Development of a new hybrid model and its application in biostatistics
Linear regression, a fundamental tool in statistical analysis, enables the exploration of relationships between variables. Despite its widespread use, traditional regression analysis encounters challenges when handling qualitative predictive variables (QPV), Multilayer Layer Feedforward Neural Netwo...
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2024
|
| Subjects: | |
| Online Access: | http://eprints.usm.my/61556/ http://eprints.usm.my/61556/1/FARAZ%20AHMED%20FAROOQI-TESIS%20P-SGD000619-E.pdf |
| _version_ | 1848884743326662656 |
|---|---|
| author | Farooqi, Faraz Ahmed |
| author_facet | Farooqi, Faraz Ahmed |
| author_sort | Farooqi, Faraz Ahmed |
| building | USM Institutional Repository |
| collection | Online Access |
| description | Linear regression, a fundamental tool in statistical analysis, enables the exploration of relationships between variables. Despite its widespread use, traditional regression analysis encounters challenges when handling qualitative predictive variables (QPV), Multilayer Layer Feedforward Neural Network (MLFFNN), and Fuzzy Linear Regression. There is a significant gap in understanding how to integrate multiple linear regression with other approaches to enhance model accuracy and predictability. This highlights the need for the development of hybrid models. Integrating Multiple Linear Regression (MLR) with advanced techniques, such as fuzzy regression and neural networks, addresses MLR's limitations in handling complex data and improves model accuracy and generalizability. This hybrid approach is crucial for overcoming challenges in biostatistics and enhancing predictive performance. This study utilizes a comprehensive methodology that integrates several techniques, such as transforming QPV, bootstrapping, MLFFNN, and employing fuzzy regression. The utility of the developed methodology is demonstrated using three secondary datasets. All obtained results demonstrate statistical significance, with high accuracy reflected in the R2 values. Additionally, small mean squared errors confirm a close alignment between predicted and actual values. All cases show the method's superiority, offering researchers precise tools for biostatistical inferences and forecasts. Future work will adapt this approach for other regression types and explore its application across various domains. |
| first_indexed | 2025-11-15T19:11:33Z |
| format | Thesis |
| id | usm-61556 |
| institution | Universiti Sains Malaysia |
| institution_category | Local University |
| language | English |
| last_indexed | 2025-11-15T19:11:33Z |
| publishDate | 2024 |
| recordtype | eprints |
| repository_type | Digital Repository |
| spelling | usm-615562025-02-11T02:37:41Z http://eprints.usm.my/61556/ Development of a new hybrid model and its application in biostatistics Farooqi, Faraz Ahmed R Medicine RA440-440.87 Study and teaching. Research Linear regression, a fundamental tool in statistical analysis, enables the exploration of relationships between variables. Despite its widespread use, traditional regression analysis encounters challenges when handling qualitative predictive variables (QPV), Multilayer Layer Feedforward Neural Network (MLFFNN), and Fuzzy Linear Regression. There is a significant gap in understanding how to integrate multiple linear regression with other approaches to enhance model accuracy and predictability. This highlights the need for the development of hybrid models. Integrating Multiple Linear Regression (MLR) with advanced techniques, such as fuzzy regression and neural networks, addresses MLR's limitations in handling complex data and improves model accuracy and generalizability. This hybrid approach is crucial for overcoming challenges in biostatistics and enhancing predictive performance. This study utilizes a comprehensive methodology that integrates several techniques, such as transforming QPV, bootstrapping, MLFFNN, and employing fuzzy regression. The utility of the developed methodology is demonstrated using three secondary datasets. All obtained results demonstrate statistical significance, with high accuracy reflected in the R2 values. Additionally, small mean squared errors confirm a close alignment between predicted and actual values. All cases show the method's superiority, offering researchers precise tools for biostatistical inferences and forecasts. Future work will adapt this approach for other regression types and explore its application across various domains. 2024-09 Thesis NonPeerReviewed application/pdf en http://eprints.usm.my/61556/1/FARAZ%20AHMED%20FAROOQI-TESIS%20P-SGD000619-E.pdf Farooqi, Faraz Ahmed (2024) Development of a new hybrid model and its application in biostatistics. PhD thesis, Universiti Sains Malaysia. |
| spellingShingle | R Medicine RA440-440.87 Study and teaching. Research Farooqi, Faraz Ahmed Development of a new hybrid model and its application in biostatistics |
| title | Development of a new hybrid model and its application in biostatistics |
| title_full | Development of a new hybrid model and its application in biostatistics |
| title_fullStr | Development of a new hybrid model and its application in biostatistics |
| title_full_unstemmed | Development of a new hybrid model and its application in biostatistics |
| title_short | Development of a new hybrid model and its application in biostatistics |
| title_sort | development of a new hybrid model and its application in biostatistics |
| topic | R Medicine RA440-440.87 Study and teaching. Research |
| url | http://eprints.usm.my/61556/ http://eprints.usm.my/61556/1/FARAZ%20AHMED%20FAROOQI-TESIS%20P-SGD000619-E.pdf |