Title:
Enhanced Stacked Sparse Autoencoder Using Hybrid Loss Function For High-Dimensional Sparse Data Prediction
Author:
Supervisor:
Department:
Applied Science
Year:
2026
URL:
Abstract:
High-dimensional sparse data frequently arises in domains such as healthcare, finance, and cybersecurity, posing substantial challenges for reliable predictive modeling. Although Stacked Sparse Autoen...
View more
Multimedia Content:
01
Methodology
1. Dataset Selection and Benchmark Design
The researcher selected three benchmark datasets representing different sparsity levels and application domains:
• Health Indicators Dataset (43% sparsit...
View more
02
Citation Analysis
1. Dominant Research Themes
The literature review focuses heavily on:
• High-Dimensional Sparse Data
• Deep Learning
• Autoencoders
• Feature Extraction
• Representation Learning
• Spar...
View more
03
Impact
1. Improved Sparse Data Classification
The proposed HSSAE significantly improves predictive performance on high-dimensional sparse datasets by learning more informative latent representations.This ...
View more
04
Benefits to Society
1. Improved Healthcare Analytics
The framework demonstrates strong performance on diabetes and arrhythmia datasets. It can potentially support clinical decision-making through earlier disease predi...
View more
05
Applications
1. Medical Diagnosis Systems
The framework can be applied to healthcare datasets involving diabetes risk prediction, cardiac arrhythmia diagnosis, and other clinical prediction tasks.
Benefit:
...
View more
06
Infographic
Summary of this Thesis
07
Patents / Products
1. Hybrid Stacked Sparse Autoencoder (HSSAE)
A novel deep learning architecture developed specifically for sparse high-dimensional data classification.
Significance:
Eliminates the need for ext...
View more
08
Experts
1. Assoc. Prof. Dr. Hanita Binti Daud
Institution: Universiti Teknologi PETRONAS (UTP).
Contribution: Principal supervisor leading the development of the HSSAE framework.
Key Work: Deep learning ...
View more