OPTIMASI PERFORMA MODEL KLASIFIKASI MENGGUNAKAN TEKNIK HYPERPARAMETER TUNING PADA ALGORITMA K-NEAREST NEIGHBORS, SUPPORT VECTOR MACHINE, DAN RANDOM FOREST

Izza Afkarina, NIM.: 20206052007 (2024) OPTIMASI PERFORMA MODEL KLASIFIKASI MENGGUNAKAN TEKNIK HYPERPARAMETER TUNING PADA ALGORITMA K-NEAREST NEIGHBORS, SUPPORT VECTOR MACHINE, DAN RANDOM FOREST. Masters thesis, UIN SUNAN KALIJAGA YOGYAKARTA.

Preview	Text (OPTIMASI PERFORMA MODEL KLASIFIKASI MENGGUNAKAN TEKNIK HYPERPARAMETER TUNING PADA ALGORITMA K-NEAREST NEIGHBORS, SUPPORT VECTOR MACHINE, DAN RANDOM FOREST) 20206052007_BAB-I_IV-atau-V_DAFTAR-PUSTAKA.pdf - Published Version Download (2MB) | Preview
	Text (OPTIMASI PERFORMA MODEL KLASIFIKASI MENGGUNAKAN TEKNIK HYPERPARAMETER TUNING PADA ALGORITMA K-NEAREST NEIGHBORS, SUPPORT VECTOR MACHINE, DAN RANDOM FOREST) 20206052007_BAB-II_sampai_SEBELUM-BAB-TERAKHIR.pdf - Published Version Restricted to Registered users only Download (4MB) | Request a copy

Abstract

Classification is the process of identifying models or patterns capable of depicting and segregating classes within a dataset. Its aim is to enable models to predict objects with unknown class labels. Classification can address various issues including data classification, image classification, text categorization, protein structure predictions, and more. Several methods can be utilized to tackle classification problems, such as K-NN, SVM, and random forest. Evaluating the performance of a classification model is done through multiple tests like accuracy, precision, recall, and f-measure. Enhancing the performance of a classification model can be achieved by adjusting hyperparameters. However, different classification algorithms come with distinct sets of hyperparameters. Thus, determining optimal hyperparameters through techniques like hyperparameter tuning is necessary to obtain parameters that maximize model performance. The objective of this research is to apply grid search and random search hyperparameter tuning techniques to several classification models capable of handling high-dimensional data, namely K-Nearest Neighbors (K-NN), Support Vector Machine (SVM), and random forest. Grid search and random search are chosen for their simplicity in implementation compared to more complex methods, yet they maintain good performance when seeking optimal hyperparameters. The hyperparameter tuning techniques applied here use CrossValidation, ensuring that good performance in model training is also reflected in the testing process. The dataset used in this research is sourced from the UCI Machine Learning repository titled "Predict Students' Dropout and Academic Success." It comprises 4424 records with 36 features and 1 target variable. Multiple classification tests were conducted on the algorithms used by applying grid search and random search hyperparameter tuning techniques. Grid search and random search successfully improved the accuracy of the K-NN and random forest algorithms compared to unoptimized models. However, there was no improvement in SVM model accuracy despite using optimal hyperparameters from tuning.

Share this knowledge with your friends :

Actions (login required)

View Item