Real-time Big Data Analytics Framework with Data Blending Approach for Multiple Data sources in Smart City Applications

Abstract

Advancement in Information Communication Technology (ICT) and the Internet of Things (IoT) has to lead tothe continuous generation of a large amount of data. Smart city projects are being implemented in various parts of the world where analysis of public data helps in providing a better quality of life. Data analytics plays a vital role in many such data-driven applications. Real-time analytics for finding valuable insights at the right time using smart city data is crucial in making appropriate decisions for city administration. It is essential to use multiple data sources as input for the analysis to achieve better and more accurate data-driven solutions. It helps in finding more accurate solutions and making appropriate decisions. Public safety is one of the major concerns in any smart city project in which real-time analytics is much useful in the early detection of valuable data patterns. It is crucial to find early predictions of crime-related incidents and generating emergency alerts for making appropriate decisions to provide security to the people and safety of the city infrastructure. This paper discusses the proposed real-time big data analytics framework with data blending approach using multiple data sources for smart city applications. Analytics using multiple data sources for a specific data-driven solution helps in finding more data patterns, which in turn increases the accuracy of analytics results. The data preprocessing phase is a challenging task in data analytics when data being ingested continuously in real-time into the analytics system. The proposed system helps in the preprocessing of real-time data with data blending of multiple data sources used in the analytics. The proposed framework is beneficial when data from multiple sources are ingested in real-time as input data and is also flexible to use any additional data source of interest. The experimental work carried out with the proposed framework using multiple data sources to find the crime-related insights in real-time helps the public safety solutions in the smart city. The experimental outcome shows that there is a significant increase in the number of identified useful data patterns as the number of data sources increases. A real-time based emergency alert system to help the public safety solution is implementedusing a machine learning-based classification algorithm with the proposed framework. The experiment is carried out with different classification algorithms, and the results show that Naive Bayes classification performs better in generating emergency alerts.

Leveraging transfer learning with deep learning for crime prediction

“…Manjunatha and Annappa [ 41 ] studied higher crime rates in cities using a predictive method based on spatial analysis and autoregressive models, highlighting the hazardous crime location. For the trial of this approach, two real-world datasets were collected in the cities of New York and Chicago, and the results demonstrate good precision in spatial and temporal crime prediction in each region.…”