lOMoAR cPSD| 44635080

lOMoARcPSD|44635080
MACHINE LEARNING-2(ML)

Coded Project Report

Easy Visa Project

Submitted to

by

Subhadeep Seal

In Partial Fulfilment of PGP-DSBA

1|Page
 lOMoAR cPSD| 44635080

Contents

Sr No Topics Page
1 Easy Visa Project 3
1.1 Problem definition 3
1.2 Objective 3
1.3 Data Dictionary 3
1.4 Data Overview 4
1.5 Exploratory Data Analysis 5
1.6 Bivariate Analysis 9
1.7 Data Pre-processing 13
1.8 Model Performance Comparison and Final Model Selection 13
1.9 Actionable Insights and Recommendations 14
2 Donald Trump Tweet analysis 17
2.1 Objective 17
2.2 Data Dictionary 17
2.3 Data Overview 17
2.4 Word cloud 21
2.5 Conclusion 21

List of Tables

No Table Description Page no

1. Table 1: Top 5 rows of the Data set 4
2. Table 2: Basic information of Data Set 4
3. Table 3: Duplicate data summary of the Dataset 4
4. Table 4: Statistical summary of Categorical fields only 4
5. Tale 5: Statistical summary of the dataset 5
6. Table 6: Histogram and boxplot of Number of employees in the employer's company 5
7. Table 7: Histogram and boxplot of prevailing wage 6
8. Table 8: observations which have less than 100 prevailing wage 6
9. Table 9: Bar plot Information of continent the employee 7
10. Table 10: Education of employee 7
11. Table 11: Barplot of has_job_experience 7
12. Table 12: Bar plot of requires job training 8
13. Table 13: Bar plot of region_of_employment 8
14 Table 14: Bar plot of Unit_of_wage 8
15 Table 15: Bar plot of case_status 9
16. Table 16: Correlation matrix of variables 10
17. Table 17: education has any impact on visa certification 10
18. Table 18: Different regions have different requirements of talent having diverse educational backgrounds 10
19. 11
Table 19: percentage of visa certifications across each region
20 Table 20: the visa status varies across different continents 11
21 11
Table 21: having work experience has any influence over visa certification
22 Table 22: employees who have prior work experience require any job training 12

23 12
Table 23: see if the visa status changes with the prevailing wage
24 12
Table 24: Checking if the prevailing wage is similar across all the regions of the US
25 Table 25: Outliers in the data 13
26 Table 26: Training performance comparison 14
27 Table 27: Testing performance comparison 14
28 Table 28: Top 5 rows of the Data set 17
29 Table 29: bottom rows of Data Set 18
30 18
Table 30: Data types of the columns for the dataset
31 Table 31: Missing values data summary of the Dataset 19
32 19
Table 32: Number of characters in tweets
33 19
Table 33: Number of words in a tweet
34 Table 34: Most active hour on twitter 20
35 21
Table 36: Word cloud

2|Page
 lOMoAR cPSD| 44635080

Easy Visa Project

1.1 Problem Definition

Business Context

Business communities in the United States are facing high demand for human resources, but one of the constant challenges is identifying and
attracting the right talent, which is perhaps the most important element in remaining competitive. Companies in the United States look for
hard−working, talented, and qualified individuals both locally as well as abroad.

The Immigration and Nationality Act (INA) of the US permits foreign workers to come to the United States to work on either a temporary or
permanent basis. The act also protects US workers against adverse impacts on their wages or working conditions by ensuring US employers'
compliance with statutory requirements when they hire foreign workers to fill workforce shortages. The immigration programs are administered
by the Office of ForeignLabor Certification (OFLC).

OFLC processes job certification applications for employers seeking to bring foreign workers into the United States and grants certifications in
those cases where employers can demonstrate that there are not sufficient US workers available to perform the work at wages that meet or
exceed the wage paid for the occupation in the area of intended employment.

1.2 Objective

In FY 2016, the OFLC processed 775,979 employer applications for 1,699,957 positions for temporary and permanent labor certifications. This
was a nine percent increase in the overall number of processed applications from the previous year. The process of reviewing every case is
becoming a tedious task as the number of applicants is increasing every year.

The increasing number of applicants every year calls for a Machine Learning based solution that can help in shortlisting the candidates having
higher chances of VISA approval. OFLC has hired the firm EasyVisa for data−driven solutions. You as a data scientist at EasyVisa have to analyze the
data providedand, with the help of a classification model:

 Facilitate the process of visa approvals.

 Recommend a suitable profile for the applicants for whom the visa should be certified or denied based on the drivers that significantly
influencethe case status.

