Sireesha Rodda
@gitam.edu
PROFESSOR
GITAM (DEEMED TO BE UNIVERSITY)
RESEARCH INTERESTS
Artificial Intelligence, Machine Learning, Data Mining
Scopus Publications
Scopus Publications
Nishanth Reddy Dereddi, Sireesha Rodda, and Soujanya Buddharaju
Springer Nature Singapore
Deepthika Challakonda, Muskaan Raza, Chinni Krishna Koppisetti, Akshay Manikonda, and Sireesha Rodda
Springer Nature Switzerland
Nammi Hemanth Kumar and Sireesha Rodda
Seventh Sense Research Group Journals
Deepika Puvvula and Sireesha Rodda
International Information and Engineering Technology Association
Vijaya Bharathi Manjeti, R Sireesha, D Dhanush, M Abhiram, and Bangaru Lakshmi Mahanthi
IEEE
Sadhana Priyadarshini and Sireesha Rodda
ENGG Journals Publications
The Recurrent Subgraph Extraction plays a key role in the Graph Mining field when our data is distributed over networks. This paper emphasizes different types of graph mining algorithms with the Giraph Distributed System to get more desirable and valuable results than existing methods. We discuss how our proposed model MapReduce Geometric Multi-way Advanced Optimized Frequent Subgraph Mining (MGMAOFSM) impacts different graph mining mechanisms for centralized and distributed systems. The comparison is done for different criteria such as memory requirement or execution time with real four datasets (Facebook Social Network, Coronavirus (COVID-19) tweets, Google web graph, Patent Citation Network) with different threshold values. We implement various algorithms such as Triangle Closing, Shortest Path, Connected Components, and PageRank algorithms, and find out our proposed algorithm that requires less memory with the Triangle closing algorithm whereas in the case of PageRank is lowest with all threshold values.
Sadhana Priyadarshini and Sireesha Rodda
IEEE
Graph Mining has been the most demanding research area for the last few decades in different fields, such as biological networks, the world wide web, mobile applications, sensors, online, social networks, etc. Frequent Subgraph Mining (FSM) plays a vital role in Graph Mining to exercise, study and generate interesting patterns from graph data. Basically, FSM techniques are classified into two types such as an apriori-based method, and a pattern growth-based method. This technique faces the problems such as the generation of the duplicate frequent subgraph, having no proper technique to rank during candidate generation, and how to map the threshold values. In this proposed system, a Dynamic PageRank GraphX- based Frequent Subgraph Mining (DPRGFSM) model that is able to extract interesting patterns from the distributed system by eliminating duplicates by ranking them to the proper level. In addition, we also use load balancing, pre-punning, and optimization techniques to improve its performance in both memory requirements and time complexity. The potency of methods defined in this paper is evaluated rigorously with different threshold values and comparative studies with different parameters with existing Spark- based Single Graph Mining (SSIGRAM) and A Ranked Frequent pattern Growth Framework (A- RAFF) and found drastic improvement with all four datasets. The proposed methodology is 1.6 times faster than the Spark-based Single Graph Mining (SSIGRAM) model and 50 times faster than the A Ranked Frequent pattern Growth Framework (A- RAFF) for recurrent subgraph extraction.
Sadhana Priyadarshini and Sireesha Rodda
IEEE
Data Mining has a subpart called Frequent Subgraph Mining (FSM) and is a demanding area for the implementation of graph classification and graph clustering which is used in the area of the social network, chemical compounds, and biological datasets, enterprise world. Many research workers have been researching on how to produce an effective and optimized technique to extract the candidate subgraphs by eliminating duplicates for the last few decades. In the case of the Giraph distributed system, a different format for input and output classes is required to take graphs into memory and put graphs after completion of its operation, which leads to excessive memory exhaustion. In this paper, a novel methodology “Giraph Dynamic Sized Structure Frequent Subgraph Mining (GDSSFSM)” has been developed to reduce the memory necessity for FSM in a graph-distributed system. The proposed approach reorganizes the inner input format class (i.e. setEdgeInputFormatClass) without any changes. Hence, it can be used by default in a customized format. The experimental analysis is done on the different datasets with an existing algorithm based on execution time and memory requirements and concludes that it decreases up to on average 52% depending on the dataset and the graph (i.e., PageRank, Connected Components, and Simple Shortest Path) edge-centric algorithm. The proposed algorithm can be used in various fields of graph mining such as social networks, bioinformatics, and web data mining
Shanti Chilukuri, Sireesha Rodda, and Lakshmana Rao Kalabarige
Inderscience Publishers
Srija Rallabhandy and Sireesha Rodda
Springer Singapore
Online shopping has gained popularity for its omnipresence. However, visually impaired people are not able to make complete use of this e-commerce shopping due to lack of user-friendly nature to the visually impaired. Here, in this paper, we have proposed a solution to make the e-commerce websites more user-friendly to the visually impaired using voice-based assistance. Our solution includes Face Recognition technology using OpenCV for login and registration into the e-commerce website. gTTS (Google Text to Speech) and speech_recognition libraries were used for making it completely speech driven. After the search results, to extract the data from the web page Web Scraping was used and the results were stored in the database to analyse the data and to choose the best-rated products. After selection of the product, the product was added to the cart using Selenium Web Driver.
