Manish Kumar Pandey

Comparative study on machine learning techniques in predicting the QoS-values for web-services recommendations

This is an era of Internet computing and computing as a service on the internet is called cloud computing. Mainly three services like SaaS (applications), PaaS, and IaaS are being accessed through internet on demand, pay as per usage basis. Quality of Service (QoS) is the main issue in internet based computing for service providers and user-dependent as well as user-independent QoS parameters. In the current work we compared different machine learning algorithms for predicting the response time and throughput QoS values using past usage data. Bagging and support vector machines are found to be better performing prediction methods in comparison with other learning algorithms.

A Novel Storage Architecture for Facilitating Efficient Analytics of Health Informatics Big Data in Cloud

Analytics of health big data are very crucial for providing cost effective quality health care. Over recent years, the analytics on healthcare big data has evolved into a challenging task for getting insights into a very large data set for improving the health services. This enormous amount of data, which is being generated incessantly over a long period of time, has put a great deal of stress on the write performance as well as on scalability. Moreover, there is a requirement of efficient storage and meaningful processing of these data which is an another challenging issue. The traditional relational databases, which were used in the storage of health data, are now unable to handle due to its massive and varied nature. Besides, these databases have some inherent weakness in terms of scalability, storing varied data format, etc. So there is a necessity for a new kind of data storage management system. This paper proposes a new big data storage architecture consisting of application cluster and a storage cluster to facilitate read/write/update speedup as well as data optimization. The application cluster is used to provide efficient storage and retrieval functions from the users. The storage services will be provided through the storage cluster.

Parallel Implementations for Solving Shortest Path Problem using Bellman-Ford

this paper, different parallel implementations of Bellman- Ford algorithm on GPU using OpenCL are presented. These variants include Bellman-Ford for solving single source shortest path (SSSP) having two variants and Bellman-Ford for all pair shortest path (APSP) problems. Also, a comparative analysis of their performances on CPU and GPU is discussed in this paper.Write-write consistency in Bellman- Ford is overcome using synchronization mechanism available in OpenCL and by explicit synchronization by modifying the algorithm.An average speed up of 13.8x for parallel bellman ford for SSSP and an average speed up of 18.5x for bellman ford for APSP is achieved by proposed algorithm.

A Fine Tuned Hybrid Implementation for Solving Shortest Path Problems using Bellman Ford

this paper a hybrid implementation for Bellman-Ford to solve shortest path problems is proposed using OpenCL. Here first parallel implementation for Bellman-Ford for single source shortest path (SSSP) problem and all pair shortest path (APSP) are analyzed on CPU and GPU and based on this analysis work is divided among CPU and GPU and hybrid implementation is done. As proper resource utilization is done here we have termed it a fine tuned implementation. We have got considerable speedup of 2.88x over parallel implementation on GPU for SSSP and 3.3x over parallel implementation of Bellman-Ford for APSP on GPU. Keywordspath problem , OpenCL , Graphical processing unit(GPU).

Social Networking and Big Data Analytics Assisted Reliable Recommendation System Model for Internet of Vehicles

The devices are becoming ubiquitous and interconnected due to rapid advancements in computing and communication technology. The Internet of Vehicles (IoV) is one such example which consists of vehicles that converse with each other as well as with the public networks through V2V (vehicle-to-vehicle), V2P (vehicle-to-pedestrian) and V2I (vehicle-to-infrastructure) communications. The social relationships amongst vehicles create a social network where the participants are intelligent objects rather than the human beings and this leads to emergence of Social Internet of Vehicles (SIoV). The big data generated from these networks of devices are needed to be processed intelligently for making these systems smart. The security and privacy issues such as authentication and recognition attacks, accessibility attacks, privacy attacks, routing attacks, data genuineness attacks etc. are to be addressed to make these cyber physical network systems very reliable. This paper presents a comprehensive survey on SIoV and proposes a novel social recommendation model that could establish links between social networking and SIoV for reliable exchange of information and intelligently analyze the information to draw authentic conclusions for making right assessment. The future Intelligent IoV system which should be capable to learn and explore the cyber physical system could be designed.