1.3 Data Dictionary

The data contains the different attributes of employee and the employer. The detailed data dictionary is given below.

 case_id: ID of each visa application

 continent: Information of continent the employee
 education_of_employee: Information of education of the employee
 has_job_experience: Does the employee has any job experience? Y= Yes; N = No
 requires_job_training: Does the employee require any job training? Y = Yes; N = No
 no_of_employees: Number of employees in the employer's company
 yr_of_estab: Year in which the employer's company was established
 region_of_employment: Information of foreign worker's intended region of employment in the US.
 prevailing_wage: Average wage paid to similarly employed workers in a specific occupation in the area of intended employment. The
purpose ofthe prevailing wage is to ensure that the foreign worker is not underpaid compared to other workers offering the same or
similar service in the same area of employment.
 unit_of_wage: Unit of prevailing wage. Values include Hourly, Weekly, Monthly, and Yearly.
 full_time_position: Is the position of work full−time? Y = Full Time Position; N = Part Time Position
 case_status: Flag indicating if the Visa was certified or denied

3|Page
 lOMoAR cPSD| 44635080

1.4 Data Overview

Structure of the data:

The data has 12 columns as mentioned in the data dictionary.

Table 1: Top 5 rows of the Data set

Table 2: Basic information of Data Set

4|Page
 lOMoAR cPSD| 44635080

Following are basic observations of data set after analyzing the data set:

1.The data frame has 25480 rows and 12 columns.

2.There are no null values in data set

3. There are no missing values in data set

Duplicate Values

Inspecting the duplicate values in customer key column

Table 3: Duplicate data summary of the Dataset

Observation:

1. There are no duplicates in the dataset.

Statistical summary of the dataset

Table 4: Statistical summary of Categorical fields only

Observations:

 The average number of employees in the employer's organization are 5667 while the median number of employees in the employer's
organization are 2109. This implies the attribute has a right skewed distribution with several positive outliers. The minimum number is
negativewhich does not appear to be a valid data point
 There are companies in the dataset with years of establishment from 1800 to 2016
 The average prevailing wage for occupation in United States is USD 74,455 while the median (~50th percentile of wages) is USD 70,308.
This indicates, slight right skewness in the data set. The minimum value of USD 2.1367 does not appear to be a valid data point. The
attribute has tobe studied in union with unit_of_wage to gather further insight
 The case ID attribute can be dropped as it is a unique ID variable and is not expected to add any value to the status of a visa being accepted
 There are 6 continents in the database, with majority of applicants from Asia
 There are 4 different levels of eduction with Bachelor's being the highest education degree for majority of applicants
 Majority of applicants do not require further job training to perform the intended occupation in the US
 There are 5 different regions in the US requiring immigrants due to Human Resource shortages, the maximum being in the NorthEast US
region
 There are 4 different units of wages with yearly being the most common. The prevailing wage and unit of wage may need to be studied in
unionto gather further insight
 Majority of the occupation with employee shortages are full time positions
 Case status is the attribute of interest (which needs to be predicted by our ML model). As per dataset, 66.7% of all applicants have a
certified visastatus and only 33.2% have a denied visa status

5|Page
 lOMoAR cPSD| 44635080

1.5 Exploratory Data Analysis

We have plotted histogram and boxplot all the data variables.

1. Number of employees in the employer's company:

Table 6: Histogram and boxplot of Number of employees in the employer's company

6|Page
 lOMoAR cPSD| 44635080

2. Prevailing wage: Average wage paid to similarly employed workers in a specific occupation in the area of intended employment. The purpose of
theprevailing wage is to ensure that the foreign worker is not underpaid compared to other workers offering the same or similar service in the
same area of employment.

Table 7: Histogram and boxplot of prevailing wage

We are checking the observations which have less than 100 prevailing wage

Table 8: observations which have less than 100 prevailing wage

Observations:

1.There are 176 observations which have less than prevailing wage

We will create bar plot of different variables

1. continent−Information of continent the employee:

7|Page
 lOMoAR cPSD| 44635080

Table 9: Bar plot Information of continent the employee

2. Education of employee: Information of education of the employee

8|Page
 lOMoAR cPSD| 44635080

Table 10: Education of employee

3.has_job_experience: Does the employee has any job experience? Y= Yes; N = No

Table 11: Barplot of has_job_experience

9|Page
 lOMoAR cPSD| 44635080

4. requires_job_training: Does the employee require any job training?

