layout

By default, the <title> tag for this page is determined by the title field (in the Content group). You can override the tag value here.

metaTitle

media

bottomSections

By default, the description tag for this post is taken from the Excerpt field. You can override the tag value here.

metaDescription

By default, the "og:image" meta tag set for social sharing this post points to the Featured Image field (in the Content group). You can override that image here.

socialImage

markdown_content

If enabled, the title suffix defined in the site configuration is appended to the title tag of this page.

addTitleSuffix

To add or override any meta tag for this page, add entries to this list. Entries defined here take precedence over any other defaults.

metaTags

title

colors

date

author

excerpt

featuredImage

Divider

## **Project Overview and Purpose**

*   Built a Predictive Model that can predict whether or not an employee will leave the company and help identify factors that contribute to leaving

## **Logistic Regression model assumptions**

*   Outcome variable is categorical

*   Observations are independent of each other

*   No severe multicollinearity among X variables

*   No extreme outliers

*   Linear relationship between each X variable and the logit of the outcome variable

*   Sufficiently large sample size

## **During the project, I tried to answer these questions:**

*   Do you notice anything odd?

*   Which independent variables did you choose for the model and why?

*   Is each of the assumptions met?

*   How well does your model fit the data?

*   Can you improve the model? Is there anything you would change about the model?

## **Model Building (Logistic Regression)**

Fit a model that predicts the outcome variable using two or more independent variables.

### Identify the type of prediction task

The goal is to predict whether an employee leaves the company, which is a categorical outcome variable. So this task involves classification. More specifically, this involves binary classification, since the outcome variable `left` can be either 1 (indicating employee left) or 0 (indicating employee didn't leave).

### Identify the types of models most appropriate for this task

Since the variable we want to predict (whether an employee leaves the company) is categorical, we need to build a **Logistic Regression model**. The binomial logistic regression suits our task because it involves binary classification.

![](/images/Screen%20Shot%202023-04-19%20at%206.14.18%20PM.png)

Next, I created a heatmap to visualize how correlated variables are and consider which variables are interesting for examining correlations between them.

![](/images/Screen%20Shot%202023-04-19%20at%206.17.16%20PM.png)

Next, I created a stacked bart plot to visualize number of employees across departments, comparing those who left with those who didn't:

![](/images/Screen%20Shot%202023-04-19%20at%206.20.22%20PM.png)

Since logistic regression is quite sensitive to outliers, it would be a good idea at this stage to remove the outliers in the `tenure` column that were identified earlier (Part 1: EDA of this dataset).

![](/images/Screen%20Shot%202023-04-19%20at%206.22.20%20PM.png)

Next, I isolated the outcome variable, which is the variable we want our model to predict:

![](/images/Screen%20Shot%202023-04-19%20at%206.25.11%20PM.png)

Next, I selected the features we want to use in our model. On this step we also need to consider which variables will help us predict the outcome variable, `left`

``![](/images/Screen%20Shot%202023-04-19%20at%206.27.18%20PM.png)

Next, we need to split the data into a training set and a testing set:

![](/images/Screen%20Shot%202023-04-19%20at%206.31.39%20PM.png)

Then construct a logistic regression model and fit it to the training dataset:

![](/images/Screen%20Shot%202023-04-19%20at%206.31.46%20PM.png)

After that,  we need to test the logistic regression model (we have to use the model to make predictions on the test set):

![](/images/Screen%20Shot%202023-04-19%20at%206.31.55%20PM.png)

## **Results**

Next, I created a confusion matrix to visualize the results of the logistic regression model:

****![](/images/Screen%20Shot%202023-04-19%20at%206.32.10%20PM.png)

The upper-left quadrant displays the number of true negatives.

The upper-right quadrant displays the number of false positives.

The bottom-left quadrant displays the number of false negatives.

The bottom-right quadrant displays the number of true positives.

