| 摘要 |
The present invention relates to a prediction method for the congestion of an urban area using a nerve network algorithm including a step of selecting an input variable and collecting data on the input variable; a step of preprocessing the collected data; a step of determining the learning conditions and the learning termination conditions of the preprocessed data; a step of determining the number and the weight of a hidden unit; and a step of verifying the accuracy of prediction by comparing an actually measured average speed with a predicted speed which is an output variable. [Reference numerals] (AA) Start; (BB) End; (S41) Select an input variable for a congestion prediction method and collect data on the input variable; (S42) Preprocess information about the input variable; (S43) Determine low data learning conditions and learning termination conditions to build a neural network model; (S44) Determine the number and the weight of a hidden unit; (S45) Check a verification result of the neural network model through a test; (S46) Verify the accuracy by comparing an actually measured average speed with a predicted speed |
