ICACT–2021

Permanent URI for this collectionhttp://repository.kln.ac.lk/handle/123456789/24483

Browse

Author: search.filters.author.Suddul, GeerishHas files: Yes

Search Results

Now showing 1 - 2 of 2
  • No Thumbnail Available
    Item
    Road Accident Severity Prediction in Mauritius using Supervised Machine Learning Algorithms
    (Faculty of Computing and Technology (FCT), University of Kelaniya, Sri Lanka, 2021) Sowdagur, Jameel Ahmad; Rozbully-Sowdagur, B. Tawheeda B; Suddul, Geerish
    Road accidents with high severities are a major concern worldwide, imposing serious problems to the socio-economic development. Several techniques exist to analyse road traffic accidents to improve road safety performance. Machine learning and data mining which are novel approaches are proposed in this study to predict accident severity. Support Vector Machine (SVM), Gradient Boosting (GB), Logistic Regression (LR), Random Forest (RF) and Naïve Bayes (NB) were applied to perform effective data analysis for informed decisions using Python programming language. The gradient boosting outperformed all the other models in predicting the severity outcomes, yielding an overall accuracy of 83.2% and an AUC of 83.9%
  • No Thumbnail Available
    Item
    A Deep Neural Network Approach for Analysis of Firewall Log Data
    (Faculty of Computing and Technology (FCT), University of Kelaniya, Sri Lanka, 2021) Lillmond, Chandesh; Suddul, Geerish
    In this paper, we propose an intelligent approach for the classification of incoming and outgoing firewall traffic packets. A firewall is a quintessential tool that ensures the control of traffic over machines’ communication over a network. It uses a set of specific rules to define the traffic and thus assists in avoiding cyber-attacks which can be very costly to an organization. Our intelligent approach is mainly through the application of the Deep Neural Network (DNN) Machine Learning algorithm so that packets going through the firewall can be automatically classified as either allow, deny or drop. Our experiments demonstrate a classification accuracy of around 94%, which is higher when compared with other approaches.