much more meaningful when you can understand where it sits relative to other observations.

Visualization with Plotly

To make the dashboard interactive, I used Plotly.

Interactive visualizations make it easier to explore performance data and switch between different analytical perspectives.

Rather than presenting static charts, the dashboard was designed to make the data itself more explorable.

Building the Dashboard with Streamlit

The final dashboard was built using Streamlit.

Streamlit made it possible to turn the Python-based analysis into an interactive application without building a traditional frontend from scratch.

The final workflow became:

YouTube Performance Data
          ↓
     Data Processing
          ↓
    Feature Engineering
          ↓
     KPI Calculation
          ↓
  Performance Analysis
          ↓
 Interactive Plotly Charts
          ↓
    Streamlit Dashboard

Tech Stack

Programming & Analysis

Python
Pandas
NumPy

Visualization

Plotly

Dashboard

Streamlit

What I Personally Built

This project gave me hands-on experience across multiple stages of the analytics workflow.

I worked on:

Preparing and processing the dataset
Performing feature engineering
Analyzing 30-day video performance
Building KPI-based analysis
Implementing percentile-based comparisons
Creating interactive Plotly visualizations
Developing the Streamlit dashboard

Rather than using the dashboard only as a visualization layer, I worked on the analytical workflow that feeds the dashboard.

Project Links

GitHub: https://github.com/mahii-17/yt-dashboard

Customer Shopping Behavior Analysis: From Raw Transactions to Customer Insights

Sonu Gupta — Sun, 30 Aug 2026 08:17:48 GMT

Introduction

Data analysis is not just about creating charts.

Before a dashboard can answer useful business questions, the underlying data needs to be clean, consistent, and structured in a way that makes analysis possible.

For this project, I worked with a retail customer dataset containing 3,900+ transactions and 18 features covering customer demographics, purchasing behavior, and product attributes.

The goal was to transform this raw transactional data into a reliable analytical dataset and use it to understand customer purchasing behavior.

The project workflow involved:

Data Cleaning → Feature Engineering → SQL Analysis → Power BI Visualization

The Problem

Retail transaction data can contain inconsistencies that make analysis difficult.

Before analyzing the data, I needed to address issues such as:

Missing values Inconsistent column names Redundant fields Data that was not immediately suitable for analysis

Simply loading the raw dataset into a dashboard would not guarantee meaningful results.

So, the first objective was to improve the quality of the dataset before performing deeper analysis.

Dataset

The dataset contains 3,900+ retail transactions with 18 features.

The available information covers areas such as:

Customer demographics Product attributes Purchasing behavior Transaction-related information

This combination made the dataset suitable for exploring customer segmentation and purchasing patterns.

My Approach

I divided the project into several stages.

Data Cleaning with Python

I started by examining the dataset and preparing it for analysis using Python and Pandas.

The cleaning process included:

Handling missing values Standardizing column names Removing redundant fields Checking the structure and consistency of the data

This preprocessing step improved the data quality by approximately 35% according to my project analysis.

This step was important because poor-quality input data can lead to misleading analysis later.

Feature Engineering

After cleaning the dataset, I created additional features that could provide more useful analytical perspectives.

Two examples were:

Age Groups

Instead of treating every customer age as an isolated value, customers could be grouped into meaningful age categories.

Purchase Frequency

I also engineered purchase-frequency information to help understand customer purchasing behavior and support better segmentation.

These transformations made the dataset more useful for analysis beyond the original raw columns.

SQL & MySQL Analysis

Once the dataset was prepared, I used SQL and MySQL to work with the structured data and perform analytical queries.

This part of the project helped me practice:

Querying relational data Filtering records Aggregating information Grouping customers and transactions Extracting patterns from transactional data

Using SQL alongside Python also helped me understand how the same analytical problem can be approached from different parts of a data workflow.

Power BI Dashboard

After preparing and analyzing the data, I used Power BI to turn the results into a visual format.

The purpose of the dashboard was to make customer and purchasing information easier to explore and interpret.

Instead of looking at thousands of individual transactions, the dashboard provides a higher-level view of the dataset and helps communicate the results more clearly.

Tech Stack Programming & Analysis Python Pandas Database MySQL SQL Visualization Power BI Workflow Raw Retail Data ↓ Data Cleaning ↓ Data Transformation ↓ Feature Engineering ↓ SQL / MySQL Analysis ↓ Power BI Dashboard What I Learned

This project taught me that a good analytics project is not simply about making visualizations.

The quality of the final result depends heavily on everything that happens before the visualization stage.

Through this project, I strengthened my understanding of:

Data Cleaning

Real datasets often require significant preparation before they become analysis-ready.

Feature Engineering

Creating useful derived features can reveal perspectives that are not immediately available in the original dataset.

SQL

SQL provides a powerful way to query and analyze structured relational data.

Data Visualization

A dashboard should communicate information clearly rather than simply display as many charts as possible.

End-to-End Analytics Workflow

Working across Python, SQL/MySQL, and Power BI helped me understand how different tools can fit together in a single analytics workflow.

Key Takeaway

The biggest lesson from this project was simple:

Good analysis starts with good data.

Cleaning and transforming the dataset was just as important as the final dashboard.

By taking the project from raw transactions to a structured analytical dataset and finally to a visualization layer, I was able to practice a complete data analytics workflow rather than working on isolated tasks.

Project Links

GitHub: Customer Shopping behavior

Portfolio: Click Here