Diabetes_analysis_Full

Diabetes Data Analysis Project

Project Explanation

Project Overview

This project analyzes a Diabetes Dataset using SQL and Tableau. The goal is to uncover insights about diabetes risk factors, patient demographics, and health indicators such as HbA1c levels, BMI, Cholesterol, and Age groups. The analysis helps in understanding trends in diabetes and visualizing key patterns for better decision-making.

Dataset Information

The dataset contains medical records with attributes such as:

• AGE - Patient's age
• GENDER - Male/Female
• HbA1c - Blood sugar levels (important for diabetes diagnosis)
• BMI - Body Mass Index
• Cholesterol, LDL, HDL - Lipid profile indicators
• CLASS - Diabetes classification (Y = Diabetic, N = Non-Diabetic, P = Pre-Diabetic)

Data Cleaning and Preparation in SQL

Before performing analysis, we first cleaned the dataset using SQL:

1. Handling NULL Values

To check for missing values in the dataset:

SELECT * FROM DiabetesDataset
WHERE AGE IS NULL OR GENDER IS NULL OR HbA1c IS NULL OR BMI IS NULL;

To replace NULL values with appropriate defaults:

UPDATE DiabetesDataset
SET BMI = (SELECT AVG(BMI) FROM DiabetesDataset WHERE BMI IS NOT NULL)
WHERE BMI IS NULL;

2. Standardizing Gender Values

To replace inconsistent gender values (F, f → F, M, m → M):

UPDATE DiabetesDataset
SET GENDER = 'F' WHERE GENDER = 'f';
UPDATE DiabetesDataset
SET GENDER = 'M' WHERE GENDER = 'm';

3. Identifying and Removing Duplicate Entries

Check for duplicates:

SELECT ID, COUNT(*)
FROM DiabetesDataset
GROUP BY ID
HAVING COUNT(*) > 1;

Remove duplicates while keeping the first occurrence:

DELETE FROM DiabetesDataset
WHERE ID NOT IN (
    SELECT MIN(ID)
    FROM DiabetesDataset
    GROUP BY No_Pation, AGE, GENDER, HbA1c, BMI
);

SQL Analysis Process

1. Exploratory Data Analysis (EDA)

• Used COUNT(), AVG(), and GROUP BY to understand the dataset.
• Analyzed age distribution and average health indicators.
• Identified correlations between HbA1c, BMI, and Cholesterol levels.

2. Key SQL Queries

1. Count of Diabetic vs. Non-Diabetic Patients

SELECT CLASS, COUNT(*) AS total_patients
FROM DiabetesDataset
GROUP BY CLASS;

2. Average HbA1c Levels by Age Group

SELECT 
    CASE 
        WHEN AGE <= 20 THEN '0-20'
        WHEN AGE <= 40 THEN '21-40'
        WHEN AGE <= 60 THEN '41-60'
        ELSE '61-80'
    END AS AgeGroup,
    AVG(HbA1c) AS Avg_HbA1c
FROM DiabetesDataset
GROUP BY AgeGroup;

3. Identifying High-Risk Patients (High HbA1c & BMI)

SELECT ID, AGE, BMI, HbA1c
FROM DiabetesDataset
WHERE HbA1c > 6.5 AND BMI > 30
ORDER BY HbA1c DESC;

Tableau Visualizations

We created interactive dashboards in Tableau to visualize key insights. Some of the most important visualizations include:

1. Age Distribution of Patients

• A histogram showing the distribution of ages among patients.

2. HbA1c vs. BMI Scatter Plot

• A scatter plot identifying relationships between HbA1c levels and BMI, helping to detect high-risk individuals.

3. Gender Distribution

• A bar chart showing the proportion of male vs. female patients.

4. Cholesterol, LDL, and HDL Levels by Age

• A line chart tracking trends in cholesterol levels with age.

5. Diabetes Classification Pie Chart

• A pie chart breaking down diabetic, pre-diabetic, and non-diabetic patients.

6. Full Tableau Dashboard

View the interactive Tableau dashboard here: Diabetes Analysis Tableau Dashboard

Screenshots

Tableau Dashboard

How to Run the Project

1. SQL Analysis

   • Import the dataset into PostgreSQL, MySQL, or any SQL database.
   • Run the SQL scripts provided in the SQLL.SQL file.

2. Tableau Visualization

   • Open Tableau Public/Desktop.
   • Load the dataset (Dataset of Diabetes CSV).
   • Import the pre-built Tableau workbook (Diabetes.twb).

Key Takeaways

• Patients with high HbA1c and BMI are more likely to be diabetic.
• The majority of patients fall within the 40-60 age group.
• Gender-based patterns in diabetes prevalence can be explored further.
• Cholesterol levels increase with age, and LDL trends correlate with diabetes risk.

Contributing

Feel free to contribute by improving the analysis, adding visualizations, or refining the queries! Open a pull request if you have any suggestions.

Contact

For any questions or discussions, reach out via GitHub Issues or email.