README.md

• This repository is for paper "The Best of Both Worlds: On Repairing Timestamps and Attribute Values for Multivariate Time Series".

File Structure

• code: source code of algorithms.
• data: dataset source files used in experiments.
• full.pdf: the full version of our manuscript, including all the complete proofs in appendix.

Dataset

• Gas: https://archive.ics.uci.edu/dataset/322/gas+sensor+array+under+dynamic+gas+mixtures
  • Manual labeling of ground truth
• Human: https://www.kaggle.com/datasets/die9origephit/human-activity-recognition?select=time_series_data_human_activities.csv
• GINS: https://github.com/i2Nav-WHU/awesome-gins-datasets
• Hum: https://www.kaggle.com/datasets/laurendobratz/iottown?select=Temp-Hum-Sensor-B.csv
  • Manual labeling of ground truth
• Lab: http://db.csail.mit.edu/labdata/labdata.html
• PMEL: http://www.pmel.noaa.gov/tao/
• PMES: https://archive.ics.uci.edu/dataset/204/pems+sf

Dependencies

java_random==1.0
jgrapht==1.5.0.3
matplotlib==3.8.4
numpy==1.23.5
pandas==1.4.2
scikit_learn==1.0.2

Instruction

To run the program, use the following command-line arguments:

• --dataname: Input file name. This parameter specifies the name of the dataset to be processed.

• --alg: Method name. This parameter allows you to select the algorithm to use.

• --w: Window size. This parameter controls the size of the window used during data processing.

• --alpha: Attribute repair cost. This parameter sets the cost associated with repairing attributes.

• --beta: Timestamp repair cost. This parameter sets the cost associated with repairing timestamps.

• --t_dratio: Timestamp noise rate. This parameter controls the proportion of noise in the timestamps.

• --v_dratio: Attribute noise rate. This parameter controls the proportion of noise in the attributes.

• --seed: Random seed.

• --s: Speed constraint. This parameter sets the constraints on speed, allowing for multiple speed values to be specified.

You can run the program using the following command line example:

main.py --dataname Lab --alg TAVLC --w 50

or you can run directly:

cd code
python main.py