What question did this study set out to answer?

The research aims to enhance packet delivery reliability in opportunistic networks by applying supervised learning techniques to routing algorithms.

June 11, 2026Open Access

Enhancing packet delivery in opportunistic networks using supervised learning techniques

Key Points

The research aims to enhance packet delivery reliability in opportunistic networks by applying supervised learning techniques to routing algorithms.
Incorporated supervised learning models to forecast packet delivery success.
Investigated various ML models, including Neural Networks, Support Vector Machines, Logistic Regression, and Random Forests.
Evaluated models on simulated datasets using metrics such as accuracy, recall, and F1-score.
Random Forest achieved a prediction accuracy of 91% and recall of 95%, indicating strong performance in dynamic environments.
The proposed framework enhances routing choices and assesses performance using network-specific indicators and machine learning metrics.
Findings suggest significant improvements in communication efficiency and packet delivery dependability using ML-based routing.

Abstract

In this paper, we analyze how Machine Learning (ML) techniques can be applied to improve packet delivery reliability in Opportunistic Networks. These networks are useful in situations where connectivity is sporadic, such disaster recovery or rural areas, because they allow communication without an established infrastructure. Nevertheless, effective routing is a significant difficulty due to OppNets’ changing architecture, constrained resources, and erratic connectivity. By incorporating supervised learning models that can forecast the likelihood of packet delivery success and choose the most dependable relay nodes, this work aims to improve routing algorithms. The three primary goals of the suggested framework are to: enhance routing choices by using predictive models; find clusters of nodes with comparable mobility patterns to maximize forwarding opportunities; and assess routing performance using both network-specific indicators and machine learning metrics. Several ML models are investigated, including Neural Networks for predicting node mobility, Support Vector Machines for prioritizing delivery based on success probability, Logistic Regression for simple yet efficient classification, and Random Forests for extracting delivery patterns from historical data. The models are trained and evaluated on simulated datasets using accuracy, recall, and F1-score to quantify predictive quality, which directly reflects routing performance in terms of delivery ratio and latency. According to the results, Random Forest exhibits excellent robustness under dynamic situations and strikes the optimal balance between prediction accuracy (91%) and recall (95%). These results imply that in difficult situations, ML-based predictive routing may greatly increase communication efficiency and packet delivery dependability. Additionally, the study provides fresh insights into how to incorporate lightweight and flexible machine learning models into practical opportunistic routing strategies

Read Full Paperexternally

KI fragen

Bookmark

View Full Paper