Table 12: Bar plot of requires_job_training

5. region_of_employment: Information of foreign worker's intended region of employment in the US

Table 13: Bar plot of region_of_employment

6. Unit_of_wage: Unit of prevailing wage. Values include Hourly, Weekly, Monthly, and Yearly.

Table 14: Bar plot of Unit_of_wage

10 | P a g
e
 lOMoAR cPSD| 44635080

7.case_status: Flag indicating if the Visa was certified or denied

Table 15: Bar plot of case_status

Observations from Exploratory Data Analysis

 More than twice the number of cases were certified than denied irrespective of the number of employees in the employer's
organization & the year of establishment of the employer's organization. These attributes are hence, not thought to have an impact on
case statuses

 Both these attributes are heavily skewed, the no_of_employees is skewed right but yr_of_estab is skewed left
 From the EDA, we infer 58% of all cases were for smaller organizations (<2500 employees) and 61% of all cases werefor
employer's established after 1990
 Only 35% of the cases were certified when the unit_of_wage is Hour−ly but 70% were certified when the unit_of_wage is notHour−ly
(i.e., Week−ly, Month−ly or Year−ly). This indicates unit_of_wage is an important attribute that can influence case statuses

 From the EDA, we infer only 8.5% of all cases were for unit_of_wage Hour−ly and the remaining 91.5% of all cases werefor
unit_of_wage not Hourly (i.e., Weekly, Monthly or Yearly)
 Majority of cases are from applicants in Asia (66%), then Europe (15%), N. America (13%) & S. America (3%); however, casesgetting
certified is highest for Europe (80% of such cases), then Africa (72% of such cases), then Asia (65% of such cases), & least for S.
America & N. America (around 60% of such cases). More cases are certified than denied irrespective of the continent. Being from
Europe is thought to be an important attribute to have an impact on case statuses

 Majority of applicants have a bachelor's (40%) or a master's degree (37.87%). A small number have only high school certification
(13.4%) or are very highly educated/ doctorate (8.6%). However, cases getting certified is highest for doctorate degree (>86%), followed
by master degree (>76%), then bachelor's (~62%). The cases getting certified is very low for those applicants with only a high school
certification (<35%). The trend observed is intuitive and one can expect attributes having adoctorate degrees & having only a high
school certification to significantly contribute to a case being certified and denied respectively

 From the EDA, we infer that 58% of all applicants have prior job experience and 42% do not. The cases getting certified is highfor
applicants with prior job experience (75% of such cases) and low for applicants without prior job experience (~56% of suchcases). This is
again an important attribute with an applicant having prior job experience significantly contributing to a case being certified

 Majority do not require the employee to receive any additional job training. This attribute was not found to have an impact onthe case
statuses

 Majority of the applications are to Northeast (28.3%), then South (27.5%), then West (25.8%), Midwest (16.9%) and least to Island
(1.5%) regions of the US. However, the cases certified follows the trend Midwest (75% of such cases), then South (70%of such cases),
then Northeast, West, & Island (60% of such cases). Region of employment being Midwest hence is an important attribute contributing
positively to a case being certified

 Majority of the jobs are full time rather than part time. This attribute was not found to have an impact on the case statuses

11 | P a g
e
 lOMoAR cPSD| 44635080

1.6 Bivariate Analysis

Correlation matrix of variables and Creating a pair plot with a kernel density estimate on the diagonal

Table 16: Correlation matrix of variables

Those with higher education may want to travel abroad for a well−paid job. Let's find out if education has any impact on visa certification

12 | P a g e
 lOMoAR cPSD| 44635080

Table 17: education has any impact on visa certification

Different regions have different requirements of talent having diverse educational backgrounds. Let's analyze it further

Table 18: Different regions have different requirements of talent having diverse educational
backgrounds

13 | P a g e
 lOMoAR cPSD| 44635080

Let's have a look at the percentage of visa certifications across each region

Table 19: percentage of visa certifications across each region

Lets' similarly check for the continents and find out how the visa status vary across different continents.

Table 20: the visa status varies across different continents

14 | P a g e
 lOMoAR cPSD| 44635080

Experienced professionals might look abroad for opportunities to improve their lifestyles and career development. Let's see if having work
experience has any influence over visa certification

Table 21: having work experience has any influence over visa certification

Do the employees who have prior work experience require any job training?

Table 22: employees who have prior work experience require any job training

15 | P a g e
 lOMoAR cPSD| 44635080

The US government has established a prevailing wage to protect local talent and foreign workers. Let's analyze the data and see if the visa status
changeswith the prevailing wage

Table 23: see if the visa status changes with the prevailing wage

Checking if the prevailing wage is similar across all the regions of the US

Table 24: Checking if the prevailing wage is similar across all the regions of the US

The prevailing wage has different units (Hourly, Weekly, etc.). Let's find out if it has any impact on visa applications getting certified.