Performance Analysis of Time Series Forecasting Using Machine Learning Algorithms for Prediction of Ebola Casualties

There is an immense concern on our vigilance for controlling the spread of pandemics such as Ebola, Zika, and H1N1 etc. through state of art technology. The dynamics become very complex of epidemics in sweeping population. Efficient descriptive, predictive, preventive and prescriptive analyses on the huge data generated by SMAC are very crucial for valuable arrangement and associated responsive tactics. In this paper, we have proposed the use of machine learning techniques for performance evaluation of time series forecasting of Ebola casualties. By experimenting without lag creation, we achieved the best results in the MAE of 7.85%, RMSE value of 61.14%, and Direction Accuracy of 85.99% with Random Tree Classifier. Thus we can conclude that by using these models for forecasting epidemic spread and developing public health policies leads the health authorities to ensure the appropriate actions for the control of the outbreak.

Performance Analysis of Ensemble Supervised Machine Learning Algorithms for Missing Value Imputation

In this era of cloud computing, web services based solutions are gaining popularity. The applications running on distributed environment seek new parameters for them to perform efficiently to satisfy end users requirements. Finding these parameters for increasing efficiency has become a talk of researchers now days. Non functional performance of a web service is described through User dependent QoS properties. These QoS parameters are generally described in WS-Policy in Service Level Agreement (SLA). Usually in web service QoS datasets, web service QoS values are missing, which makes missing value imputations an important job while working with cloud web services. In the current work we compared the prediction accuracy of two groups of supervised machine learning ensembles based Meta learners: bagging and additive regression (boosting) with a fusion of the seven base learners in both. Random forest is found to be better performing in both Meta learners: bagging and boosting than other learning algorithms.

OpenCL Parallel Blocked Approach for Solving All Pairs Shortest Path Problem on GPU

Pairs Shortest Path Problem (APSP) finds a large number of practical applications in real world. This paper presents a blocked parallel approach for APSP using an open standard framework OpenCL, which provides development environment for utilizing heterogeneous computing elements of computer system and to take advantage of massive parallel capabilities of multi-core processors such as graphics processing unit (GPU) and CPU. This blocked parallel approach exploits the local shared memory of GPU, thereby enhancing the overall performance. The proposed solution is for directed and dense graphs with no negative cycles and is based on blocked Floyd Warshall (FW) and Kleenes algorithm. Like Floyd Warshall this approach is also in-place and therefore requires no extra memory.

A note on boundaries in atlas maps

Thematic maps at very small scales like 1:1 million, 1:10 million and smaller in various atlases show smooth boundaries of areal units as well as linear features. Atlas maps are normally meant for visualization of distributional patterns. But many users believe that the boundaries are extracted/delineated/mapped at larger scales and are then reduced to go into atlases or into various publications and hence are taken to be reasonably accurate as the maps carry a stamp of premier map making organizations and/or publishers. Problem arises when the boundaries, both political as well as of natural units, are borrowed and linear or areal measurements are made from them for analysis of any thematic units in them. In a recent case, it so happened with the authors involved in a piece of work connected with Indus and Ganga-Yamuna river basins within Indian political boundary, the basin boundaries from a map at 1:14 million scale published in National Atlas and Thematic Mapping Organization Atlas (NATMO, 1999) are borrowed for some

Parallel Implementation of Travelling Salesman Problem using Ant Colony Optimization

In this paper we have proposed parallel implementation of Ant colony optimization Ant System algorithm on GPU using OpenCL. We have done comparison on different parameters of the ACO which directly or indirectly affect the result. Parallel comparison of speedup between CPU and GPU implementation is done with a speed up of 3.11x in CPU and 7.21x in GPU. The control parameters α, β, ρ is done with a result of best solution at 1, 5 and 0.5 respectively.