Overcoming the Challenge of Limited Data in Classification Tasks using Neighborhood-Based Positive Unlabeled Learning using Decision Trees

Classification tasks are a fundamental aspect of machine learning, enabling systems to categorize data into predefined classes. However, many real-world applications lack sufficient labeled data, making it challenging to develop accurate classification models. In this study, we present a novel method that addresses this problem by incorporating neighborhood-based positive unlabeled learning using decision trees (NPULUD).

Traditional binary supervised classification tasks require both negative and positive samples in the training dataset to construct a classification model. However, in certain applications, only one class of samples may be obtainable, making it necessary to adopt alternative approaches. One such approach is learning from positive and unlabeled (PU) data, which has been shown to be effective in overcoming the limitations of traditional supervised learning methods.

NPULUD leverages the concept of neighborhood-based learning to identify patterns in the data that are characteristic of a specific class. By analyzing the relationships between samples, NPULUD can distinguish between positive and negative instances, even when only one class is present in the training dataset. This approach has significant implications for applications where labeled data is scarce or difficult to obtain.

Entropy, as an interdisciplinary journal, provides a platform for researchers to share their findings and advance our understanding of entropy and information studies. As part of this endeavor, we recognize the importance of open access publishing, allowing readers worldwide to access cutting-edge research without barriers.

In addition to its commitment to open access, Entropy also prioritizes rapid publication, with an average time-to-publication of just 2.9 days. This enables authors to share their work quickly and efficiently, maximizing the impact of their research.

As a leading journal in the field, Entropy has established itself as a go-to destination for researchers seeking to publish high-quality papers. With its strong reputation for rapid publication, high visibility, and commitment to open access, Entropy is an ideal choice for authors looking to share their work with the global academic community.

Entropy magazine