Dr. Shaker A Ali - Academia.edu (original) (raw)

Papers by Dr. Shaker A Ali

Educational Administration: Theory and Practice, 2024

The telecommunication industry has grown exponentially to accommodate such technologies as the In... more The telecommunication industry has grown exponentially to accommodate such technologies as the Internet of Things, 5G, and cloud computing. This intricacy necessitates advanced fault management solutions, particularly for anomaly detection. While conventional methods, such as threshold-based alarms, remain reactive and insufficient, there is an increasing thrust on proactive solutions capable of preemptively identifying anomalies. The current research investigates the use of big data analytics and machine learning (ML) as a potential game-changer in realizing proactive fault management in telecommunication networks. It employs disparate ML paradigms – supervised, unsupervised, and semi-supervised – to develop a Python-based model called AUTO-PRO. This novel approach aims to significantly enhance the capability to hastily detect and mitigate anomalies, thereby elevating the reliability, security, and performance of contemporary telecommunication infrastructures.
This research addresses distinct yet interrelated objectives. First, it identifies the optimal methods for leveraging big data in detecting anomalies. Second, it evaluates the efficacy of various ML algorithms, considering the multifaceted nature of network faults. The third objective revolves around streamlining ML models by focusing on feature selection and dimensionality reduction to augment their efficiency and precision. Finally, the research seeks to identify and surmount the challenges of actualizing big data and ML-based solutions in tangible telecommunication environments.
The study employs a six-step research methodology. Initially, the telecommunication network data, specifically from the E1/T1 port alarm indication signal, undergoes data collection and robust preprocessing, including data cleansing, normalization, and transformation. Feature engineering accentuates the process of optimizing the data for ML applications. Subsequent stages encompass feature selection and dimensionality reduction, using principal component analysis to identify the most salient features and reduce computational demands. ML models, notably random forest and support vector machines, are then developed and thoroughly evaluated using precision, recall, and F1-score metrics. Lastly, a comparative analysis juxtaposes these models against conventional methods to ascertain their efficacy in detecting anomalies.
The findings underscore the superior performance of the isolation forest model in anomaly detection within telecommunication networks. The technique demonstrated superior performance in anomaly detection, achieving a precision of 0.863 and a recall of 0.943, overshadowing the commendable precision of the One-Class SVM model at 0.832. Notably, while the One-Class SVM exhibited an impeccable recall of 1.0, this study accentuates the importance of multi-metric evaluation. It highlights the potential pitfalls of over-reliance on a single metric. Moreover, the integration of SMOTEENN effectively addressed class imbalances, enriching the dataset quality and ensuring a more representative learning environment. Conclusively, the research reaffirms the potency of big data and ML in enhancing anomaly detection, as it introduces a new and optimized approach. Therefore, it significantly augments the extant body of knowledge, provides strategies to counter real-world challenges, and sets new benchmarks in telecommunication network anomaly detection.

Educational Administration: Theory and Practice, 2024

The telecommunication industry has grown exponentially to accommodate such technologies as the In... more The telecommunication industry has grown exponentially to accommodate such technologies as the Internet of Things, 5G, and cloud computing. This intricacy necessitates advanced fault management solutions, particularly for anomaly detection. While conventional methods, such as threshold-based alarms, remain reactive and insufficient, there is an increasing thrust on proactive solutions capable of preemptively identifying anomalies. The current research investigates the use of big data analytics and machine learning (ML) as a potential game-changer in realizing proactive fault management in telecommunication networks. It employs disparate ML paradigms – supervised, unsupervised, and semi-supervised – to develop a Python-based model called AUTO-PRO. This novel approach aims to significantly enhance the capability to hastily detect and mitigate anomalies, thereby elevating the reliability, security, and performance of contemporary telecommunication infrastructures.
This research addresses distinct yet interrelated objectives. First, it identifies the optimal methods for leveraging big data in detecting anomalies. Second, it evaluates the efficacy of various ML algorithms, considering the multifaceted nature of network faults. The third objective revolves around streamlining ML models by focusing on feature selection and dimensionality reduction to augment their efficiency and precision. Finally, the research seeks to identify and surmount the challenges of actualizing big data and ML-based solutions in tangible telecommunication environments.
The study employs a six-step research methodology. Initially, the telecommunication network data, specifically from the E1/T1 port alarm indication signal, undergoes data collection and robust preprocessing, including data cleansing, normalization, and transformation. Feature engineering accentuates the process of optimizing the data for ML applications. Subsequent stages encompass feature selection and dimensionality reduction, using principal component analysis to identify the most salient features and reduce computational demands. ML models, notably random forest and support vector machines, are then developed and thoroughly evaluated using precision, recall, and F1-score metrics. Lastly, a comparative analysis juxtaposes these models against conventional methods to ascertain their efficacy in detecting anomalies.
The findings underscore the superior performance of the isolation forest model in anomaly detection within telecommunication networks. The technique demonstrated superior performance in anomaly detection, achieving a precision of 0.863 and a recall of 0.943, overshadowing the commendable precision of the One-Class SVM model at 0.832. Notably, while the One-Class SVM exhibited an impeccable recall of 1.0, this study accentuates the importance of multi-metric evaluation. It highlights the potential pitfalls of over-reliance on a single metric. Moreover, the integration of SMOTEENN effectively addressed class imbalances, enriching the dataset quality and ensuring a more representative learning environment. Conclusively, the research reaffirms the potency of big data and ML in enhancing anomaly detection, as it introduces a new and optimized approach. Therefore, it significantly augments the extant body of knowledge, provides strategies to counter real-world challenges, and sets new benchmarks in telecommunication network anomaly detection.