October 1, 2021 Thesis Open Access

HABTU REDA

Citation Style Language JSON Export

{
  "DOI": "10.20372/nadre:4654", 
  "author": [
    {
      "family": "HABTU REDA"
    }
  ], 
  "issued": {
    "date-parts": [
      [
        2021, 
        10, 
        1
      ]
    ]
  }, 
  "abstract": "<p>Estimating public bus arrival times and delivering accurate arrival time information to<br>\npassengers are critical for making public transportation more user-friendly and thereby<br>\nincreasing its competitiveness among various forms of transportation. However public bus<br>\narrival time prediction remains major bottlenecks With traffic heterogeneity in composition and<br>\ndiversity of vehicles, as well as a big pedestrian population combined with inadequate lane use,<br>\npredicting the arrival time of public buses at stations is a severe concern.. The main objective of<br>\nthis study is to apply machine learning algorithms to predict bus arrival time. The data was<br>\ncollected from Addis Ababa Sheger Public Bus Transport. Random Forest, Gradient Boosting,<br>\nArtificial Neural Network, K-Nearest Neighbors and Support Vector Machine algorithms are<br>\napplied to build the models and to compare and choose the best model to predict the bus arrival<br>\ntime. After selecting the features and algorithms, different data preprocessing tasks like checking<br>\noutliers, missing values and data reduction are done. Finally, 140,000 instances of dataset are<br>\nused to train and build the model. The prepared dataset is partitioned into 90% training and 10%<br>\ntesting set. Beginning Date, Beginning Time, End Date, Time Range, Mileage, Duration, Initial<br>\nlatitude, Initial longitude, Final latitude, Final longitude, and End Time were used as input<br>\nfeatures for developing the model. Based on the experiment result the Random Forest algorithm<br>\nachieved a better performance with R-squared score of 0.994, MAE of 0.812, RMSE of 3.780<br>\nand MSE of 14.28.</p>", 
  "title": "PUBLIC BUS ARRIVAL TIME PREDICTION USING MACHINE LEARNING: IN CASE OF ADDIS ABABA", 
  "type": "thesis", 
  "id": "4654"
}