Impact of Holy Quran Recitation on Psychological Wellbeing Among Muslim Youth

The present study is aimed to examine the relation of Holy Qur’ān recitation and psychological wellbeing among the Muslim Youth. People find no time for offering prayers and reciting Holy Qur’ān because they are very busy in their daily routines and if they do so, they do it for short period of time when they are in some trouble. The objectives include firstly the exploration of the relationship of Holy Qur’ān recitation and psychological wellbeing. Second objective of the study is to explore the correlation of the Holy Qur’ān recitation with depression, anxiety and stress among young Muslims. Study sample comprises of 100 young Muslims (43 males and 57 females) falling between 17 and 25 years from Rawalpindi and Islamabad. Instruments used for data collection include psychological wellbeing scale by Kamman and Flett (1983) and DASS (depression anxiety stress scale) by Lovibond and Lovibond (1995) and a demographic sheet. Results of the study showed that significant positive relationship exists between Holy Qur’ān recitation and psychological wellbeing among young Muslims and Holy Qur’ān recitation negatively relates with depression, anxiety and stress. The present study findings support that those young Muslims who had more rate of Holy Qur’ān recitation were psychologically more stable as compared to non-frequent reciters. In the light of findings of current study, it can be declared as a quintessence that Holy Qur’ān Recitation can serve as an influential element in ensuring the positive mental health of youth. Educators and Policy makers can play a crucial role in promotion of Holy Qur’ān familiarity which will make certain the psychological and mental health of youth and of the society at large.

Development of Feature Selection Algorithms for High-Dimensional Binary Data

There has been a growing interest in representing real-life applications with data sets having binary-valued features. These data sets due to the advancements in computer and data management systems consist of tens or hundreds of thousands of features. In this dissertation, we investigate two problems in machine learning which have been relatively less studied for high-dimensional binary data. The first problem is to select a subset of features useful for supervised learning applications from the entire feature set and is known as the feature selection (FS) problem. The second problem is to compare two orderings of features induced by feature ranking (FR) algorithms and to determine which one is better. For the feature selection problem, we have proposed a new feature ranking measure termed as the diff-criterion. Its distinct attribute is that it estimates the usefulness of binary features by using their probability distributions. The diff-criterion has been evaluated against two well-known FS algorithms with four widely used clas- sifiers on six binary data sets on which it has achieved up to about 99% reduction in the feature set size. To further improve the performance, we have suggested a two-stage FS algorithm. The novelty of our two-stage algorithm is that the first stage provides the second stage with a reduced subset without losing valuable in- formation about the class. Two-stage feature selection used with the diff-criterion not only significantly improves the classification accuracy but also exhibits up to about 99% reduction in the feature set size. We have also compared our proposed FS algorithms against the winning entries of the “Agnostic Learning versus Prior Knowledge” challenge. The algorithms have shown results better or comparable to the winners of the challenge. For the problem of ranking features using FR algorithms, different FR algorithms estimate the importance of features with respect to the class variable differently thus generating different orderings. To determine which ordering is better, we propose a new evaluation method termed as feature ranking evaluation strategy (FRES). It uses the individual predictive power of features for estimating howAbstract correct is an ordering of features. We found that compared to Relief and mu- tual information algorithms our proposed diff-criterion generates the most correct orderings of binary features.