Hybrid artificial bee colony based neural network and dynamic threshold technique for predicting moving vehicle location and co-located objects

• Published in: Sadhana (Bangalore) 2023-04, Vol.48 (2), Article 68
• Main Authors: Baby Anitha, E, Sivaprakash, S, Velmurugan, S, Saranya, S S
• Format: Article
• Language: English
• Subjects: Engineering
• ISSN: 0973-7677; 0973-7677
• DOI: 10.1007/s12046-023-02128-w

Co-location pattern discovery contains processing data as well as a spatial context to observe the classes of spatial objects that are often located together. Moreover, the existing moving vehicle location prediction methodology doesn't examine the vehicles movement co-location instances. So, improving the previous method process by mining spatially co-located moving objects utilize spatial data mining (SDM) methods. To improve this, Hybrid Artificial Bee Colony based Feed Forward Back propagation Neural Network for Predicting Moving Vehicle future Location and Co-Located Objects (Hyb-ABC-FFBPNN-PMVFL-CLO) is proposed. Initially, the data comes from the Honda Egocentric View-Intersection (HEV-I) dataset. In order to provide the future location of the vehicle, it is necessary to predict the driving trajectory of vehicles traveling on the road in certain segments. For predicting moving vehicles future location, FFBPNN is used. Generally, FFBPNN does not reveal any adoption of optimization techniques for computing optimal parameters to ensure accurate prediction. Therefore, in this work, proposed an Artificial Bee Colony (ABC) algorithm is utilized for optimizing the weight parameters of Feed Forward Back propagation Neural Network (FFBPNN). Also, Co-location Pattern Mining with Dynamic Threshold Technique (CPMDTT) is proposed to predict co-located objects with spatial data mining technique. In this, an undirected graph is constructed using the information on predicted future location of vehicles. Finally, the proposed method is implemented in MATLAB. The evaluation metrics such as accuracy, number of Co locations, and run time are analyzed. Then, the proposed approach provide 14.68%, 7.142%, and 4.65% higher accuracy, 38.18%, 12.02%, and 7.59% lower running time compared with existing methods, like Ego-centric vision-based future vehicle localization for intelligent driving assistance systems, Performance of design options in automated Auto-Regressive Integrated Moving Average method construction for dynamic vehicle GPS location prediction and a deep learning method for next location prediction respectively.