mrap 1.0.0

Machine-Readable Data Analysis Results with Function Wrappers

mrap

100% AI-free: we did not use any AI technologies in developing this package.

The goal of mrap is to provide wrapper functions to reduce the user's effort in writing machine-readable data with the dtreg library. All-in-one wrappers (such as scipy_f_oneway) will cover functions from scipy and other widely used libraries. The mrap.analytic_instances module contains wrappers for analytical schemata used by TIB Knowledge Loom.

To write the results of your data analysis as JSON-LD:

• Select a wrapper from mrap, check the types of arguments it requires, and create an instance.
• Modify the instance by setting or correcting its fields manually.
• Write the finalised instance as a machine-readable JSON-LD file.

Installation

## activate a virtual environment (optional)
python3 -m venv .venv
source .venv/bin/activate
## install from PyPi:
pip install mrap

Example

Let us assume you intend to report the classification performance of your_fancy_algorithm. From the help page you know that the wrapper of choice is algorithm_evaluation(). The required arguments are:

• code_list: a string "N/A" if not given or a list of two strings - the library and the line of code used for implementing the analysis.
• input_data: a pd.DataFrame, a dictionary of pd.Series with names, or a string which is either the data URL or the file name.
• dictionary_results: a dictionary with metrics names as keys.

The wrapper writes information about the data, your results, your Python version, the library version, etc. You can add or correct any information in relevant fields.

from mrap.analytic_instances import algorithm_evaluation
## run your analysis and write the results as a dictionary
results_dict = {"F1": 0.46, "recall": 0.51}
## create your instance with the wrapper
eval_instance = algorithm_evaluation(["the_library", "the_line_of_code"], 
                                   "my_URL", 
                                   results_dict)
## add information about the algorithm and the task                                   
eval_instance.evaluates = "my_fancy_algorithm"
eval_instance.evaluates_for = "Classification"

It is also possible to write the results of a few algorithms evaluated on the same data for the same task.

from mrap.list_algorithm_evaluations import list_algorithm_evaluations
## run your analysis and write a nested dictionary
results_nested = {"ABC": {"F1": 0.46, "recall": 0.64},
             "KLM": {"F1": 0.55, "recall": 0.38},
             "XYZ": {"F1": 0.87, "recall": 0.78}}
## create the list of instances
eval_instance = list_algorithm_evaluations("N/A", "my URL", "Classification", my_results)
## when manually modifying the instance, select an element of the list
eval_instance[2].label = "XYZ evaluation"

You can include the evaluation instance in data analysis instance, convert it into a machine-readable JSON-LD string, and write as a file.

from mrap.analytic_instances import data_analysis
from mrap.to_jsonld import to_jsonld
da_instance = data_analysis(eval_instance)
da_instance.is_implemented_by = "code_URL"
my_json = to_jsonld(da_instance)
with open('data_analysis.json', 'w') as f:
    f.write(my_json)

For more information, please see the help page.