idsw 1.4.1

Full workflow for ETL, statistics, and Machine learning modelling of (usually) time-stamped industrial facilities data.

Industrial Data Science Workflow

Industrial Data Science Workflow: full workflow for ETL, statistics, and Machine learning modelling of (usually) time-stamped industrial facilities data.

Not only applicable to monitoring quality and industrial facilities systems, the package can be applied to data manipulation, characterization and modelling of different numeric and categorical datasets to boost your work and replace tradicional tools like SAS, Minitab and Statistica software.

• Check the project Github: https://github.com/marcosoares-92/IndustrialDataScienceWorkflow

• Check our Steel Industry Simulator on: https://github.com/marcosoares-92/steelindustrysimulator

  • The Ideal Tool for Process Improvement, and Data Collection, Analyzing and Modelling Training.
  • User interface available in:

  https://colab.research.google.com/github/marcosoares-92/steelindustrysimulator/blob/main/steelindustry_digitaltwin.ipynb

• Check our Model Project using Industrial Data Science Workflow in Crop Science on: https://github.com/marcosoares-92/crop_model_proj_case_soybean_uci

Authors:

• Marco Cesar Prado Soares, Data Scientist and Data Engineer Specialist; Data Coordinator

  • [e-mail]

• Gabriel Fernandes Luz, Senior Data Scientist

  • [e-mail]

• Ivan Alexander Yapita Vargas, Senior Data Analyst

  • [e-mail]

• Sergio Guilherme Neto, Senior Data Analyst

  • [e-mail]

• If you cannot install the last version from idsw package directly from PyPI using pip install idsw:

1. Open the terminal and:

Run:

git clone "https://github.com/marcosoares-92/IndustrialDataScienceWorkflow" 

to clone all the files (you could also fork them).

2. Go to the directory called idsw.
3. Now, open the Python terminal and:

Navigate to the idsw folder to run:

pip install .

• You can use command cd "...\idsw", providing the full idsw path to navigate to it. Alternatively, run pip install ".\*.tar.gz" in the folder terminal.

After cloning the directory, you can also run the package without installing it:

1. Copy the whole idsw folder to the working directory where your python or jupyter notebook file is saved.

• There must be an idsw folder on the python file directory.

2. In your Python file:

Run the command or run a cell (Jupyter notebook) with:

from idsw import *

for importing all idsw functions without the alias idsw; or:

import idsw

to import the package with the alias idsw.

Alternatively, if you do not want to clone the repository, you may download the file load.py and copy it to the working directory.

1. After downloading load.py and copying it to the working directory, in your Python environment, run:

   import load

2. After conclusion of this step, you may import the package as:

   from idsw import *

or as:

import idsw

The load.py file runs the following code, which may be copied to your Python environment and run:

class LoadIDSW:

  def __init__(self, timeout = 60):  
    self.cmd_line1 = """git clone https://github.com/marcosoares-92/IndustrialDataScienceWorkflow IndustrialDataScienceWorkflow"""
    self.msg1 = "Cloning IndustrialDataScienceWorkflow to working directory."
    self.cmd_line2 = """mv IndustrialDataScienceWorkflow/idsw ."""
    self.msg2 = "Subdirectory 'idsw' moved to root directory. Now it can be directly imported."
    self.timeout = timeout

  def set_process (self, cmd_line):
    from subprocess import Popen, PIPE, TimeoutExpired
    proc = Popen(cmd_line.split(" "), stdout = PIPE, stderr = PIPE)
    return proc

  def run_process (self, proc, msg = ''):
    try:
        output, error = proc.communicate(timeout = self.timeout)
        if len(msg > 0):
          print (msg)
    except:
        output, error = proc.communicate()       
    return output, error

  def clone_repo(self):
    self.proc1 = self.set_process (self.cmd_line1)
    self.output1, self.error1 = self.run_process(self.proc1, self.msg1)
    return self

  def move_pkg(self):
    self.proc2 = self.set_process (self.cmd_line2)
    self.output2, self.error2 = self.run_process(self.proc2, self.msg2)
    return self

  def move_pkg_alternative(self):
    import shutil
    source = 'IndustrialDataScienceWorkflow/idsw'  
    destination = '.'
    dest = shutil.move(source, destination)    
    return self

loader = LoadIDSW(timeout = 60)
loader = loader.clone_repo()
loader = loader.move_pkg()

try:
  from idsw import *
except ModuleNotFoundError:
  loader = loader.move_pkg_alternative()

msg = """Package copied to the working directory.
	To import its whole content, run:
	
	    from idsw import *
	"""
print(msg)

History

1.2.0

Fixed

• Deprecated structures

Added

• New functionalities added.

Reshape of project design.

• New division into modules and new names for functions and classes.

Removed

• Removed support for Python < 3.7

1.2.1

Fixed

• Setup issues.

1.2.2

Fixed

• Setup issues: need for rigid and specific versions of the libraries.

1.2.3

Fixed

• Setup issues.

1.2.4

Fixed

• Import bugs.
• Introduced function for Excel writing.

1.2.5

Fixed

• Matplotlib export figures bugs.

• 'quality' argument is no longer supported by plt.savefig function (Matplotlib), so it was removed.

• This modification was needed for allowing the correct functioning of the steelindustrysimulator, which is based on idsw.

• Check simulator project on: https://github.com/marcosoares-92/steelindustrysimulator

  • The Ideal Tool for Process Improvement, and Data Collection, Analyzing and Modelling Training.

1.2.6

Fixed

• Export of figures generated a message like with '{new_file_path}.png.png'. Fixed to '{new_file_path}.png'.

1.3.0

Added

• New functionalities added.

Reshape of project design.

• New division of functions and classes and correspondent modules.
• Refactoring of functions and classes to improve code efficiency.
• Added new pipelines for fetching data and modified the storage of connectors.
• It includes pipelines for fetching table regions in Excel files, even if they are stored in a same tab; and a pipeline for downloading files stored in MS SharePoint.
• Added ControlVars dataclass to store if the user wants to hide results and plots.

1.3.1

Improved

• Benford algorith for fraud and outlier detection.
• Pipeline for fetching SharePoint and downloading files.

1.4

• Module datafetch.texts added: functions based on LangChain for extracting texts from PDFs, DOCX, CSV, HTML and TXT and creating a text database.
• Several bugs fixed, specially in functions based on Pandas and NumPy deprecated structures.

1.4.1

• Bug fixed in linear regression function.
• Bug fixed: Python ~ operator deprecated, and replaced by "not".