Vehicle Incident Hot Spots Identification: An Approach for Big Data

Abstract

A note on versions:The version presented here may differ from the published version or from the version of record. If you wish to cite this item you are advised to consult the publisher's version. Please see the repository url above for details on accessing the published version and note that access may require a subscription. Abstract-In this work we introduce a fast big data approach for road incident hot spot identification using Apache Spark. We implement an existing immuno-inspired mechanism, namely SeleSup, as a series of MapReduce-like operations. SeleSup is composed of a number of iterations that remove data redundancies and result in the detection of areas of high likelihood of vehicles incidents. It has been successfully applied to large datasets, however, as the size of the data increases to millions of instances, its performance drops significantly. Our objective therefore is to re-conceptualise the method for big data. In this paper we present the new implementation, the challenges faced when converting the method for the Apache Spark platform as well as the outcomes obtained. For our experiments we employ a large dataset containing hundreds of thousands of Heavy Good Vehicles incidents, collected via telematics. Results show a significant improvement in performance with no detriment to the accuracy of the method.

Fuzzy Hot Spot Identification for Big Data: An Initial Approach

“…An increase in the value of the confidence threshold tends to lead to an increase in the number of identified hot spots, as incidents require a higher degree of membership to be reduced by existing hot spots, and are therefore more likely to form new hot spots instead. A similar effect was observed in [6] following an increase in the defined mileage range. The difference in the fuzzy version is that within the set of identified hot spots, a variety of different radiuses will be present, depending on the distribution of incidents.…”

“…In this subsection we provide a high-level description of the streaming SeleSup-HSID algorithm, based on the earlier works on SeleSup-HSID in [5] and [6], and used here as the method into which we introduce fuzzy concepts. A full description can be found in [8] and our implementation is available on GitHub 1 (using Apache Spark Streaming [14]).…”

“…In recent work, we have proposed the use of instance selection techniques, usually employed to pre-process data prior to data mining tasks [4], for the task of HSID. In [5] and [6], an immune-inspired instance selection technique [7] was applied to find hot spots in large telematics datasets. Experimental results showed the success of this method (SeleSup-HSID) as a HSID approach.…”

A graph CNN-LSTM neural network for short and long-term traffic forecasting based on trajectory data

“…It is focused on investigating the dynamic traffic relationships of different locations. This is achieved by characterizing similar traffic measure values from one road to another and then grouping the locations in clusters that divide the road network into correlated groups (Triguero et al, 2017). This is possible due to traffic's correlations in the temporal and spatial domain.…”

Identifying Heavy Goods Vehicle Driving Styles in the United Kingdom

“…Statistics for the UK between October 2014 and September 2015 show that HGVs delivered 1.63 billion tonnes of freight within the UK, and 8.5 million tonnes were imported and exported [1]. Due to the importance of HGVs in the country's economy, there are great efforts being employed to reduce their incident numbers [2], [3]. Such incidents, which incur in significant losses, are caused by the characteristics of the vehicle, road, weather conditions, defiance of traffic rules, company policies and mostly by driving behaviour.…”