Understanding the factors that influence L1 work visa for US
With increasing number of visa applications to US, USCIS has intensified their processes to approve candidates work visas. It is important for an applicant and their employer to know if the visa application would be approved or not. On one hand, applicants make a life changing decision of leaving their home country and starting a new life in the US. On the other, employers spend a lot of money looking to recruit the right candidate and applying for their visa. Thus, knowing about the chances of visa approval and the levers that can be adjusted (to an impact) will help them put their best foot forward. In this project, we choose to restrict our scope of analysis to the L1 visa category.
When an employer makes a decision to sponsor L1 visa for their employee, they are impressed by the skills of the employee and want to make use of the employee’s talents in their US office. From USCIS perspective these visas have to be granted to an applicant with high qualifications without affecting the employment opportunity of a US citizen with similar qualifications. So, these applications are thoroughly vetted to make sure that there are no qualified US employees to fill that job and also to make sure that the foreign applicants are paid wages as per the requirements for that position. Given that L1 visa applicants have higher chance to qualify for an EB-1 category Green card application, they also have to perform careful background checks to verify the credentials of the applicant and any associated risks pertaining to their country of origin. Our goal is to identify the factors that have an influence in the application decision as it would help the applicants and employers to identify drawbacks and make better informed decisions.
In order to understand these factors, we begin by collecting and cleaning the data. We then move on to determine what factors are statistically significant, and which ones are not. Moving on, we will train a couple of machine learning algorithms and understand which feilds/ factors does the model use to make decisions. Lastly, we will train a bunch of models and compare their performances based on precision, recall and accuracy metrics.