True negatives: The number of people who did not leave that the model accurately predicted did not leave.

False positives: The number of people who did not leave the model inaccurately predicted as leaving.

False negatives: The number of people who left that the model inaccurately predicted did not leave

True positives: The number of people who left the model accurately predicted as leaving

A perfect model would yield all true negatives and true positives, and no false negatives or false positives.

## **Model Evaluation**

Next, I created a classification report that includes precision, recall, f1-score, and accuracy metrics to evaluate the performance of the logistic regression model.

In this step, we have to check the class balance in the data - check the value counts in the `left` column. Since this is a binary classification task, the class balance informs the way we interpret accuracy metrics.

***

There is an approximately 83%-17% split. So the data is not perfectly balanced, but it is not too imbalanced. If it was more severely imbalanced, you might want to resample the data to make it more balanced. In this case, we can use this data without modifying the class balance and continue evaluating the model.

![](/images/Screen%20Shot%202023-04-19%20at%206.37.37%20PM.png)

*Our evaluation metrics:*

*   **AUC** is the area under the ROC curve; it's also considered the probability that the model ranks a random positive example more highly than a random negative example.

*   **Precision** measures the proportion of data points predicted as True that are actually True, in other words, the proportion of positive predictions that are true positives.

*   **Recall** measures the proportion of data points that are predicted as True, out of all the data points that are actually True. In other words, it measures the proportion of positives that are correctly classified.

*   **Accuracy** measures the proportion of data points that are correctly classified.

*   **F1-score** is an aggregation of precision and recall.

## **Results**

The classification report shows that the logistic regression model achieved a precision of 80%, recall of 83%, f1-score of 80% (all weighted averages), and accuracy of 83%, on the test set.

## **Conclusion and Recommendations**

The models and the feature importances extracted from the models confirm that employees at the company are overworked.

To retain employees, the following recommendations could be presented to the stakeholders:

*   Cap the number of projects that employees can work on.

*   Consider promoting employees who have been with the company for at least four years, or conduct a further investigation about why four-year tenured employees are so dissatisfied.

*   Either reward employees for working longer hours, or don't require them to do so.

*   If employees aren't familiar with the company's overtime pay policies, inform them about this. If the expectations around workload and time off aren't explicit, make them clear.

*   Hold company-wide and within-team discussions to understand and address the company work culture, across the board and in specific contexts.

*   High evaluation scores should not be reserved for employees who work 200+ hours per month. Consider a proportionate scale for rewarding employees who contribute more/put in more effort.



Employee Retention. Part 2: Predictive Modeling

## **Project Overview and Purpose**

*   Perform Exploratory Data analysis to figure out which variables have an impact on employee retention.
*   Provide data-driven suggestions and insights for the HR department, based on my understanding of the data
*   Built a Predictive Model that can predict whether or not an employee will leave the company and help identify factors that contribute to leaving

## **Dataset**

In this dataset, there are 14,999 rows, 10 columns, and these variables:

![](/images/Screen%20Shot%202023-04-19%20at%203.13.43%20PM.png)

## **Technology Stack:**

*   Python Packages: Scikit-Learn, NLTK, Pandas, SciPy, Seaborn  (load, explore, extract, and organize information)
*   Jupyter Notebook (Exploratory Data Analysis)

## **Perform the Project**

## Step 1. Imports and load dataset

![](/images/Screen%20Shot%202023-04-19%20at%203.19.57%20PM.png)

![](/images/Screen%20Shot%202023-04-19%20at%203.31.52%20PM.png)

## Step 2. Data Exploration: Initial EDA and Data Cleaning

*   Understand variables

*   Clean the dataset (missing data, redundant data, outliers)

**Gather basic information about the data**

![](/images/Screen%20Shot%202023-04-19%20at%203.38.18%20PM.png)

**Gather descriptive statistics about the data**

****![](/images/Screen%20Shot%202023-04-19%20at%203.38.32%20PM.png)

**Rename columns**
As a data cleaning step, we need to rename some columns.  The column names were standardized - all in snake_case, correct column names that were misspelled, also column names were made more concise.

