ERTool 0.2.6

A Python Package for Efficient Implementation of the Evidential Reasoning Approach for Multi-Source Evidence Fusion

ERTool

ERTool is a Python Package for Efficient Implementation of Multi-source Evidence Fusion Based on the Evidential Reasoning Approach. It aims to provide an intuitive and flexible approach for integrating ER processes, particularly suitable for data analysis and decision support systems.

Features

• Easy-to-use implementation of Evidential Reasoning.
• Efficient in handling complex ER tasks.
• Flexible interface suitable for various application scenarios.

Installation

You can install ERTool directly from PyPI using pip:

pip install ertool

Using Instruction

er_algorithm

ertool.er.er_algorithm(W, DBF, numOfEvidence, numOfPropositions)

er_algorithm() can implement the Evidential Reasoning (ER) algorithm.

Input Variables

• W: A one-dimensional array of floats. It represents the weights of each piece of evidence. These weights are used in the algorithm to adjust the influence of each evidence.
• DBF: A two-dimensional array of floats. It stands for "Degrees of Belief" and is one of the main inputs to the algorithm, used to represent the initial belief degree of each proposition supported by each evidence.
• numOfEvidence: An integer. It indicates the number of evidence to be combined. In the DBF array, this typically corresponds to the number of rows.
• numOfPropositions: An integer. It indicates the number of propositions or evidential grades. In the DBF array, this typically corresponds to the number of columns.

Output Values

• B Array: Upon completion of the algorithm, the B array is updated with the final calculation results. It reflects the degree of belief of each proposition or evidential grades for the object being assessed after combining all availble evidence. The pre-Numofproposition values in the B represent the belief degree of each proposition after evidence fusion. The last value of the B represents the belief degree of the global uncertainty.
• False (Boolean): It returns True if the algorithm successfully executes and completes all computations. If any error is encountered during execution (e.g., division by zero), it returns False.

dempster_shafer

ertool.er.dempster_shafer(DBF, numOfEvidence, numOfPropositions)

dempster_shafer() can implement the original Dempster-Shafer evidence theory.

Input Variables

• DBF: A two-dimensional array of floats. It stands for "Degrees of Belief" and is one of the main inputs to the algorithm, used to represent the initial belief degree of each proposition supported by each evidence. The pre-Numofproposition values in the B represent the belief degree of each proposition after evidence fusion. The last value of the B represents the belief degree of the global uncertainty.
• numOfEvidence: An integer. It indicates the number of evidence to be combined. In the DBF array, this typically corresponds to the number of rows.
• numOfPropositions: An integer. It indicates the number of propositions or evidential grades. In the DBF array, this typically corresponds to the number of columns.

Output Values

• B Array: Upon completion of the algorithm, the B array is updated with the final calculation results. It reflects the degree of belief of each proposition or evidential grades for the object being assessed after combining all availble evidence.
• False (Boolean): It returns True if the algorithm successfully executes and completes all computations. If any error is encountered during execution (e.g., division by zero), it returns False.

show_er_result

ertool.er.show_er_result(B, P = None)

er.show_er_result() can visualize the result of evidential reasoning algorithm.

Input Variables

• B: The ER result of belief degree.
• P: The name array of propositions.

run_algorithm_from_file

ertool.er.run_algorithm_from_file(file_path, algorithm = ’ER’)

run_algorithm_from_file() reads CSV or XLSX files and performs multi-source evidence fusion on the data using ER approach or Dempster-Shafer’s theory.

Input Variables

• file_path: A string. The location of the CSV or XLSX file. Note that the format of data strictly follows the format of the provided template.
• algorithm: ’ER’ or ’DS’. ‘ER’ stands for using the ER approach, and ’DS’ stands for using the Dempster-Shafer theory.

Output

• B Array: Upon completion of the algorithm, the B array is updated with the final calculation results. It reflects the degree of belief in each proposition or evaluation grade for the object being assessed after combining all available evidence. The first numofPropositions members in the B represent the belief degree in each proposition after evidence fusion. The last member of the B represents the belief degree in the global uncertainty.
• False (Boolean): It returns True if the algorithm successfully executes and completes all computations. If any error is encountered during execution (e.g., division by zero), it returns False.

Quick Start

Here is a basic usage example of ERTool.

Consider a medical scenario. There are three medical experts (weights 10, 8, and 5). For one patient, the three experts rated the different likelihood of the diagnosis of cold, common pneumonia, COVID-19, and other diseases. As shown in the table.

Experts & Diseases	Expert 1	Expert 2	Expert 3
Cold	90%	0	0
Common Pneumonia	0	90%	0
COVID-19	0	0	90%

In this case, the numOfEvidence is 3 (the number of experts) and the numOfPropositions is 3 (cold, common pneumonia and COVID-19).

The W array is the weights array of every expert and the ERTool package can normalize them automatically.

We can write the code using the ERTool package:

from ertool import er
import numpy as np

W = np.array([10,8,5])
DBF = np.array([[0.9, 0, 0], 
                [0, 0.9, 0], 
                [0, 0, 0.9]])

# List or numpy array are both OK.
# W = [10,8,5]
# DBF = [[0.9, 0, 0], 
#        [0, 0.9, 0], 
#        [0, 0, 0.9]]

numOfEvidence = 3
numOfPropositions = 3
B = er.er_algorithm(W, DBF, numOfEvidence, numOfPropositions)
print("The result: ", B)

P = ['Cold', 'Common Pneumonia', 'COVID-19']
er.show_er_result(B, P)

With the code, we can calculate the probability that the patient will be diagnosed with each disease using evidential reasoning.

The result: [0.43317251 0.307059 0.16390311 0.09586537]

The calculation results show that the probability of the patient being diagnosed with a cold, common pneumonia, and COVID-19 are 0.43317251, 0.307059, and 0.16390311, respectively. The last member of B represents global uncertainty, and it is 0.09586537 in this example.

Contributing

Contributions to ERTool are welcome. Please contact us for how to contribute to the project.

Reference

Tongyue Shi, Liya Guo, Zeyuan Shen, Guilan Kong. ERTool: A Python Package for Efficient Implementation of the Evidential Reasoning Approach for Multi-source Evidence Fusion. 2024.

Contact

This project is supported by Peking University. For any questions or suggestions, please contact us at [e-mail].