Urban Computing ToolBox

Introduction

Urban Computing ToolBox is a package providing spatial-temporal predicting models. It contains both conventional models and state-of-art models.

Currently the package supported the following models: (Details)

• HM
• HMM
• ARIMA
• XGBoost
• DeepST
• ST-ResNet
• DCRNN
• AMulti-GCLSTM

UCTB is a flexible and open package. You can use the data we provided or use your own data, the data structure is well stated in the tutorial chapter. You can build your own model based on model-units we provided and use the model-training class to train the model.

Installation

Step 1 : Install TensorFlow

You can skip to step 2 if you already installed tensorflow.

You can refer to this page https://www.tensorflow.org/install to install tensorflow, if you have a Nvidia GPU installed on you computer, we highly recommend you to install GPU version of tensorflow.

Step 2 : Install UCTB

pip install --upgrade UCTB

The following required package will be installed or upgraded with UCTB:

xxxxxxxxxx
'hmmlearn',
'keras',
'GPUtil',
'numpy',
'pandas',
'python-dateutil',
'scikit-learn',
'scipy',
'statsmodels',
'wget',
'xgboost',
'nni',
'chinesecalendar',
'PyYAML'

UCTB Docker

You can also use UCTB by docker. First pull uctb docker from docker hub.

xxxxxxxxxx
docker pull dichai/uctb:v0.2.0

And you then can run it.

xxxxxxxxxx
docker run  --runtime=nvidia  -it -d dichai/uctb:v0.2.0 /bin/bash

Quick start

• Quick start with AMulti-GCLSTM

​x
from UCTB.dataset import NodeTrafficLoader
from UCTB.model import AMultiGCLSTM
from UCTB.evaluation import metric
​
# Config data loader
data_loader = NodeTrafficLoader(dataset='Bike', city='NYC', graph='Correlation',
                                closeness_len=6, period_len=7, trend_len=4, normalize=True)
​
# Init model object
AMulti_GCLSTM_Obj = AMultiGCLSTM(closeness_len=data_loader.closeness_len,
                                 period_len=data_loader.period_len,
                                 trend_len=data_loader.trend_len,
                                 num_node=data_loader.station_number,
                                 num_graph=data_loader.LM.shape[0])
​
# Build tf-graph
AMulti_GCLSTM_Obj.build()
# Training
AMulti_GCLSTM_Obj.fit(closeness_feature=data_loader.train_closeness,
                      period_feature=data_loader.train_period,
                      trend_feature=data_loader.train_trend,
                      laplace_matrix=data_loader.LM,
                      target=data_loader.train_y,
                      sequence_length=data_loader.train_sequence_len)
​
# Predict
prediction = AMulti_GCLSTM_Obj.predict(closeness_feature=data_loader.test_closeness,
                                       period_feature=data_loader.test_period,
                                       trend_feature=data_loader.test_trend,
                                       laplace_matrix=data_loader.LM,
                                       target=data_loader.test_y,
                                       output_names=['prediction'],
                                       sequence_length=data_loader.test_sequence_len)
​
# Evaluate
print('Test result', metric.rmse(prediction=data_loader.normalizer.min_max_denormal(prediction['prediction']),
                                 target=data_loader.normalizer.min_max_denormal(data_loader.test_y), threshold=0))

• Quick start with other models

Tutorial

• Compare different version of AMulti-GCLSTM

• Use datasets from UCTB

• Build your own datasets

• Use build-in models from UCTB

• Build your own model using UCTB

API Reference

• Model Class

• ModelUnit Class

• Evaluation Class

• Training Class

Full examples

• Experiments on bike traffic-flow prediction

• Experiments on metro traffic-flow prediction

• Experiments on didi order demand prediction

• Experiments on charge station demand prediction

Contribute to UCTB (Delay)

• Contribute Data

• Contribute Models

• Contribute Model-Units