Tesis

Optimization in machine learning and deep learning models for action prediction of concept map log data with hyperparameter tuning / F.ti Ayyu Sayyidul Laily

Abstrak
A concept map is a depiction of an individual s idea that connects two concepts and forms a proposition. Concept maps are recognized as a simple learning representation but are very influential in supporting teaching and learning activities. When learners interact with this concept map their behavior such as adding new vertices forming relationships between concepts and restructuring content results in unique interaction patterns that exist in the concept map creation activity data log. Log data has an important role in understanding how users interact with the system but currently presents many challenges due to the complexity and volume of the data. In recent years deep learning techniques have become fundamental in technology one of which is like LSTM BiLSTM and RNN which can capture the sequential nature of data improving system accuracy and performance. In this study action prediction was carried out with two datasets with a total of 15 194 rows of data using deep learning algorithms namely LSTM BiLSTM and RNN to measure model performance using evaluation results based on performance measures namely accuracy precision and recall and hyperparameter tuning was carried out using random search to determine the effect of tuning modifications on the model which is expected to improve accuracy. The research stage through data collection is an open-ended concept map of 2 different materials namely databases and SQL. The process begins with data collection then data preprocessing modeling and hyperparameter tuning for each model algorithm and finally evaluation. The research results show that the use of deep learning is proven to be more effective than conventional machine learning methods in capturing complex patterns in user interaction log data. Applied models such as LSTM BiLSTM and RNN are able to identify action patterns more accurately. The use of hyperparameter tuning with random search produces better effects compared to models without tuning for LSTM BiLSTM and RNN based on accuracy precision and recall.

Informasi Detail

DDC

Rt 621.3076 LAI o

Prodi

Universitas Negeri Malang. Program Studi Teknik Elektro, 2025.

Deskripsi Fisik

xiii, 85 lembar. : ilus. ; 30 cm.

Bahasa

No Reg

00704/RT/25

Edisi

Tesis (Pascasarjana)--Universitas Negeri Malang. 2025

Subjek

1. TEKNIK ELEKTRO - PETA KONSEP
2. MODEL MACHINE LEARNING DAN DEEP LEARNING
3. ELECTRICAL ENGINEERING - CONCEPT MAP

Pembimbing

1. Dr.eng Didik Dwi Prasetya, S.t., M.t.; 2. Dr.eng Anik Nur Handayani, S.t., M.t

Lampiran Berkas