# Causal-Discovery-from-Continuous-Time-Series-by-Compression
## 1. Code Summary
In data directory, we provide all data that we use in our experiments and a data generator of MATLAB Simulink to generate mechatronic engineering data.
In cute.py, there are methods provided by [Causal Inference on Event Sequences](http://kailashbuki.github.io/manuscript/cute.pdf) to allow our methods to compare with others.
In granger.py, we provide Granger Causality test method.
In Util.py, we provide all functions that will be used in DISC and our experiments.
In Disc.py, core methods of DISC are provided.
synthetic_data_test.py and real_data_test.py provide synthetic data test and real world data test results separately.
## 2. Time Complexity
**synthetic_data_test.py**
time_window()
time_weighted()
time_weighted_window()
## 3. Validation for Encoding Method
**synthetic_data_test.py**
test_causality_consistency()
test_no_causality_consistency()
## 4. Synthetic Experimental Results
#### 4.1 Simple Linear Causality
**synthetic_data_test.py**
test_linear_data()
#### 4.2 Nonlinear Monotonic Causality
**synthetic_data_test.py**
test_non_linear()
## 5. EMPIRICAL RESULTS ON REAL WORLD DATA
#### 5.1 River Data
**real_data_test.py**
river_test()
river_test2()
#### 5.2 Temperature and Ozone Data
ozone_test(1)
ozone_test(2)
ozone_test(3)
#### 5.3 Mechatronic Engineering Data
test_engineering(1, 0)
test_engineering(2, 0)
test_engineering(1, 1)
test_engineering(2, 1)
## Supplement to our code
The methods above only provide main results of our experiments. The code for other models(G-test, cute, icute) is annotated. These functions can also be used.
From the point of view of code simplicity, all of our synthetic data related tests are imported from test_data(). Make sure to read comments before you use this method by yourself.
## Experiment Results
We provide detailed results of all the experiments in this paper in directory **Experiment Result**.
In addition, as the noise is random, the last real world data experiment will get different results every time you run. Therefore we store the data that we use in this paper in excel of the directory.
matlab导入excel代码-TimeSeries:时间序列
需积分: 10 158 浏览量
2021-05-21
09:54:58
上传
评论
收藏 7.5MB ZIP 举报
TimeSeries-master.zip （82个子文件） 
TimeSeries-master 
Util.py 16KB dy_v.eps 32KB value1 21KB complex.eps 136KB normal3 7KB synthetic_data_test.py 49KB 2013.csv 1.5MB data 
ozone2.txt 7KB
ozone1.xlsx 15KB engineering_data_generation optimal_control.m 763B Readme.txt 563B slope_position.slx 23KB torque.slx 24KB adaptive_control.m 282B base_mix.m 2KB mix.slx 30KB torque_discrete.slx 24KB position.slx 24KB position_discrete.slx 24KB rhein.dat 22KB ozone3.txt 7KB ozone2.xlsx 15KB saar.csv 82KB temperature.txt 3KB saar.dat 17KB rhein.csv 93KB TrapezoidalWave.xlsx 323KB ozone3.xlsx 15KB ozone1.txt 7KB value5 19KB data_generation.py 2KB normal1 7KB
price.xls 373KB cute.py 4KB Experiment Result Synthetic Data non-linear synthetic data(tanh).xlsx 9KB linear synthetic data.xlsx 9KB Validation for Encoding Method CausalityNormalDistributionImprovedCUTE.xlsx 16KB NoCausalityNormalDistributionDISC.xlsx 16KB CausalityGMMImprovedCUTE.xlsx 16KB NoCausalityGMMOurDISC.xlsx 16KB non-linear synthetic data(ex).xlsx 9KB Real World Data Trapezoidal wave with noise output2.xlsx 290KB iCUTE.txt 1KB Granger.txt 1KB CUTE.txt 1KB Trapezoidal wave with noise output1.xlsx 290KB DISC.txt 1KB value2 21KB Soybean.xlsx 17KB tmp_causes2 86KB tmp_effects 42KB values2 0B LICENSE 11KB normal5 6KB tmp_causes 42KB cute.pyc 6KB dy.eps 177KB 2012.csv 1.5MB values 41KB value4 23KB 2013.xlsx 1.48MB Corn.xlsx 18KB normal2 8KB cause.eps 185KB data_generation.pyc 2KB Util.pyc 15KB value.eps 29KB real_data_test.py 218KB 2012.xlsx 1.48MB synthetic_data_test.pyc 29KB Addtional.py 22KB README.md 2KB value3 21KB Disc.pyc 7KB real_data_test.pyc 137KB normal4 6KB granger_test.py 818B 2010.csv 1.45MB granger_test.pyc 972B 2010.xlsx 1.38MB Disc.py 9KB tmp_effects2 85KB资源评论