![](/images/Screen%20Shot%202023-04-19%20at%203.47.56%20PM.png)

**Check for any missing values in the data**

****![](/images/Screen%20Shot%202023-04-19%20at%204.07.39%20PM.png)

![](/images/Screen%20Shot%202023-04-19%20at%204.08.56%20PM.png)

**Check for outliers in the data**

![](/images/Screen%20Shot%202023-04-19%20at%204.10.27%20PM.png)

![](/images/Screen%20Shot%202023-04-19%20at%204.11.17%20PM.png)

## Step 3. Continue EDA: Analyze Relationships Between Variables

Now, I started examining the variables and creating plots to visualize relationships between variables in the data, comparing employees who stayed versus those who left

### **Number of projects and average monthly hours**

**First,** I created a stacked boxplot (visualizing distributions within data) showing `average_monthly_hours` distributions for `number_project`,  comparing the distributions of employees who stayed versus those who left.

Also, I plotted a stacked histogram to visualize the distribution of `number_project` those employees who stayed and those who left, because, although box plots are very useful in visualizing distributions within data, they can be deceiving without the context of how big the sample sizes that they represent are.

![](/images/Screen%20Shot%202023-04-19%20at%204.21.54%20PM.png)

It might be natural that people who work on more projects would also work longer hours. This appears to be the case here, with the mean hours of each group (stayed and left) increasing with the number of projects worked. However, a few things stand out from this plot:

1.  There are **two groups of employees** who left the company: **(A)** those who worked considerably less than their peers with the same number of projects, and **(B)** those who worked much more. Of those in group A, it's possible that they were fired. It's also possible that this group includes employees who had already given their notice and were assigned fewer hours because they were already on their way out the door. For those in group B, it's reasonable to infer that they probably quit. The folks in group B likely contributed a lot to the projects they worked on; they might have been the largest contributors to their projects.

2.  Everyone with **7 projects** left the company, and the interquartile ranges of this group and those who left with six projects was ~255–295 hours/week—much more than any other group.

3.  The **optimal number of projects** for employees to work on seems to be **3–4**. The ratio of left/stayed is very small for these cohorts.

4.  If we assume a work week of 40 hours and two weeks of vacation per year, then the average number of working hours per month of employees working Monday–Friday `= 50 weeks * 40 hours per week / 12 months = 166.67 hours per month`. This means that, aside from the employees who worked on two projects, every group—even those who didn't leave the company—worked considerably more hours than this. **It seems that employees here are overworked.**

As the next step, I needed to confirm that all employees with 7 projects left:

![](/images/Screen%20Shot%202023-04-19%20at%204.26.38%20PM.png)

This confirms that all employees with 7 projects did leave.

### **Average monthly hours versus the satisfaction levels**

![](/images/Screen%20Shot%202023-04-19%20at%204.33.04%20PM.png)

The scatterplot above shows that there was a **sizeable group of employees who worked ~240–315 hours per month**. 315 hours per month is over 75 hours per week for a whole year. **It's likely this is related to their satisfaction levels being close to zero.**

The plot also shows **another group of people who left**, those who had more **normal working hours**. Even so, their **satisfaction** was only around **0.4**. It's difficult to speculate about why they might have left. It's possible they felt pressured to work more, considering so many of their peers worked more. And that pressure could have lowered their satisfaction levels.

Finally, there is a **group who worked ~210–280 hours** per month, and they had **satisfaction** levels ranging **~0.7–0.9.**

### **Satisfaction levels and tenure**

It might be interesting to visualize satisfaction levels by tenure:

![](/images/Screen%20Shot%202023-04-19%20at%204.43.35%20PM.png)

**My observations from this plot:**

*   Employees who **left** fall into two general categories: **dissatisfied** employees with **shorter tenures** and very **satisfied** employees with **medium-length tenures**.

