Boston Train Station's Passenger Flow Prediction

Using Genetic Algorithm to optimize BP Neural Network's initial weight and threshold value.

Two kind prediction: Whole Day prediction and Time Period Prediction.

Description

• Data Pre-Processing.ipynb: Generating input and output csv files

• MBTA_MATLAB Folder: Running in MATLAB to do prediction

Tools

• Neural Network Training Toolbox

• Sheffield Genetic Algorithms Toolbox

• MatLab

• Python Jupyter

• Numpy / Pandas

• Matplot

• Sklearn

Data

• MBTA historical passenger flow data, available in MBTA.

• Weather (Rain, Snow, Temperature) historical data, has been uploaded in current folder.

• US Holiday Data, Week, Month...

Input

Output

Design Structure

• Setting up BP Neural Network structure

• Using GA optimize initial weight and threshold

• Training and Testing BP Neural Network

BP Neural Network Structure

Genetic Algorithm Parameters

Outcomes Display

BP NN Prediction without GA && BP NN Prediction with GA

BP NN & BP NN + GA Comparison

Prediction Error - BP NN without GA

Prediction Error - BP NN with GA

Evaluation / Performance / Training State / Error Histogram