A combination of methodology building for multi-layer feed-forward neural network (MLFF) and linear modeling: an application in biometry modeling

• Main Author: Jusoff, Muhammad Khairan Shazuan
• Format: Thesis
• Language: English
• Published: 2024
• Subjects: R Medicine; RA440-440.87 Study and teaching. Research
• Online Access: http://eprints.usm.my/60710/1/MUHAMMAD%20KHAIRAN%20SHAZUAN%20BIN%20JUS-FINAL%20THESIS%20P-SGM001221%28R%29-E.pdf; http://eprints.usm.my/60710/

Biostatistics, also known as biometry, is a field of statistics that focuses on the application of statistical methods to the field of biomedicine and health sciences. Biostatistics can assist researchers and healthcare professionals in identifying risk factors, evaluating intervention effectiveness and many more. However, biostatistics has not been entirely embraced by medical professionals due to several reasons. One of the main reasons is that the medical field is challenging, and maintaining a high level of accuracy is critical. In addition, many previous studies focused on individual modeling technique that has limited ability to capture the dynamic and complexities in the medical field. This study aims to develop a biometry model that combines several statistical techniques, namely bootstrap, Multi-Layer Feed-Forward Neural Network (MLFF) and Multiple Linear Regression (MLR). This study will propose two distinct models: (i) Hybrid MLFF-MLR model with case resampling and (ii) Hybrid MLFF-MLR model without case resampling. The two models will be compared using the Mean Square Error of Neural Network (MSE.net) and the Mean Square Error of the Linear Model (MSE.lm). The model with lower MSE.net and MSE.lm values will be deemed superior. The analysis results from both models show that the hybrid MLFF-MLR model with case resampling yields a more accurate output. This research contributes to the body of knowledge by exploring the potential of biometry modeling and can be a reference for future researchers in the same field.