Dr. Muhammad Salman Pathan has completed his PhD from Beijing University of Technology (BJUT), China in 2019. He was awarded by 4 years fully funded Chinese Scholarship Council (CSC) Scholarship to complete his PhD studies. His PhD research focuses on malicious attack pattern detection and its prevention in ad-hoc networks. He has applied different machine learning techniques in diverse technologies like smart homes, cloud computing, mobile ad-hoc networks in order to observe the behavior of the attacker and its detection efficiently. He also has published his work in some reputable research journals and conferences.

Since September 2019 till January 2021, he has worked as a Postdoctoral fellow at BJUT to study the event onset prediction using time series data. Analysis of time series data is a kind of analytical technique that is used to study different patterns of data so that the actual root cause behind the occurrence of any event can be understood in order to predict the onset. Most of these future events are critical and detrimental for the specific applications. Successful detection of the onset of these critical events could save a lot of effort and resources.

In addition to generous research funding, ELITE-S offers training and development programmes for scientific excellence and career development opportunities. ELITE-S program supports the recruited fellows by providing an opportunity to work in companies along with academic research study at a host institute that enhances employability and career development. Such training programs will provide an opportunity to work and hone my scientific skills collaboratively with researchers and technology experts. Through this mentored research training, I will acquire the professional skills needed to pursue the desired career path, while contributing to innovations in translational research.

Project Summary

Analyzing Onset of Critical Events Using a Machine-Learning Framework

With the advancement of computing resources and the availability of plethora of data, machine-learning techniques are now extensively used for various applications in Internet of Things (IoT) devices. Such techniques assist in the user’s decision progress by predicting future events, based on historical data. Most of these future events are critical and detrimental for the specific applications.

Therefore, it is important for us to accurately predict the onset of these critical events. Successful detection of the onset of these critical events could save a lot of effort and resources. In this fellowship, we propose to develop machine learning based models that can perform accurate and timely predictions of critical events using historical time series data. We focus our work on three diverse applications — atmospheric study, optical communications and medical analytics.

Furthermore, our work will not only be used for predictions, but also useful in the identification of the root cause that caused the onset of the critical event. This will help in reducing the number of false predictions. Our proposed machine learning framework will be a strong candidate for analyzing the enormous time-series data generated from various IoT devices, which in turn can save a lot of cost related to extra hardware usage. We will ensure to propose computationally minimalistic models, in order to reduce extra computational overhead in the IoT device, and also ensuring a high Quality of Service (QoS) score for the application.