PREDICTION OF ITB ALUMNI’S SALARY WITH MULTIPLE LINEAR REGRESSION MODELS USING RIDGE AND LASSO REGULARIZATION

• Main Author: Valentino, Andreas
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/77613
• Institution: Institut Teknologi Bandung

Bandung Institute of Technology (ITB), as one of the top universities in Indonesia, continually innovates to maintain the quality of its graduates. One of these innovations is the Tracer Study survey, which is later summarized in a report book containing descriptive statistics and data analysis. The survey results can be processed using predictive models through machine learning. This final project aims to predict the salaries of ITB alumni using multiple linear regression models. The data used consists of 5034 rows and 13 columns, and a preprocessing process will be conducted first to select variables and observations. Subsequently, the model will be subjected to Ridge and LASSO regularization to enhance the accuracy of the multiple linear regression model. The results of Ridge and LASSO regularization will be compared to determine which regularization method is better for predicting the salaries of ITB alumni. Based on the experimental results, Ridge regularization performs better for the predictive model with an adjusted R-squared value of 0.4346, which is higher than LASSO's value of 0.4245. The coefficient values of each predictor in the prediction model and the adjusted R-squared values between the two types of regularization have an average difference of one decimal place.