Installations The project is on predicting the salary of programmers based on the StackOverFlow survey done in 2017.
The libraries used are:
pandas
numpy
seaborn
random
sklearn
matplotlib
missingno
Project Motivation The project is part of Udacity data Science nano-degree and is directed towards answering the following three questions:
-
In Which countries people earned more than the rest.
-
What is the relation between salary and eduction in these counries.
-
How to forecast the salary based on the survey data considering categrical data like country and eduction level?
File description
The analysis will follow the CRISP-DM steps:
-
Business Understanding
-
Data Understanding (access and explore)
-
Data cleaning and preparation of some categotrical columns into dummies to go into a linear model
-
Linear modeling
-
Evaluation
-
Deployment
Takeaways and summary fo teh analysis
1- The majority of respondents to the survey came from 10 countries; United States, India, United Kingdom, Germany, Canada, France, Poland, Australia, Russian Federation, and Spain. Programers in the United States earned the most while India and Russia earned the lowest.
2- The salary of a programmer is not related to education level as many earned more through a university degree. This could highlight the role of online education in building competent skillsets.
3- Our linear regression model that relied mainly on education, country and company size did not manage to predict salaried above $50,000 with good accuracy. More feature engineering work is required to enhance model performance.
Author: Mustafa Adel Amer