joseenrique-dev/h1b-visa-petitions

Predicting whether an individual will be able to obtain an h1B Visa

H1B-Visa-Petitions

Predicting whether an individual will be able to obtain an h1B Visa

Overview

H-1B visas allow U.S employers to employ foreign workers in highly technical occupations. However, not all applicants receive the immigration status and many are rejected during the certification process.

Can we use machine learning and data analytics to predict who will be successfully granted an H-1B Visa?

I will use both Logistic Regression and a Deep Neural Network developed in TensorFlow and compare the two methods and their respective accuracy.

The dataset can be found on Kaggle

Summary Statistics

Here are some basic statistics about the wages of the applicants

Statistic	Prevailing Wage
count	3,002,373
mean	146,998
std	5,287,609
min	0
25%	54,371
50%	65,021
75%	81,432
max	6.997607e+09

Here are some statistics on the qualitative variables

Statistic	Employers	Job Title	Full Time Position	Worksite
Number	263014	287550	3	18622
Most Popular	INFOSYS LIMITED	PROGRAMMER ANALYST	FULL TIME	NEW YORK, NEW YORK

H-1B Destination Visuallization

This map shows where in the US are the applicants trying to work

Neural Network Training

Architecture

Training

Conclusions

The accuracy of the Logistic Regression Model is 0.954635108481 or 95%

The accuracy of a Neural Network with 2 hidden layers with 10 neurons each (a very small network) is 0.955852508545 or 96%