phyosandarwin / sc1015-ntu-project

Job Placement Status Mini-project for SC1015- Introduction to Data Science and Artificial Intelligence

SC1015 Project: JobMatch

About

This is a Mini-Project for SC1015 (Introduction to Data Science and Artificial Intelligence) which focuses on job placement statuses in developing countries from Job Placement Status Kaggle Dataset.

For detailed walkthrough, please view the source code in order from:

1. Data Preparation and Cleaning
2. Exploratory Data Analysis (EDA)
   • Numeric
   • Categorical
3. Machine Learning Models

Presentation Slides

Video

Problem Statement

Which variables are the most important in predicting whether one gets hired or not?

Problem Significance

Growing need for educated and talented individuals, especially in developing countries, recruiting fresh graduates is a routine practice for organizations. As undergraduate students, we felt that this issue was relevant to us. As individuals who are not so familiar with trends in developing countries, we decided to investigate the hiring practices of fresh graduates there.

Models used

• Decision Tree Classifier
• Logistic Regression
• Support Vector Machine (SVM)

Conclusion/Data-driven insights

• Logistic Regression performs the best out of the three models, especially in terms of F1 score (good balance between precision and recall)

  features selected are SSC%, HSC%, Degree %, MBA %, and the lack of work experience

• Unexpectedly in developing countries, hiring practices have an increasing focus on work experience (although academic qualifications remain to be highly desirable) and a decreasing focus on gender for job roles.
• Our recommendations to students from LDC are:

1. Do well academically across all levels
2. Gain some work experience before entering the workforce, e.g. internships

What did we learn from the project?

• Use of chi-square statistic for correlation analysis of categorical varibles
• Use of F1 score, precision, recall, ROC-AUC score as model performance evaluation metrics
• Use of RFE (Recursive Feature Elimination) as a feature selection technique
• Logistic regression and SVM from sklearn
• Collaborating on GitHub

References

• Chi-square Test
• Logistic Regression Code Explanation
• SVM
• One Hot Encoding
• EDA on categorical variables

Contributors

• @phyosandarwin: EDA (Numeric), Logistic Regression, Chi-Square Statistic in EDA (Categoric), Data-driven insights
• @cheyenneseet: Data Preparation and Cleaning, Decision Tree Classifier, Data-driven insights, Video Presentation
• @senchiagladine: EDA (Categoric), Support Vector Machine, Data-driven insights, Video Presentation