The goal of our project was to dive into an HR Analytics dataset from Kaggle and uncover trends in the data scientist hiring process by analyzing the candidate pool in terms of various demographic and professional characteristics. As students and current practitioners of data science, we are always intrigued by questions such as “What makes a good candidate for a data scientist?” and “Where do I stand in the job market in this region for this year?” Our motivation was to dissect the anonymized dataset with as much detail as possible from gender, education level, city, and years of experience to understand which candidates may be looking for jobs, whether it’s their very first foot in the door or a second/third position as a Data Scientist.
We decided to include
- A navigation for the users to easily toggle between the intro, the write up, the different data analyses, both macro and micro (gender and education-specific), and also our own contact information if they wish to get in touch with us for further inquiry.
- A user form where we predict the visitor’s likelihood of looking for a new job as a Data Scientist based on the provided dataset to engage and encourage the users to not only understand the general summary statistics of the given dataset but also to utilize it to see how it can applied to their current employment/job search situation. After the user types in their name, we thank them with a green popup “Thanks, 'user’s name!'" to motivate them to provide further information. We made the years of experience response a range scale and the rest a dropdown style, so that we receive as clean of data points as possible. We made sure to layer with nuance, that in cases where it’s difficult to accurately make the prediction, we surface the message “we’ll need more information from you!” to inform them of the limitations of the anonymized dataset. The alternative solution we considered was creating a separate tab for this, but ultimately decided to keep it as a sidebar that follows the user across different tabs, to keep the app dynamic and fun.
- Box plots, bar graphs, pie graphs, scatter plots, and snapshots of the raw dataset along with an interactive drop down filters just for gender and education where once the user selects a specific gender/education, they're able to further drill down the visualizations. We hypothesized that gender and education biases play a significant role in HR analytics especially in tech and in the data science function and also hypothesized that our visitors are curious to learn the insights.
Each of us created more than six graphs per person. Before we chose the datasets, we discussed our common interests and found the datasets from kaggle. We met three times for two weeks and spent about 10 hours finishing this assignment. Since both of us did not have experience in developing apps or using streamlit, we first had to learn how to code in the Streamlit. Then, we brainstormed and planned out how to clean and analyze the dataset. After creating graphs, debugging took the most time. We had to figure out what caused the errors by searching the error code online.