Browsing by Author "Krishnadeva, K"

Now showing 1 - 2 of 2

Embargo
Intruder detection using deep learning and association rule mining
(IEEE, 2016-12-08) Thilina, A; Attanayake, S; Samarakoon, S; Nawodya, D; Rupasinghe, L; Pathirage, N; Edirisinghe, T; Krishnadeva, K
With the upsurge of internet popularity, nowadays there are millions of online transactions that are being processed per minute thus increasing the possibilities of intruder attacks over the recent times. There have been various intruder detection techniques such as using traditional machine learning based algorithms. These algorithms were widely used to identify and prevent intruder activities in the recent past. Furthermore, multilayer neural networks[5] were also used in this regard to perform the detection. Hence multi-layer neural networks inherit fundamental drawbacks due to its inability to perform training due the problems such as overfitting, etc. In contrast, deep learning algorithms were introduced to overcome these issues effectively. We propose a novel framework to perform intruder detection and analysis using deep learning nets and association rule mining. We utilize a recurrent network to predict intruder activities and FP-Growth to perform the analysis. Our results show the effectiveness of our framework in detail.
Embargo
Supervised learning based approach to aspect based sentiment analysis
(IEEE, 2016-12-08) Pannala, N. U; Nawarathna, C. P; Jayakody, J. T. K; Rupasinghe, L; Krishnadeva, K
Aspect base sentiment analysis is a very popular concept in the machine learning era which is under the research domain still at the movement. This research mainly consist of the way of exploring the sentiment analysis based on the trained data set to provide the positive, negative and neutral reviews for different products in the marketing world. Most of the existing approaches for opinion mining are based on word level analysis of texts and are able to detect only explicitly expressed opinions. In aspect-based sentiment analysis (ABSA) the aim is to identify the aspects of entities and the sentiment expressed for each aspect. The ultimate goal is to be able to generate summaries listing all the aspects and their overall polarity. For this research mainly natural language and machine learning techniques are used. To train the application for the given data sets SVM (support vector machine) and ME (Maximum Entropy) classification algorithms have been used. Differentiation of the performance of the each algorithm will be analyzed through this research using the proven technologies available in the world like "Re call", "F-Measure" and Accuracy.