Throughput prediction of fifth-generation cellular system using hybrid feature selection and enhanced sequential decision tree machine learning algorithm

This paper proposes enhanced sequential decision tree (ESDT) for the prediction of fifth-generation (5G) cellular network throughput. The dataset which is used as input for machine learning (ML) model without preprocessing steps is called as dataset 1 and contains 49,706 no. of records. Missing valu...

Full description

Saved in:
Bibliographic Details
Published in:Wireless networks Vol. 31; no. 3; pp. 3025 - 3042
Main Authors: Sharma, Abhilasha, Pandit, Shweta, Talluri, Salman Raju
Format: Journal Article
Language:English
Published: New York Springer US 01.03.2025
Springer Nature B.V
Subjects:
Algorithms
Communications Engineering
Computer Communication Networks
Datasets
Decision trees
Electrical Engineering
Engineering
Feature selection
IT in Business
Machine learning
Networks
Original Paper
Performance enhancement
Preprocessing
Support vector machines
Throughput prediction
Random forest
Support vector machine
Decision tree
Enhanced sequential decision tree
K-nearest neighbor
ISSN:1022-0038, 1572-8196
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This paper proposes enhanced sequential decision tree (ESDT) for the prediction of fifth-generation (5G) cellular network throughput. The dataset which is used as input for machine learning (ML) model without preprocessing steps is called as dataset 1 and contains 49,706 no. of records. Missing value imputation, one hot encoding, embedded features selection (FS) followed by recursive FS and standardization are performed as the preprocessing steps and the resulting dataset containing 68,118 records is named as dataset 2. These steps are applied to enhance the classification and regression accuracy of 5G throughput prediction. After preprocessing, embedded feature selection is carried out to remove the features with less information. To select the best subset of features, forward recursive feature selection technique is employed. The score of the model is computed after training the model and hyperparameter tuning is done to improve the performance. Different machine learning models such as support vector machine, decision tree, random forest, k-nearest neighbor are used to compare the performance of ESDT model using dataset 1 and dataset 2. In order to improve generalization and prove that our proposed model is not biased to a single dataset, two more datasets called as the 4G dataset and the 5G production dataset have been used to evaluate the presented methodology.The proposed model ESDT has achieved the best 5G throughput prediction results compared to other ML models on the datasets.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1022-0038
1572-8196
DOI:10.1007/s11276-025-03917-3