COPE (Covid19 Offset for Performance Evaluation)

Machine Learning based we application that factors in individual COVID19 impact on the performance evaluation of the employees

Comment

Inspiration

COVID19 has impacted the performance of employees •End of year reviews for many companies will start now in 4Q 2020 •Most affected employees include:

  1. Those who tested positive for COVID19
  2. Single Moms/Dads
  3. Moms/Dads with children
  4. Those taking care of seniors •Many women (~800,000 women as cited by NYTimes: https://www.nytimes.com/2020/10/03/us/jobs-women-dropping-out-workforce-wage-gap-gender.html) forced to leave their job to take care of dependents. Our motivation is how can the management evaluate employee not just based on performance at work, but also based on impact on performance due to COVID19 factors. This app helps show empathy towards employees affected by COVID19 and ensure employee satisfaction and reduces attrition in workforce in this difficult time.

What it does

Our Application uses a Machine Learning based model, that considers three types of data features: •Features related to past performance of employee, actual performance evaluation by management etc. •EHR features like gender, children and dependents, COVID19 positive/negative •Dependent factors like whether employee has homeschooled children or newborns, seniors at home.

It evaluate the unbiased performance score by management for the employee, and approximates the COVID19 impact on the employee in the form of a score. Our output is in two forms:

  1. Whether the employee may be impacted by COVID19
  2. COVID19 impact in the form of a score that can be used to offset the performance score by management

Employees have the option to opt in or out of this COVID19 based employee evaluation, as it requires personal data

How we built it

We explored the data using jupyter notebook and google cloud storage (which did not work entirely). We trained the Decision Tree ML model, KNN classifier and pushed it to our app. We built the python based application using streamlit.io, and pushed our trained model. Our model is having security for login, so that the application data is visible only to management and admins. Our model has two key features: Plot - to visualize overall employee data Prediction - This uses individual employee data and predicts the COVID19 impact. Here the management has the option to consider or not consider the COVID19 impact score in overall performance score of employee.

Challenges we ran into

1.Data: Finding the accurate dataset for employee performance

2.WebApp: Understanding streamlit.io and its packages Difficult to get the models to work in the application and give predictions

3.Overall: Finding relevant hackathon idea of app that will help us in the current COVID19 scenario

Accomplishments that we're proud of

  1. Proud to build an app that is relevant to COVID19 and will show empathy to company employees
  2. We are happy that our app is running and can be used with real-world data
  3. Our team of two (spanning two countries - US and India) could build this Machine Learning based app

What we learned

  1. We learned using streamlit.io to quickly pull and show ML apps.
  2. We also dabbled with Google Cloud (storage, prediction).
  3. Proxy a dataset and get good ML results
  4. Overall our learning curve was good this weekend

What's next for COPE (Covid19 Offset for Performance Evaluation )

  1. We wish to dabble with real-world employee data
  2. We hope our idea is actually used by companies. This will enable employee retention and less attrition.
  3. Write a blog outlining our efforts in this hackathon

Built With

Share this project:

Updates