.png){kind=icon}
.png){kind=icon}
Return
A parallel grid-search-based SVM optimization algorithm on Spark for passenger hotspot prediction
.png){kind=icon}
PRE
AI
DOI：10.1007/s11042-022-12077-x.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Predicting passenger hotspots helps drivers quickly pick up travelers, reduces cruise expenses, and maximizes revenue per unit time in intelligent transportation systems. To improve the accuracy and robustness of passenger hotspot prediction (PHP), this paper proposes a parallel Grid-Search-based Support Vector Machine (GS-SVM) optimization algorithm on Spark, which provides an efficient methodology to search for passengers in a complex urban traffic network quickly. Specifically, to effectively locate passenger hotspots, an urban road network is gridded on the Spark parallel distributed computing platform. Moreover, to enhance the accuracy of PHP, the grid search (GS) approach is employed to optimize the radial basis function (RBF) of the support vector machine (SVM), and the cross-validation method is utilized to find out the global optimal parameter combination. Finally, the SVM optimization algorithm is implemented on Spark to improve the robustness of PHP. In particular, the proposed GS-SVM algorithm is applied to successfully predict passenger hotspots. By analyzing seven groups of data sets and comparing with serval state-of-the-art algorithms including autoregressive integrated moving average (ARIMA), support vector regression (SVR), long short-term memory (LSTM), and convolutional neural network (CNN), the results of an empirical study indicate that the MAPE value of our GS-SVM algorithm is lower than that of comparative algorithms at least 78.4%.
Keywords:
Visual analytics
Data visualization
Passenger hotspot prediction
Parallel SVM optimization algorithm
Spark
Journal
IF:
3
Papers:
1.9W
Citations:
3.2W
View more organizations