1 data description
./data
----  / jz_agg			CDR data aggregation ( results of _01_AggCDR data in base station.zip )
----------  avg.csv			Average by sum.csv and count.csv ( avg = sum / count)
----------  count.csv			Count by different days of a week
----------  sum.csv			Summarize by different days of a week
----  JZpoint&gsm_Theissen.csv		CDR data aggregated in base station
----  su_jz for cluster2.csv		CDR data aggregated in streetunit, used for DCAE training and hidden feature extraction
----  su_gsm2016_avg_forCls(test).csv	CDR data used for DCAE training test


2 required dependent libraries
Python==3.5
Keras==2.2.0 
Pytorch==0.4.1
Cuda==8.0
Scikit_learn==0.15

3 Trained model parameters
./models
----  *.h5				Trained DCAE model with different parameters


4 Results in this manuscript
./results
----  figure*			Figures of data visualization used in manuscript
./ outClassification result_1.csv		Experimental result with different parameters (via methods)
./ outClassification result_2.csv		Experimental result with different parameters (sensitivity analysis)


5. Code
./ _01_AggCDR data in base station.zip	Aggregating CDR data in base station (C# code in vs2015)
./ _02_plotting_for_two_norms.py	Visualization of two normalizations (python)
./ _03_vae_f1f2.py			DCAE model, training & hidden feature extraction & kmeans cluster (python)
./ _03_*.py				Sensitivity analysis models of DVAE, and methods for comparison (python)
./ _03_lstm_ae.zip			LSTM+AE model (python)
./ _04_autoencoder_result_forAll.py	Load DCAE trained model,  hidden feature extraction & kmeans cluster (python)
./_05_DCAE_with_noise.zip The classification results with different level additive noise (0.2, 0.5 and 0.8 RSM noise) have been present in the Table VI.
./_06_RNN_LSTM_source_code.zip             LSTM_results.