16 | P a g e
 lOMoAR cPSD| 44635080

1.7 Data Pre−processing

Outlier Check

Let's check for outliers in the data.

Table 25: Outliers in the data

Data Preparation for modelling

 We want to predict which visa will be certified.

 Before we proceed to build a model, we'll have to encode categorical features.
 We'll split the data into train and test to be able to evaluate the model that we build on the train data.

Model evaluation criterion

Model can make wrong predictions as:

1. Model predicts that the visa application will get certified but in reality, the visa application should get denied.
2. Model predicts that the visa application will not get certified but in reality, the visa application should get certified.

Which case is more important?

 Both the cases are important as:
 If a visa is certified when it had to be denied a wrong employee will get the job position while US citizens will miss the opportunity to
work onthat position.

17 | P a g e
 lOMoAR cPSD| 44635080

 If a visa is denied when it had to be certified the U.S. will lose a suitable human resource that can contribute to the economy.

How to reduce the losses?

 F1 Score can be used the metric for evaluation of the model, greater the F1 score higher are the chances of minimizing False Negatives
and FalsePositives.
 We will use balanced class weights so that model focuses equally on both classes.

1.8 Model Performance Comparison and Final Model Selection

 Hyperparameter tuning has decreased the over fit and increased F1 score, however, this model is not performing as optimally
as thehyperparameter tuned decision tree
 Bagging classifier is also overfitting the training data
 Bagging − Hyperparameter Tuning is still found to over fit the training data, as the training metrics are high but the testing metrics are not
 Unlike the decision tree, random forest, or the bagging classifier; the AdaBoost classifier is not found to over fit the training data. It is
giving ageneralized performance on the training & testing data with a F1 score 0.819 & 0.816
 The hyperparameter tuned model is giving similar performance to the default AdaBoost model
 There is not much difference in the model performance after hyperparameter tuning
 Decision tree, Random forest (default & tuned) & Bagging classifier (default & tuned) were found to over fit the training dataset

18 | P a g e
 lOMoAR cPSD| 44635080

 Decision tree tuned and Adaboost (default & tuned) were found to give generalized performance on the training & testing data sets.

Table 26: Training performance comparison

Table 27: Testing performance comparison

 mind, and perhaps a revaluation of cases getting denied can be carried out in case there is a prevailing human resource shortage
in the US. Themodel is still helpful, as only a small subset of data will need further re−evaluation significantly decreases time spent
in the process

1.9 Actionable Insights and Recommendations

 Based on the EDA the following features were identified as important for visas getting certified than denied

 (1) Education of employee; an employee with only a high school certification has over 65% chance of visa getting denied in
comparison to an employee with a doctorate degree with over a 85% chance of visa getting certified
 (2) Unit of wage; an employee with an hourly pay likewise has over 65% chance of visa getting denied in comparison to an
employee with anon−hourly pay (weekly, monthly or yearly) with over 70% chance of visa getting certified
 (3) The continent the employee is from (e.g., if Europe, over 80% chance of visa getting certified), if the employee has prior job
experience (over 75% chance of visa getting approved if an employee has prior work experience but 50% chance of visa getting
denied if an employee has no work experience) are other important attributes
 (4) Likewise, the region of the US the employment opportunity is in is also an important deciding factor with over 70% cases
getting certified if the region is Midwest or South
 Interestingly, attributes like if the job opportunity is full time/ part time; if an employee requires further job training; the annual prevailing
wage of the occupation in the US; year of establishment of the employer or the number of employees in the organization are not
important attributes & do not have much bearing on a case getting certified vs denied

 The hyperparameter tuned ML model is able to give generalized prediction on training & testing datasets (not prone to overfitting) and is
able to explain over 80% of information (accuracy of 75% on test dataset & F1 score of 82% on test dataset).

 The precision & recall are likewise both high (77% & 88% respectively)
 The confusion matrix is able to identify a higher % of cases getting certified, but only a smaller % of cases getting denied
correctly. This limitation has to be borne in mind, and perhaps a revaluation of cases getting denied can be carried out in case
there is a prevailing human resource shortage in the US. The model is still helpful, as only a small subset of data will need further
re−evaluation significantly decreases time spent in the process

19 | P a g e
 lOMoAR cPSD| 44635080

20 | P a g e
 lOMoAR cPSD| 44635080

21 | P a g e
 lOMoAR cPSD| 44635080

22 | P a g e
 lOMoAR cPSD| 44635080

23 | P a g e
 lOMoAR cPSD| 44635080

24 | P a g e