ESSAY SCORING USING MACHINE LEARNING MODEL

• Main Author: Qotrunnada, Farah
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/55166
• Institution: Institut Teknologi Bandung

Essay scoring is an assessment method that is often used by teachers to evaluate students’ learning. However, in reality, the application of this method often takes the time of teachers who should spend more time involving students in the actual learning process. In addition, in assessing a large number of essays, the teacher who is a human can still experience fatigue and allow for inconsistency of assessment between students. Assessment of essay answers using a computer to increase efficiency can be an alternative to solving these problems. In this study, the assessment of essay answers using a computer was built using machine learning models of Linear Regression, Ridge Regression, and XGBoost. Essay answer data in the form of text is processed into numbers that are included in the syntax features and vector representation of the number of occurrences of gram words (Bag of Words). The features that can already be understood by the computer become input for model development. Evaluation is done to measure the performance of the model that has been built, namely using Quadratic Weighted Kappa (QWK). This evaluation measures how accurate the model is by considering different weights for different errors. In this study, the model has been successfully built with the best performance based on QWK is the XGBoost model using all features input with a result of 85%. Furthermore, based on computational efficiency, the Linear Regression and Ridge Regression models are better than the XGBoost model.