Parth4786 / Data-_Scientist_-Salary_-Prediction

This is a GitHub repository for “Data Scientist Salary Prediction”. It predicts salaries based on location, experience, education, skills, and industry. It uses machine learning algorithms such as linear regression, decision trees, and random forests. The project is open-source and written in Python.

Project Overview

• Created a machine learning model that estimates salary of data scientist based on the features like rating, company_founded, etc.
• Engineered features from the text of each job description to quantify the value companies put on python, excel, tableau and sql

How will this project help?

• This project helps data scientist/analyst to negotiate their income for an existing or a new job

Resources Used

• Packages: pandas, numpy, sklearn, matplotlib, seaborn.
• Dataset by Ken Jee: https://github.com/PlayingNumbers/ds_salary_proj

Exploratory Data Analysis (EDA) and Data Cleaning

• Removed unwanted columns: 'Unnamed: 0'
• Plotted bargraphs and countplots for numerical and categorical features respectively for EDA
• Numerical Features (Rating, Founded): Replaced NaN or -1 values with mean or meadian based on their distribution

• Categorical Features: Replaced NaN or -1 values with 'Other'/'Unknown' category
• Removed unwanted alphabet/special characters from Salary feature
• Converted the Salary column into one scale i.e from (per hour, per annum, employer provided salary) to (per annum)

Feature Engineering

• Creating new features from existing features e.g. job_in_headquaters from (job_location, headquarters), etc.

• Trimming columns i.e. Trimming features having more than 10 categories to reduce the dimensionality
• Handling ordinal and nominal categorical features
• Feature Selection using information gain (mutual_info_regression) and correlation matrix


• Feature Scaling using StandardScalar

Model Building and Evaluation

Metric: Negative Root Mean Squared Error (NRMSE)
• Multiple Linear Regression: -27.523
• Lasso Regression: -27.993
• Random Forest: -17.637
• Gradient Boosting: -24.429
• Voting (Random Forest + Gradient Boosting): -19.136
Note: Evaluation scores are obtained using cross validation.

Model Prediction