Nirmal Dhara, Shashank K
Department of Computer Science, Christ University, Bangalore, India.
Abstract— The social networking sites are adding enormous load to the servers. Balancing such an increasing load for faster response is really challenging. We propose a cost effective, loosely coupled and highly scalable middleware architecture, Free Scale Architecture (FSA), which helps the sites to handle increasing load with minimal changes to its fundamental architecture by dividing the load among several servers. FSA uses Classification Data Engine (CDE), Tiny Intelligent Agent (TIA), Logical Router (LR), Dynamic Bottleneck Tracker (DBT) and SOLR as part of the architecture which helps fetch data and classify, into clusters, shrink Master and Slave data, balance the increasing load, help find the reason of any bottleneck with balancing and routing the load among servers, and cache the most frequently used data for faster retrieval, respectively. Further, CDE requires starting a separate thread to copy the clusters into a local domain, and this is known as parallel process. The parallel process saves nearly 80% of the time. We also discuss the implementation of FSA and SOLR architecture for social networking sites, and how social networking sites can use the FSA and SOLR architecture to increase scalability and handle the unexpected load.
Keywords-Free Scale Architecture (FSA); Logical Router (LR); Classification Data Engine (CDE); Social Networks (SN); Failure Management (FM); Dynamic Bottleneck Tracker (DBT); Simple Failure Handling (SFH); Permanent Failure (PF); Temporary Failure (TF); Critical Failure (CF); Tiny Intelligent Agent (TIA).
Chandra J, Nachamai.M
Associate Professor, Department of Computer Science, Christ University, Bangalore, Karnataka, India.
Anitha S Pillai
Professor and Head, Department of MCA, Hindustan University, Chennai, Tamil Nadu, India.
Abstract — Investment prediction is a method to decide the future values of stock indexes and commodity exchanges or trading of financial services. The aim of the model is to perform optimized prediction on commodities and stock market indexes. The investment prediction is an important task for an investor to maximize his or her return on investment. The purpose of the paper is to propose an optimized model using computational intelligence and it is a step by step method that follows an integrated approach which can solve several complex problems in predictive analytics. The integrated approach in this paper uses genetic algorithm, pearson’s correlation coefficient and multilayer perceptron adaline feed forward neural network to predict the next business day high values of stock indexes and commodities trading. As an integrated method, the model uses genetic algorithm as first step to check the data optimization, since the data is considered as an important element in data analytics. The optimized data is extracted using correlation coefficient and the classifier prediction is done with multilayer perceptron adaline feed forward neural network for making the prediction. The proposed model was implemented continuously on three months data to evaluate the performance and to check the accuracy on the NSEindia. The predicted values were checked against the next business day of original values, the predicted result is very close to the original values. The model is evaluated with the statistical parameter MRE, MMRE and the accuracy rate. In comparison with other existing methods, the current method outperforms other testing patterns.
Keywords- Predictive Analytics (PA); Computational Intelligence (CI); Genetic Algorithm (GA); Pearson’s Correlation Coefficient(PCC); Multi Layer Perceptron (MLP); Adaptive Linear Element (ADALINE); Neural Network (NN).