International Journal of Computer (IJC - Global Society of Scientific Research and Researchers, GSSRR)
Not a member yet
459 research outputs found
Sort by
Sentiment Analysis of News Event-based Social Network using Data Mining Technique
Get PDFThe increasing popularity of social media takes the attention of the internet users across the word wide to discuss and share the events/things they are interested on social media blogs/sites. Consequently, an explosive increase of social media data spread on the web has been promoting the development of analysis of social media news depending on the news or events, the latest trend of the social big data. The sentiment analysis of news event becomes an important research area for many real-world applications, such as public opinion monitoring for government and news recommendation of news websites. In this paper, we perform sentiment analysis for news events based on posts, and comments of the users upon a news event. We use two data mining techniques namely naïve Bayesian and support vector machine to reveal what the polarity/meaning of the post is such as positive, negative or polarity. There are two main stages in performing this task called training and testing phases. The first phase uses the training datasets of the news event and the second phase use newly inputted data of the user to classify the polarity of the user news posts or comments. We then execute the experiments for each algorithm and then collect the experimental results and compare them with accuracy with known and unknown test data with different volumes of tweet transactions. According to the results, both of them can accurately reveal the opinions of the social network users
To have an Idea on NoSQL Databases
Get PDFNoSQL databases (initially non-SQL, then Not Only SQL) are specifically designed to handle large amounts of data. They have been developed since the 1970s, but they have gained the interest of academia and industry for about two decades. This is because of their powerful characteristics and lack of relational databases, which are the most widely used data sources around the world. Indeed, these databases are based on the relational model, which is materialized by a relational database management system (RDBMS). Although RDBMS efficiently manage data (tables), they have many drawbacks that make them unsuitable for managing current data, which come mainly from Internet applications. They are called Big Data and they are used for example by Twitter, FaceBook, LinkedIn, .... They are very numerous and tend to change quickly. In fact, among the disadvantages of relational databases, we can mention: non-flexibility, non-scalability, ... On the contrary, NoSQL databases evolve very well (scaling) and almost all NoSQL databases are schema-free (we can add or delete an entity or a relationship at any time during execution). In this article, we begin by giving an overview of relational databases and their characteristics. We then describe the NoSQL databases and their main characteristics, knowing that there are as many different characteristics as "NoSQL databases" products. We then give the taxonomy of NoSQL databases, which distinguishes four main types of NoSQL databases: key-value, wide-column, document and graphical databases. We will then give some elements of each type of database through the use of a product, an implementation of a kind of such a database
Comparative Study for Text Document Classification Using Different Machine Learning Algorithms
Get PDFClassification is a supervised learning method: the goal is finding the labels of the unknown object. In the real world, the tedious amounts of manual works are required to label the unknown documents. The system is initially trained by labeled documents by using one of the supervise machine learning algorithm and then applied trained model to predict the label of the unknown documents. The framework of text document classification consists of: input text document, pre-processing, feature extraction and classification. The analysis four common classification methods are performed: Naïve Bayes, Decision Tree, Support Vector Machine and K-nearest neighbors for text document classification. The main focus of this paper is to present comparative study of different exiting classification methods for text document classification. The experiment performed different classification methods on the Enron Email Dataset and measure classification accuracy, true positive, true negative, false positive and false negative to compare the performance of different classification methods
Network Vulnerability Analysis
Get PDFA network vulnerability analysis (NVA) is the process of identifying and evaluating those security loopholes that might exist in an enterprise before the network is hacked. This paper discussed network vulnerability analysis and its processes. It also demonstrate some approaches to conducting vulnerability analysis for any network environment using manual tools and showed other automatic tools used for conducting vulnerability analysis. The paper also addressed the reasons for network vulnerability analysis and finally discussed vulnerabilities of the network. Findings from this study indicate that network vulnerability analysis is important. However, conducting vulnerability analysis does not necessarily prevent security on its own; instead, it reflects a snapshot of the environment at a particular point in time and pointed out the rationale for regular conducting of vulnerability analysis
Survey on Detection Methods for Self-Driving Cars
Get PDFAccurate vehicle detection or classification plays an important role for self-driving cars. Objects classification and detection can be used in various such as Robotics, Medical Diagnosis, Safety, Industrial Inspection and Automation, Human Computer Interface, Advanced Driver Assistance System and Information Retrieval. In this article, we investigated the methods of detection and classification in context images and videos. SIFT, HOG, SVM, CNN, faster RCNN and YOLO methods are reviewed to detect and recognize the objects. The paper aims to know the methods that detect the obstacles on the way to reduce the traffic accidents. We summarize the results, faster-RCNN is better than the other methods for real-time citing the advantages and disadvantages of existing methods
