Over the past few weeks, I’ve taken a deep dive into HR analytics using Python. Starting with a dataset of employee records, I explored everything from basic data cleaning to advanced dimensionality reduction with PCA. This post is a reflection of what I’ve learned—broken down into four key stages: Exploratory Data Analysis (EDA), Business Analysis, Data Visualization, and PCA.
Whether you're an aspiring data analyst or an HR professional curious about data-driven decision-making, this walkthrough will show you how Python can turn spreadsheets into strategy.
Part A: Basic Exploratory Data Analysis (EDA)
Before diving into insights, I had to understand the data:
- Loaded the dataset using Pandas and previewed the first few rows
- Checked the shape to see how many rows and columns I was working with
- Inspected column types to identify numerical, categorical, and date fields
- Counted unique values to spot identifiers and categorical features
- Identified missing values using
.isnull()and planned data cleaning - Described numerical columns with
.describe()to understand distributions - Plotted salary distribution with Matplotlib to detect skewness
- Calculated average age from the DOB column using datetime operations
- Compared employment status (active vs terminated) using
.value_counts() - Identified largest departments using Seaborn’s
countplot()
Part B: Business Analysis
Next, I tackled questions that HR teams care about:
- Average salary by department using
groupby() - Employment status breakdown with a pie chart
- Gender pay comparison using Seaborn’s
boxplot() - Top recruitment sources via
.value_counts() - Diversity Job Fair attendance calculated from a Boolean column
- Engagement scores by department with a barplot
- Race-based salary averages using
groupby()and.mean() - Projects vs salary correlation visualized with a scatterplot
- Marital status and salary compared using a barplot
- Manager team sizes identified with
groupby().size()
Part C: Data Visualization
To make the data speak visually:
- Salary histogram to show distribution
- Department headcount with a countplot
- Satisfaction scores by department using a barplot
- Termination trends over time with datetime plots
- Gender-based salary boxplot to highlight disparities
- Performance vs salary stripplot to spot trends
- Correlation heatmap to reveal relationships between variables
- Engagement vs satisfaction scatterplot to explore alignment
- Stacked bar chart of employment status across departments
- Absenteeism distribution with a histogram
Part D: PCA (Dimensionality Reduction)
Finally, I explored Principal Component Analysis (PCA) to simplify the dataset:
- Standardized features using
StandardScaler()to prep for PCA - Applied PCA and interpreted the first two components
- Plotted explained variance to understand dimensional importance
- Visualized PCA-reduced data colored by department
- Identified top contributing variables to PC1 and PC2
- Condensed engagement, satisfaction, and absences into one dimension
- Grouped employees by performance in PCA space
- Compared clustering before and after PCA using KMeans
- Created a PCA biplot to show feature loadings
- Discussed PCA use cases in HR—like simplifying survey data or improving clustering
Final Thoughts
This journey taught me how to:
- Clean and explore data with Pandas
- Visualize insights with Seaborn and Matplotlib
- Answer strategic HR questions with analytics
- Simplify complexity using PCA
HR analytics isn’t just about dashboards—it’s about understanding people through data. Whether you're optimizing recruitment, improving engagement, or analyzing performance, Python gives you the tools to make smarter decisions.
Thanks for reading! If you’ve worked with HR data or PCA, I’d love to hear your experiences. Drop a comment or share your favorite Python trick for workforce analytics.
Top comments (0)