- Created a tool that estimates data science salaries to help data scientists negotiate their income when they get a job.
- Scraped over 1000 job descriptions from glassdoor using python and selenium.
- Engineered features from the text of each job description to quantify the value companies put on python, excel, aws, and spark.
- Optimized Linear, Lasso, and Random Forest Regressors using GridsearchCV to reach the best model.
- Built a client facing API using flask.
Python Version: 3.7
Packages: pandas, numpy, sklearn, matplotlib, seaborn, selenium, flask, json, pickle
For Web Framework Requirements: pip install -r requirements.txt
Scraper Github: https://github.com/arapfaik/scraping-glassdoor-selenium
Scraper Article: https://towardsdatascience.com/selenium-tutorial-scraping-glassdoor-com-in-10-minutes-3d0915c6d905
Flask Productionization: https://towardsdatascience.com/productionize-a-machine-learning-model-with-flask-and-heroku-8201260503d2
We scraped the data from glassdoor using our selenium scraper for the job postings of "Data Scientist" in the United States. Following features were extracted from the web :
- Job title
- Salary Estimate
- Job Description
- Rating
- Company
- Location
- Company Headquarters
- Company Size
- Company Founded Date
- Type of Ownership
- Industry
- Sector
- Revenue
- Competitors
The following steps were taken for cleaning our data and making it ready for model building.
- Removed rows without salary
- Parsed numeric data out of salary
- Added a column for if the job was at the company’s headquarters
- Made columns for if different skills were listed in the job description
Looked into distributions for the numerical columns and made bar charts/pivot tables to understand the categorical data too. Also made a correlation plot to check the dependency between our columns.
First, I transformed the categorical variables into dummy variables. I also split the data into train and tests sets with a test size of 20%.
I tried three different models and evaluated them using Mean Absolute Error. I chose MAE because it is relatively easy to interpret and outliers aren’t particularly bad in for this type of model.
1.Multiple Linear Regression – Baseline for the model
2.Lasso Regression – Because of the sparse data from the many categorical variables, I thought a normalized regression like lasso would be effective.
3.Random Forest – Again, with the sparsity associated with the data, I thought that this would be a good fit.
The Random Forest model far outperformed the other approaches on the test and validation sets.
- Random Forest : MAE = 11.06
- Linear Regression : MAE = 18.85
- Lasso Regression : MAE = 19.66
In this step, I built a flask API endpoint that was hosted on a local webserver by following along with the TDS tutorial in the reference section above. The API endpoint takes in a request with a list of values from a job listing and returns an estimated salary.