Comparing Machine Learning Algorithms for Vehicle Route Prediction
View/ Open
Date
2023-06-22Author
Pudu, Prithvidhar
0000-0001-9671-9332
Metadata
Show full item recordAbstract
Route prediction is an important tool in understanding the trajectories of vehicles.
This information can be used to make intelligent decisions in city planning such as
traffic avoidance and pedestrian safety. Machine Learning can predict these routes by
learning common patterns from previous vehicle paths. These vehicle paths are plentiful
due to the abundance of IoT devices and advancements in computer vision. The most
suitable machine learning algorithm that could accurately predict destinations was
determined by passing vehicle data to three popular machine learning algorithms: Decision
Trees, Artificial Neural Networks, and Naïve-Bayes. The training data for these algorithms
was converted into a suitable format using the Pandas python library. Moreover, the hyperparameters for each algorithm were tuned to maximize the accuracy of the prediction.
Respectively, decision trees were observed to provide the highest accuracy of 99.7% with
the Naïve-Bayes and ANN showing 99.5% and 98%