Sadhana Priyadarshini and Sireesha Rodda
Springer Singapore
In the present time, Graph Mining has become the most research-oriented field in the advance technologies for its importance in many areas, such as citation graphs, web data mining, chemical structures, protein interaction, social networks, etc. The rapid change in Graph Mining research work is fully dependent on the field of Graph Partitioning (GP) as well as Frequent Subgraph Mining (FSM). In this paper, we define Geometric Multi-Way Frequent Subgraph Mining (GMFSM) approach, which is based on Geometric Partition of a Single Large Graph Database with Frequent Subgraph Mining (FSM) approach that uses filtration technique to reduce number of candidate subgraphs. After partitioning the large graph database, we execute FSM algorithm simultaneously on each subparts which produce the desire result much faster (one-third to half) than existing algorithms. In addition, we use two-way partitioning algorithm recursively to obtain multi-way partition which drastically changes the performance of the algorithm.
Pritee Parwekar and Sireesha Rodda
IGI Global
The energy of a sensor node is a major factor for life of a network in wireless sensor network. The depletion of the sensor energy is dependent on the communication range from the sink. Clustering is mainly used to prolong the life of a network with energy consumption. This paper proposes optimization of clustering using genetic algorithm which will help to minimize the communication distance. The cluster overhead and the active and sleep mode of a sensor is also considered while calculating the fitness function to form the cluster. This approach helps to prolong the network life of sensor network. The proposed work is tested for different number of nodes and is helping to find the correct solution for the selection of cluster heads.
Pritee Parwekar, Sireesha Rodda, and Parmeet Kaur
IGI Global
A WSN consists of a large number of limited computation and storage capability wireless sensor nodes, which communicate wirelessly. These sensor nodes typical communicate in short range and collaborate to accomplish the network function. To increase the range of sensing and with the advent of MEMS, mobile sensors and sinks is the technology the world is moving to. This paper presents a network of mobile sensors and a sink. A mobile sink is selected as check-point to have the recoverability of the network. A Fuzzy Rule based system (FRS) is used to construct and select efficient static sensor nodes having adequate resources as Check Point Storage Nodes (CPSNs). The objective of FRS is to increase the probability of recovery of check-pointed data subsequent to a failure, thereby allowing a distributed application to complete its execution successfully. Simulations show FRS's better recovery probabilities in comparison to a random check-pointing arrangement.
G. Chakravarthy, K. Anupam, P. Varma, G. H. Teja and S. Rodda
Visually impaired people face a lot of challenges in day-to-day life. Having seen the difficulties faced by them, our primary objective is to facilitate confidence and to empower them to lead a life free from threats related to their safety and well-being. The lack of ability to identify known individuals in the absence of auditory or physical interaction cues drastically limits the visually challenged in their social interactions and poses a threat to their security. Over the past few years many prototype models have been developed to aid this population with the task of face recognition. This application will reduce the inherent difficulty for recognition of a person. It will present a facial recognition application with an intuitive user interface that enables the blind to recognise people and interact socially. The carefully designed interface lets the visually challenged to be able to access and use it without any requirement for visual cues as the users are acquainted by a voice assistant to navigate through the application. The entire build is designed to run efficiently on a Raspberry Pi 3 model B module using the Android Things platform. The Open CV library has been used for the detection and recognition of people in this project. This enables the scope for the software to be run on a multitude of devices such as camera embedded glasses to warn users of their surroundings and identify people to interact safely. Since everything in the application is done in real time with no requirement for prior datasets to be hardcoded it drastically improves the versatility of the software. We hope to make the visually impaired feel closer, comfortable and more secure with the world surrounding them through our application.
