Title: An Experimental Investigation of Mobile Network Traffic Data; Management and Prediction Accuracy
Speaker: Prof. Samuel Ajila, Systems and Computer Engineering
Date and Time: Friday April 7, 1:00-2:00 pm
Location: Room 4359 MacKenzie Building
Abstract: The growth in the number of mobile subscriptions has led to a substantial increase in the mobile network bandwidth demand. The mobile network operators need to provide enough resources to meet the huge network demand and provide a satisfactory level of Quality-of-Service (QoS) to their users. This talk presents an application of data analytics techniques to the processing, analysis, and resource prediction of a commercial mobile network traffic dataset. The processing and analysis provide an insight into the network traffic and the resource usage vis-à-vis the network base stations.
The approach in this research consists of multiple steps – identification of the relationship between features using correlation analysis, dimensionality reduction using Principal Component Analysis (PCA), application of clustering algorithms (K-Means and Fuzzy K-Means) for categorizing data records, data standardization, and finding the best Time Series algorithm for resource prediction. Prediction accuracy of three algorithms is considered: Multi-Layer Perceptron (MLP), Multi-Layer Perceptron with Weight Decay (MLPWD) and Support Vector Machine (SVM). MLP and MLPWD are variants of Artificial Neural Networks (ANN). The tools used in the research include Apache Hadoop, a software platform for analyzing and processing large datasets; Mahout, a machine learning library built on top of Hadoop; the R package, the WEKA tools; and the additive decomposition of Time Series using STL.
Notes:
- Bring your lunch if you like, there will be coffee and juices.
- Sponsored by the Real-time and Distributed Systems (RADS) Lab