Browsing by Author "Rafiq, Fatama Binta"

Filter results by typing the first few letters
Now showing 1 - 4 of 4
  • No Thumbnail Available
    Item
    Assessing the Effectiveness of Topic Modeling Algorithms in Discovering Generic Label with Description
    (Springer, 2020-02-13) Rahman, Shadikur; Hossain, Syeda Sumbul; Arman, Md. Shohel; Rawshan, Lamisha; Toma, Tapushe Rabaya; Rafiq, Fatama Binta; Md. Badruzzaman, Khalid Been
    Analyzing short text or documents using topic modeling becomes a popular solutions for the increasing number of documents produced in everyday life. For handling the large amount of documents, many topic modeling algorithms are used e.g. LDA, LSI, pLSI, NMF. In this study, we have used LDA, LSI, NMF and also lexical database wordNet synset for candidate labels in our topics labeling. And finally compare the effectiveness of topic modeling algorithms for short documents. Among those LDA gives the better result in terms of WUP similarity. This study will help to select the proper algorithm for labeling topics and can easily identify the meaning of topics.
  • No Thumbnail Available
    Item
    Context-based News Headlines Analysis Using Machine Learning Approach
    (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Springer, 2019-08-09) Rahman, Shadikur; Hossain, Syeda Sumbul; Islam, Saiful; Chowdhury, Mazharul Islam; Rafiq, Fatama Binta; Badruzzaman, Khalid Been Md.
    An increasing number of people are changing their way of thinking by reading news headlines. The interactivity and sincerity present in online news headlines are becoming influential to society. Apart from that, news websites build efficient policies to catch people’s awareness and attract their clicks. In that case, it is a must to identify the sentiment polarity of the news headlines for avoiding misconception. In this paper, we analyze 3383 news headlines generated by five major global newspapers during a minimum of four consecutive months. In order to identify the sentiment polarity (or sentiment orientation) of news headlines, we use 7 machine learning algorithms and compare those results to find the better ones. Among those Bernoulli Naïve Bayes technique achieves higher accuracy than others. This study will help the public to make any decision based on news headlines by avoiding misconception against any leader or governance and will help to identify the most neutral newspaper or news blogs.
  • No Thumbnail Available
    Item
    Performance Assessment of Multiple Machine Learning Classifiers for Detecting the Phishing URLs
    (Scopus, 2020-01-09) Rahman, Sheikh Shah Mohammad Motiur; Rafiq, Fatama Binta; Toma, Tapushe Rabaya; Hossain, Syeda Sumbul; Biplob, Khalid Been Badruzzaman
    In the field of information security, phishing URLs detection and prevention has recently become egregious. For detecting, phishing attacks several anti-phishing systems have already been proposed by researchers. The performance of those systems can be affected due to the lack of proper selection of machine learning classifiers along with the types of feature sets. A details investigation on machine learning classifiers (KNN, DT, SVM, RF, ERT and GBT) along with three publicly available datasets with multidimensional feature sets have been presented on this paper. The performance of the classifiers has been evaluated by confusion matrix, precision, recall, F1-score, accuracy and misclassification rate. The best output obtained from Random Forest and Extremely Randomized Tree with dataset one and three (binary class feature set) of 97% and 98% accuracy accordingly. In multiclass feature set (dataset two), Gradient Boosting Tree provides highest performance with 92% accuracy.
  • No Thumbnail Available
    Item
    Real-time Driver Drowsiness Detection using Deep Learning
    (Scopus, 2021) Dipu, Md. Tanvir Ahammed; Hossain, Syeda Sumbul; Arafa, Yeasir; Rafiq, Fatama Binta
    Every year thousands of lives pass away worldwide due to vehicle accidents, and the main reason behind this is the drowsiness in drivers. A drowsiness detection system will help to reduce this accident and save many lives around the world. To defend this problem, we propose a methodology based on Convolutional Neural Networks (CNN) that illustrates drowsiness detection as a task to detect an object. It will detect and localize whether the eyes are open or close based on the real-time video stream of drivers. The Mobile Net CNN Architecture with Single Shot Multibox Detector is the technology used for this object detection task. A separate algorithm is used based on the output given by the SSD_MobileNet_v1 architecture. A dataset that consists of around 4500 images was labeled with the object’s face yawn, no-yawn, open eye, and closed eye to train the SSD_MobileNet_v1 Network. Around 600 randomly selected images are used to test the trained model using the PASCAL VOC metric. The proposed approach is to ensure better accuracy and computational efficiency. It is also affordable as it can process incoming video streams in real-time and does not need any expensive hardware support. There only needs a standalone camera to be implemented using cheap devices in cars using Raspberry Pi 3 or other IP cameras.