MOHD AZRI BIN ABDUL AZIZ

Scopus Publications

Graduate Data Job Classification Using Support Vector Machine with Radial Basis Function Kernel
M. H. H. Hisham, M. A. Abdul Aziz, and A. A. Sulaiman
IEEE
The escalating rates of unemployment among recent graduates constitute a pressing concern, with farreaching implications for a nation’s future. Graduates often encounter challenges in aligning their skills and interests with suitable positions, while employers grapple with identifying the ideal candidates for their job openings. To address this issue, this study focuses on graduate-job classification using a Support Vector Machine (SVM) with a Radial Basis Function (RBF) kernel, based on graduates’ data. The SVM - RBF model’s performance was evaluated with a consistent C value of 10, while the Gamma value underwent variations (0.125, 0.25, and 0.75). In addition, a linear SVM was included for comparative analysis. Various metrics including classification accuracy, Root Mean Square Error (RMSE), and the receiver operating characteristic (ROC) curve were employed to ascertain the optimal classifier performance. The results indicate that the SVM - RBF model with a Gamma value of 0.125 demonstrated the most robust performance, surpassing SVM - RBF models with Gamma values of 0.25 and 0.75, as well as the linear SVM.


WiFi Approximated Strength Measurement Method with Brute Force Algorithm for a Minimum Number of AP and Maximum WiFi Coverage
Wan Nur Fatihah Wan Mustapha, Mohd Azri Abdul Aziz, Marianah Masrie, Rosidah Sam, and Mohd Nor Md. Tan
IEEE
The implementation of a wireless network in indoor premises has increased due to its easy and flexible access. This however requires a good strategy in placing the access point (AP) in order to cover as much area as possible with a small number of AP as possible. This paper proposed a WiFi approximated signal quality measurement method to be used with a Brute Force algorithm in looking for the best placement of AP in indoor locations. Only one time measurement of WiFi signal quality for each AP was done and our proposed algorithm will predict the strength of this AP as it was installed in a different location. The result shows the approximated signal quality generated by our algorithm almost equals the actual strength measured with an acceptable error. The new placement of AP proposed by our algorithm also manages to ensure a minimum of 84% WiFi strength in each room if all 4 APs were used. Experimental results have also shown a minimum of 2 APs is adequate to ensure at least 72% of WiFi signal quality can be received in each room.