Browsing by Author "Rath, Santanu Kumar"
Now showing 1 - 4 of 4
- Results Per Page
- Sort Options
Article Citation Count: 126Co-LSTM: Convolutional LSTM model for sentiment analysis in social big data(Elsevier Sci Ltd, 2021) Mısra, Sanjay; Jena, Monalisa; Rath, Santanu Kumar; Misra, Sanjay; Computer EngineeringAnalysis of consumer reviews posted on social media is found to be essential for several business applications. Consumer reviews posted in social media are increasing at an exponential rate both in terms of number and relevance, which leads to big data. In this paper, a hybrid approach of two deep learning architectures namely Convolutional Neural Network (CNN) and Long Short Term Memory (LSTM) (RNN with memory) is suggested for sentiment classification of reviews posted at diverse domains. Deep convolutional networks have been highly effective in local feature selection, while recurrent networks (LSTM) often yield good results in the sequential analysis of a long text. The proposed Co-LSTM model is mainly aimed at two objectives in sentiment analysis. First, it is highly adaptable in examining big social data, keeping scalability in mind, and secondly, unlike the conventional machine learning approaches, it is free from any particular domain. The experiment has been carried out on four review datasets from diverse domains to train the model which can handle all kinds of dependencies that usually arises in a post. The experimental results show that the proposed ensemble model outperforms other machine learning approaches in terms of accuracy and other parameters.Article Citation Count: 12Comparative Study of Real Time Machine Learning Models for Stock Prediction through Streaming Data(Graz Univ Technolgoy, inst information Systems Computer Media-iicm, 2020) Mısra, Sanjay; Das, Sushree; Rath, Santanu Kumar; Misra, Sanjay; Damasevicius, Robertas; Computer EngineeringStock prediction is one of the emerging applications in the field of data science which help the companies to make better decision strategy. Machine learning models play a vital role in the field of prediction. In this paper, we have proposed various machine learning models which predicts the stock price from the real-time streaming data. Streaming data has been a potential source for real-time prediction which deals with continuous flow of data having information from various sources like social networking websites, server logs, mobile phone applications, trading floors etc. We have adopted the distributed platform, Spark to analyze the streaming data collected from two different sources as represented in two case studies in this paper. The first case study is based on stock prediction from the historical data collected from Google finance websites through NodeJs and the second one is based on the sentiment analysis of Twitter collected through Twitter API available in Stanford NLP package. Several researches have been made in developing models for stock prediction based on static data. In this work, an effort has been made to develop scalable, fault tolerant models for stock prediction from the real-time streaming data. The Proposed model is based on a distributed architecture known as Lambda architecture. The extensive comparison is made between actual and predicted output for different machine learning models. Support vector regression is found to have better accuracy as compared to other models. The historical data is considered as a ground truth data for validation.Article Citation Count: 18Distributed Centrality Analysis of Social Network Data Using MapReduce(Mdpi, 2019) Mısra, Sanjay; Rath, Santanu Kumar; Misra, Sanjay; Damasevicius, Robertas; Maskeliunas, Rytis; Computer EngineeringAnalyzing the structure of a social network helps in gaining insights into interactions and relationships among users while revealing the patterns of their online behavior. Network centrality is a metric of importance of a network node in a network, which allows revealing the structural patterns and morphology of networks. We propose a distributed computing approach for the calculation of network centrality value for each user using the MapReduce approach in the Hadoop platform, which allows faster and more efficient computation as compared to the conventional implementation. A distributed approach is scalable and helps in efficient computations of large-scale datasets, such as social network data. The proposed approach improves the calculation performance of degree centrality by 39.8%, closeness centrality by 40.7% and eigenvalue centrality by 41.1% using a Twitter dataset.Article Citation Count: 32Large Scale Community Detection Using a Small World Model(Mdpi, 2017) Mısra, Sanjay; Rath, Santanu Kumar; Misra, Sanjay; Damasevicius, Robertas; Maskeliunas, Rytis; Computer EngineeringIn a social network, small or large communities within the network play a major role in deciding the functionalities of the network. Despite of diverse definitions, communities in the network may be defined as the group of nodes that are more densely connected as compared to nodes outside the group. Revealing such hidden communities is one of the challenging research problems. A real world social network follows small world phenomena, which indicates that any two social entities can be reachable in a small number of steps. In this paper, nodes are mapped into communities based on the random walk in the network. However, uncovering communities in large-scale networks is a challenging task due to its unprecedented growth in the size of social networks. A good number of community detection algorithms based on random walk exist in literature. In addition, when large-scale social networks are being considered, these algorithms are observed to take considerably longer time. In this work, with an objective to improve the efficiency of algorithms, parallel programming framework like Map-Reduce has been considered for uncovering the hidden communities in social network. The proposed approach has been compared with some standard existing community detection algorithms for both synthetic and real-world datasets in order to examine its performance, and it is observed that the proposed algorithm is more efficient than the existing ones.