A Big Data Application for Energy Consumption Management of Commercial Buildings

The demand for energy is increasing rapidly and, after a few years, it may surpass the available energy, which may lead the energy providers to increase the cost of energy consumption to compensate the cost for the production. This paper provides design and implementation details of a prototype big data application developed to help large buildings to automatically manage their energy consumption by setting energy consumption targets, collecting periodic energy consumption data, storing the data streams, displaying the energy consumption graphically in real-time, analyzing the consumption patterns, and generating energy consumption graphs and reports. The application is connected to Mongo NoSQL backend database to handle the large and continuously changing data. This big data energy consumption management system is expected to help the users in managing energy consumption by analyzing the patterns to see if it is within or above the desired consumption targets and displaying the data graphically.

Classifying UNSW-NB15 Network Traffic in the Big Data Framework using Random Forest in Spark

The focus of this work is on detecting and classifying attacks in network traffic using a binary as well as multi-class machine learning classifier, Random Forest, in a distributed Big Data environment using Apache Spark. The classifier is tested using the UNSW-NB15 dataset. Major problems in these types of datasets include high dimensionality and imbalanced data. To address the issue of high dimensionality, both Information Gain as well as Principal Components Analysis (PCA) were applied before training and testing the data using Random Forest in Apache Spark. Binary as well as multi-class Random Forest classifiers were compared in a distributed environment, with and without using PCA, using various number of Spark cores and Random Forest trees, in terms of performance time and statistical measures. The highest accuracy was obtained by the binary classifier at 99.94%, using 8 cores and 30 trees. This study obtained higher accuracy and lower FAR rates than previously achieved, with low testing times.

A LIGHTWEIGHT SYSTEM TOWARDS VIEWING ANGLE AND CLOTHING VARIATION IN GAIT RECOGNITION

In this paper, the authors have proposed a computationally efficient, robust, and lightweight system for gait recognition. The proposed system contains two main stages: In the first stage, a classification network identifies optical flow corners in the normalized silhouette and calculates the distances traveled in every viewpoint which is further used by a regression model to identify the viewing angle. In the second stage, a feature extraction network computes the Gait Energy Image (GEI) for every viewpoint and then uses Principal Component Analysis (PCA) to extract low dimensional feature vectors from these GEI images. Finally, a multi-layer perceptron model is trained using the extracted principal components for every viewing angle. The performance of a system is comprehensively evaluated on the CASIA B and OULP gait dataset. The experimental results demonstrate the superior performance of a proposed system in viewing angle classification (100% accuracy), gait recognition (100% accuracy in normal walk), computational efficiency, robustness to clothing, and viewing angle variation.

Deep Learning-Based Service Discovery for Business Process Re-Engineering in the Era of Big Data

In recent years, business process re-engineering has played an important role in the development of large-scale web-based applications. To re-engineer business processes, business services are developed and coordinated by reusing a set of open APIs and services on the internet. Yet, the number of services on the internet has grown drastically, making it difficult for them to be discovered to support the changing business goals. One major challenge is therefore to search for a suitable service that matches a specific business goal from a large number of available services in an efficient and effective manner. To address this challenge, this paper proposes a deep learning approach for massive service discovery. The approach, thus called MassRAFF, employs a combination of the recurrent attention and feature fusion methods. This paper first presents the MassRAFF approach and then reports on an experiment for evaluating this approach. The experimental results show that the MassRAFF approach has performed reasonably well and has potential to be improved further in future work.

An Application of an Intelligent Data Warehouse for Modelling Spatiotemporal Objects

Data warehouse (DW) systems provide the best solution for intelligent data analysis and decision-making. Changes applied to data gradually in real life have to be projected to the DW. Slowly changing dimension (SCD) refers to the potential volatility of DW dimension members. The treatment of SCDs has a significant impact over the quality of data analysis. A new SCD type, Type N, is proposed in this research paper, which encapsulates volatile data into historical clusters. Type N preserves complete history of changes, additional tables, columns, and rows are not required, extra join operations are omitted, and surrogate keys are avoided. Type N is implemented and compared to other SCD types. Good candidates for practicing SCDs are spatiotemporal objects (i.e., objects whose shape or geometry evolves slowly over time). The case study used and implemented in this paper concerns shape-shifting constructions (i.e., buildings that respond to changing weather conditions or the way people use them). The results demonstrate the correctness and effectiveness of the proposed SCD Type N.

Detecting Safe Routes During Floods Using Deep Learning

Floods are one of the most devastating and frequently occurring natural disasters throughout the world. Floods can cause blockage of roads and hence create trouble for civilians and authorities to navigate in the flooded area. This paper proposes an automated system that uses a road extraction algorithm to extract roads from satellite images to create a highlighted map of all the available roads during floods. The road extraction algorithm the authors developed uses U-net model architecture, a fully convolutional neural network, to extract roads from aerial images (satellite images and drone images). Convolutional Neural Network is robust to shadows and water streams, able to obtain the characteristics of roads adequately and most importantly, able to produce output quickly, which is necessary for flood evacuations and relief. The developed system can be deployed as an Application Programming Interface or stand-alone system, loaded on drones, which will provide the users with a map highlighting safe paths to traverse the flooded areas.