*   **Four-year** employees who left seem to have an unusually **low** satisfaction level. It's worth investigating changes to company policy that might have affected people specifically at the four-year mark, if possible.

*   The **longest-tenured** employees **didn't leave**. Their satisfaction levels aligned with those of newer employees who stayed.

*   The histogram shows that there are relatively few longer-tenured employees. It's possible that they're the higher-ranking, higher-paid employees.

### **Mean and median satisfaction scores of employees who left and those who didn't**

![](/images/Screen%20Shot%202023-04-19%20at%204.47.56%20PM.png)

As expected, the mean and median satisfaction scores of employees who **left** are **lower** than those of employees who stayed. Interestingly, among employees who stayed, the mean satisfaction score appears to be slightly below the median score. This indicates that satisfaction levels among those who stayed might be skewed to the left.

### **Salary levels and different tenures**

Let's examine salary levels for different tenures:

![](/images/Screen%20Shot%202023-04-19%20at%204.50.03%20PM.png)

The plots above show that **long-tenured** employees were **not disproportionately comprised** of **higher-paid** employees.

### **Average monthly hours and  evaluation scores**

Next, I explored whether there's a correlation between working long hours and receiving high evaluation scores. I created a scatterplot of `average_monthly_hours` vs `last_evaluation:`

``![](/images/Screen%20Shot%202023-04-19%20at%204.53.19%20PM.png)

**The following observations can be made from the scatterplot above:**

*   The scatterplot indicates two groups of employees who left: **overworked** employees who performed very **well** and employees who worked slightly **under** the nominal monthly average of 166.67 hours with **lower** evaluation scores.

*   There seems to be a **correlation** between **hours** worked and **evaluation** **score**.

*   There isn't a **high percentage** of employees in the **upper left** quadrant of this plot, but working long hours doesn't guarantee a good evaluation score.

*   Most of the employees in this company work well over **167** hours per month.

### **Average monthly hours and promotion**

Next, let's examine whether employees who worked very long hours were promoted in the last five years:

![](/images/Screen%20Shot%202023-04-19%20at%204.57.27%20PM.png)

The plot above shows the following:

*   very few employees who were promoted in the last five years left

*   very few employees who worked the most hours were promoted

*   all of the employees who left were working the longest hours

### **Distribution of employees who left across departments**

Next, I  inspected how the employees who left are distributed across departments:

![](/images/Screen%20Shot%202023-04-19%20at%205.01.18%20PM.png)

![](/images/Screen%20Shot%202023-04-19%20at%205.01.36%20PM.png)

There doesn't seem to be any department that differs significantly in its proportion of employees who left to those who stayed.



### **Correlation heatmap between variables in the dataset**

Lastly, I  checked for strong correlations between variables in the data:

![](/images/Screen%20Shot%202023-04-19%20at%205.04.36%20PM.png)

The correlation heatmap confirms that the **number of projects**, **monthly hours**, and **evaluation scores** all have some positive correlation with each other, and whether an employee leaves is negatively correlated with their **satisfaction level**.



### **Insights**

It appears that employees are leaving the company as a result of **poor management**. Leaving is tied to longer **working hours**, **many projects**, and generally **lower satisfaction levels**. 

It can be ungratifying to work long hours and not receive promotions or good evaluation scores. There's a sizeable group of employees at this company who are probably burned out. 

It also appears that if an employee has spent more than six years at the company, they tend not to leave.





NYC Citi Bike Share Program Analysis. Part 2: The findings report

Usage Trends

Common Locations

Recent posts

Employee Retention. Part 2: Predictive Modeling

Employee Retention. Part 1: Exploratory Data Analysis

NYC Citi Bike Share Program Analysis. Part 2: The findings report

NYC Citi Bike Share Program Analysis. Part 1: Performing analysis