Parking Spaces Prediction and Dynamic Programming of Taipei city

Introduction

This is a website that can get your location (Only for Taipei City) and predict the number of remaining parking spaces in the nearby parking lot in the next 30 minutes, and calculates the best route to each parking lot using the Google API, reducing the time for waiting for parking spaces.

Demonstration

Require

• python>=3
• Apache
• Two Server (For deploy the website and the prediction model)
• Tensorflow<2
• keras 2.3.1

Two Flasks

Start Up

1. Change the sys.path in website/app.wsgi to your own Apache website path
   
2. Change line 37 in webiste/server.py to your prediction model server address and port
   
3. Launch the website by Apache
   
4. Change line 86 in parking_map/map.py to the same port as the website/server.py
   
5. Launch the prediction model
   

python ./parking_map/map.py

6. Now you can open the website for checking the nearby parking spaces
   

Train your own LSTM model

1. Unzip model_training/data/parking_data_all_V3.csv.zip and model_training/data/Google_api_V1.csv.zip

unzip ./model_training/data/parking_data_all_V3.csv.zip
unzip ./model_training/data/Google_api_V1.csv.zip

2. Run model_training/model/preprocessing.py for preprocessing csv data to pickle

python ./model_training/model/preprocessing.py

3. Train LSTM model

python ./model_training/model/LSTM.py

4. Move the new LSTM model to prediction model folder

mv ./model_training/model/lstm_model.h5 ./parking_map/lstm_model.h5