This ongoing project aims to develop a comprehensive machine learning classification package. A notable feature of this package is its integration of Dempster-Shafer theory, enhancing classification accuracy compared to conventional methods like logistic regression.
- Train the model: assign subjective probabilites (masses) to all subsets of the data.
- Predict
Dempster-Shafer theory offers a novel perspective on evidence consolidation. By initially considering a set of possibilities, it partitions them into two key components: belief (indicating the strength of evidence) and plausibility (encompassing all potential values, including the evidence). This unique approach has the potential to elevate classification performance, even if the improvements are incremental, when compared to widely-used methods such as logistic regression.
The versatility of this algorithm is particularly evident in its application to various data types, with a notable focus on image classification. Theoretically, the algorithm can harness evidence to accurately classify images and determine their representations.
Getting started with our package is a breeze. Simply install it using pip (Python 3) with the following command:
pip install dempster_shafer_classificationOur ongoing efforts are focused on optimizing parameters to enhance belief strength, a process currently undergoing refinement through a grid search. This will ultimately contribute to an optimized grid search technique that further enhances the effectiveness of Dempster-Shafer theory in classification tasks.
We invite you to join us in this exciting journey of advancing classification accuracy through innovative approaches. Feel free to contribute, provide feedback, and help shape the future of classification techniques.
We value and appreciate contributions from the open-source community. If you're interested in contributing to the Dempster-Shafer Enhanced Classification Package, please review our Contribution Guidelines for detailed information on how to get involved.
This project is licensed under the MIT License. Your use of this software signifies your acceptance of the terms and conditions outlined in the license agreement.