Identifying Features Eligiblity For Blood Donors' Preferences Using Artificial Neural Networks Prediction Performances

Che Khalid, Nor Syuhada (2017) Identifying Features Eligiblity For Blood Donors' Preferences Using Artificial Neural Networks Prediction Performances. Masters thesis, Universiti Teknikal Malaysia Melaka.

[img] Text (24 Pages)
Identifying Features Eligiblity For Blood Donors' Preferences Using Artificial Neural Networks Prediction Performances - Nor Syuhada Che Khalid - 24 Pages.pdf

Download (558kB)

Abstract

Blood donation is an activity that has required people to contribute blood to help others during need for critical and near with fatal conditions such as organ transplant, post-partum haemorrhage, thalassemia, bowel operation, and orthopaedic surgery. Blood supplies extremely needed without fail. Therefore, blood donation service should retrieved useful information, especially to attract specific target groups of donors. However, this information required must be up-to-date, prepared systematically as prediction components, needed to scale down based on specifics target groups to make information extraction become better, and prediction algorithm has to adaptable with several sample sizes and features types because data may origin from different sources will have variety of datasets. Furthermore, blood donors’ preferences are based on human opinions, which could cause different priority, conditions, and data retrieval method based on different communities, organisation, or places. As a solution, these research focuses are to collect new data on blood donors’ preferences, construct Features Arrangement (FA) as dataset preparation for prediction model and criteria to distinguish between leading features (LF), features, and main leading features as main targets, and apply prediction algorithm which is artificial neural network. There is main dataset has collected from survey questionnaires. Features Arrangement has applied Pearson correlation between potential leading features and features as measurement to main leading features’ criteria. This study has found out about main leading features which have influenced directly by less number of positive significant relationships with their attributes or features that have known as member features (MF). Therefore, decreasing number of positive significant relationships, regardless numbers of significant relationships, have yield better performance of blood donors’ preferences predictor on main leading features as priority groups of respondents. FA has been implemented to select most and least associated features sets, from LFs and MFs. As summary, main blood donors’ preference in Malaysia at 2015 is gender; meanwhile least preference is donation fear. Another recommended main preferences besides than gender as additional information are donating as religious purpose, donated more than once per year experience, health self-awareness and save another people, longer donation experience, overcome donation fear, high overall donation volume, tend to donate for family or acquaintances, donate frequently, up to date donation, information announcement medium such as social media, donation experience, and favourite donation center. Another least preferences recommended by FA are donation fear, favourite donation center, marriage status, up to date donation, interested to overcome donation fear, high overall donation volume, and donation motivation by celebrities. These findings of this study contribute as beneficial information to improve blood donation or healthcare service, as guide to collect and arrange data into prediction or another data mining problems, and extend another study for flexible algorithms with various datasets.

Item Type: Thesis (Masters)
Uncontrolled Keywords: Artificial intelligence, Neural networks (Computer science), Artificial neural networks, Blood donors
Divisions: Library > Tesis > FTMK
Depositing User: Nor Aini Md. Jali
Date Deposited: 24 Aug 2020 09:06
Last Modified: 02 Jul 2021 11:29
URI: http://eprints.utem.edu.my/id/eprint/23162
Statistic Details: View Download Statistic

Actions (login required)

View Item View Item