M. Pranay, Hari Varshini Rajkumari, Sireesha Rodda, Y. Srinivas, and P. Anuradha
Springer Singapore
Technology cannot run essentially without the input of a human. The rate of technological advancement is increasing with time, society is looking to create and develop easier ways to live and lengthen their lives. The internet is a massive source of information that millions of people use and depend on every day. Artificial Intelligence (AI) is intended to do the thinking for us often thinking through things very quickly that we do not have enough information or time to process ourselves. So Gideon will help to do daily task and it acts like artificial companion. As compared to other application Gideon has additional feature face recognition which can able to detect faces in real world and try to recognize known faces, which we may have forgotten. It can able to listen to you and provide appropriate response precisely and quickly.
Many aspects of our life now continually rely on computers and internet. Data sharing among networks is a major challenge in several areas, including communication, national security, medicine, marketing, finance and even education. Many small scale and large scale industries are becoming vulnerable to a variety of cyber threats due to increase in the usage of computers over network. We propose Fuzzy-ECOC frame work for network intrusion detection system, which can efficiently thwart malicious attacks. The focus of the paper is to enforce cyber security threats, generalization rules for classifying potential attacks, preserving privacy among data sharing and multi-class imbalance problem in intrusion data. The Fuzzy-ECOC framework is validated on highly imbalanced benchmark NSL_KDD intrusion dataset as well as six other UCI datasets. The experimental results show that Fuzzy-ECOC achieved best detection rate and least false alarm rate.
M. Vijaya Bharathi and Sireesha Rodda
Inderscience Publishers
M. Vijaya Bharathi and Sireesha Rodda
Inderscience Publishers
AJAX build 2.0 web applications depend in light of state full unique client and server correspondence and client side control of the DOM tree, which makes not the same as standard web applications. Provoking to more slip-ups and harder to set. Another methodology for this AJAX named ATUSA based web applications has been perceived for such accuses that can occur in any state and for making the test suite covering the ways. This approach called as ATUSA, realised by using a gadget which offers invariant checking of portions, module instrument. We portray this framework in three phases with six segments and furthermore accuse revealing limits, versatility, manual effort and level of motorisation testing. This paper mainly concentrates on rookie's vantage point of testing modern web application based on so far accomplished potential research done by software practitioners and experts.
Kavya Devabhakthuni, Bhavya Munukurthi, and Sireesha Rodda
Springer Singapore
Surface transportation in urban cities is inevitable to move from one place to another place for carrying out regular activities. Taxis are assumed as one of the essential parts for transportation in New York. This paper focuses on the selection of the top profitable areas using New York City (NYC) taxi trips dataset. The data used in the current work is captured from the NYC taxi and analyzed using Hadoop Big Data to find the profitable locations for the taxi driver, so that they can increase their income by waiting in most profitable locations.
Pritee Parwekar, Sireesha Rodda, and Parmeet Kaur
IGI Global
A WSN consists of a large number of limited computation and storage capability wireless sensor nodes, which communicate wirelessly. These sensor nodes typical communicate in short range and collaborate to accomplish the network function. To increase the range of sensing and with the advent of MEMS, mobile sensors and sinks is the technology the world is moving to. This paper presents a network of mobile sensors and a sink. A mobile sink is selected as check-point to have the recoverability of the network. A Fuzzy Rule based system (FRS) is used to construct and select efficient static sensor nodes having adequate resources as Check Point Storage Nodes (CPSNs). The objective of FRS is to increase the probability of recovery of check-pointed data subsequent to a failure, thereby allowing a distributed application to complete its execution successfully. Simulations show FRS's better recovery probabilities in comparison to a random check-pointing arrangement.
Sireesha Rodda
Springer Singapore
With the growth of network activities and data sharing, there is also increased risk of threats and malicious attacks. Intrusion detection refers to the act of successfully identifying and thwarting malicious attacks. Traditionally, the help of network security experts is sought owing to their familiarity with the network technologies and broad knowledge. Recently, data mining techniques have been increasingly adopted to perform network intrusion detection. This paper presents the comparison between multi-layer perceptron and radial basis function networks for designing network intrusion detection system. Multi-layer perceptron proved to be more effective than radial basis function when applied on the benchmark NSL_KDD dataset.
Sireesha Rodda and Uma Shankar Rao Erothi
Springer Singapore
A wide range of malicious attacks and threats are increasing day by day with the growth and development of internet and network technologies. Enforcing network security is important to protect data or information in the computer network against attacks from intruders. The right of privacy of the user must be respected even on the network-resident data. This paper evaluates the performance of four different classifiers on a standard network intrusion detection dataset. The original values in the dataset are anonymized in order to protect the user’s privacy. All the experiments were performed on IBM SPSS Premium Modeler. The effectiveness of the techniques is tested using different evaluation measures.
Pritee Parwekar and Sireesha Rodda
Inderscience Publishers