Classification of Students Based on Academic Ability Using Profile Matching and Linear Interpolation Weighting
Get PDFHigher education institutions play an important role in learning activities, both academic and non-academic, including establishing a social transition to adjust to the Fourth Industrial Revolution (4IR). Higher education in Indonesia is generally divided into classes with heterogeneous characteristics that cause less conducive teaching and learning process. Clustering of students in a particular group (homogeneous) is expected to improve acceleration and effectiveness of learning. Multicriteria analysis needs to be done to avoid errors of judgment in the determination of the class. Selection methods may affect the quality of the resulting decisions. This research profile matching method applying in determining the clustering of students, which is assessed based on the ideal profile of a superior class. The criteria that form the basis of assessment is the value of two semesters learning achievement in the first year, the value of the course, the expertise, and mastery of programming languages as well as activity in the organization\u27s activities. Weighting difference in value (gap) with a certain range is calculated using linear interpolation. Output in the form of a ranking system that helps decision-makers to the cluster of students accurately and efficiently
Evaluation of Similarity Metrics Under the Context of an Autonomous Reactive System
Get PDFCurrently, in the field of robotics, institutions and researchers are working on the design and development of autonomous navigation systems on robots for dynamic environments. The most advanced implementations of autonomous behaviors are found on vehicles or wheeled devices, allowing them to move on controlled environments and even on rough terrain. In this paper, it is presented the design of an autonomous reactive system for humanoid robots. This system requires to know the current state of the robot, during a specific activity, to make the right reactive action for a specific situation. In the context of inquiring the current state of the robot, we consider the implementation of a knowledge base populated with diverse states of the joints and their possible reactive actions. To recover the possible reactive actions from the knowledge base, it is required to search for the current state of the robot in the knowledge base. However, this process may incur in high computational cost depending on the size of the knowledge base. Therefore, in this work, we carried out a comparative study of six similarity metrics, with the objective of identifying the metric that offers the best computational time. In these studies, it is identified that the metrics with lower mathematical complexity showed the best results. Additionally, we used Wilcoxon and Friedman statistics tests to assess the performance of the similarity metrics. Finally, we included an analysis of the characteristics and functionality of several similarity metrics, which showed that some of them are not suitable in the context of our proposal. On the other hand, other metrics were identified as viable and with potential for future works
A Survey in Deep Learning Model for Image Annotation
Get PDFImage annotation is generating the human-understandable natural language sentence for images. Annotating the image with sentence is one kind of the computer vision process that includes in the artificial intelligence. Annotation is working by combining computer vision and natural language processing. In image annotation, there are two types: sentence based annotation and single word annotation. Deep learning can get the more accurate sentence for the image. This paper is the survey for image annotation that applied the deep learning model. This discusses existing methods, technical difficulty, popular datasets, evaluation metrics that mostly used for image annotation
Review on Blockchain Technology for Healthcare Records
Get PDFFor a long time, the healthcare system has been facing many problems such as incomprehensible doctor’s hand-writing, problematic retrieval of patient information and patients are unable to monitor their own data. In addition to that, there are problems afflicting the medical record systems such as how to share the medical data for various purposes without risking data privacy and security. Due to technology advancements, it is now possible to improve the current situations and to minimise these problems by providing more personalised services to the patients. This paper explains the interesting area of blockchain in healthcare, evaluating the blockchain technology from the multiple perspectives around healthcare data and explaining some platforms which could be used for the blockchain system development. Finally, the challenges of implementing the blockchain technology are deliberated.
Machine Learning Technique and Normalization Cross Correlation Model Applied for Face Recognition
Get PDFFace recognition systems just like any other biometric systems have continued to stand the test of time as a reliable means of human verification and identification. The high rate of fraud, crime, and terrorism in Nigeria and the world at large makes it increasingly necessary to have recognition systems that will be compatible with security devices currently deployed. However, the accuracy of facial recognition system is dependent on the adequacy of the model applied. This work applies a combination of Support Vector Machine (SVM) and Normalization Cross Correlation (NCC) starting with a preprocessing stage that involves filtering, cropping, normalization as well as histogram equalization of the face images. The facial images were trained and classified using Support Vector Machine then verified by NCC. The experimental study of the model with benchmarked face images showed that the model is very suitable for obtaining a better accuracy level. The False Acceptance Rate (FAR), False Rejection Rate (FRR), Genuine Acceptance Rate (GAR) and Total Error Rate (TER) values established the superiority of the proposed model over some related ones