User-Level Handover Decision Making Based on Machine Learning Approaches

João Lima; Alvaro Medeiros; Eduardo Aguiar; Vicente Angelo de Sousa Junior; Tarciana Guerra

cs.LGSep 18, 2026

User-Level Handover Decision Making Based on Machine Learning Approaches

Authors: João Lima, Alvaro Medeiros, Eduardo Aguiar, Vicente Angelo de Sousa Junior, Tarciana Guerra

Organizations: CPQD · Federal University of Juiz de Fora, Brazil · Munster Technological University, Cork, Ireland · Federal University of Rio Grande Norte, Brazil

Abstract

This letter covers a broad comparison of methods for classification and regression applications for a user-level handover decision making in scenarios with adverse propagation conditions involving buildings, coverage holes, and shadowing effects. The simulation campaigns are based on network simulator ns-3. The comparison encompasses classical machine learning approaches, such as KNN, SVM, and neural networks, but also state-of-the-art fuzzy logic systems and latter boosting machines. The results indicate that SVM and MLP are the most suitable for the classification of the best handover target, although fuzzy system SOFL can perform similarly with lower processing time. Additionally, for the download time estimation, LightGBM provides the smallest error with short processing time, even in hard propagation scenarios.

Figures & tables

Event	Description
A1	Primary cell (PC) signal power becomes better than a threshold
A2	PC signal power becomes worse than a threshold
A3	Secondary cell (SC) signal power becomes better than PC by an offset
A4	SC signal power becomes better than a threshold

TABLE I: Characteristic events for 3GPP Handover [ 1 ] .

Method	Accuracy (%)	Std. Dev.	Time (s)
MLP	99.72	0.10	5.25
SVM	99.74	0.08	0.54
SOFL	99.11	0.22	12.51
T2FLS	98.94	0.68	5845.78
ALMMo	99.61	0.11	257.47

TABLE II: Initial results for Scenario 1.

Method	Accuracy (%)	Std. Dev.	Time (s)
MLP	86.22	0.96	15.94
SVM	86.34	0.33	21.26
SOFL	85.32	0.71	14.60
T2FLS	72.40	0.98	7672.62
ALMMo	68.06	1.16	357.74

TABLE III: Initial results for Scenario 2.

${G}_{1}$	${G}_{2}$	$p$ -value	Low. b.	Upp. b.	${H}$
SVM	ALM	3.56E-07	0.0009	0.0018	1
SVM	MLP	0.425	-2.70E-4	6.34E-4	0
SVM	SOFL	4.66E-19	0.0055	0.0071	1
SVM	T2FL	1.17E-07	0.0056	0.0105	1
MLP	ALM	2.13E-05	0.0007	0.0017	1
MLP	SOFL	5.12E-19	0.0053	0.007	1

TABLE IV:

T

-test results for Scenario 1.

${G}_{1}$	${G}_{2}$	$p$ -value	Low. b.	Upp. b.	${H}$
SVM	ALM	1.26E-44	0.1786	0.1871	1
SVM	MLP	0.5019	-0.0024	0.0048	0
SVM	SOFL	1.66E-9	0.0075	0.013	1
SVM	T2FL	2.47E-44	0.1358	0.1431	1
MLP	ALM	1.86E-60	0.1764	0.1869	1
MLP	SOFL	5.76E-5	0.0049	0.0132	1

TABLE V:

T

-test results for Scenario 2.

Method	$MAE$	Std. Dev.	Time (s)
MLP	0.27163	0.00800	21.85
KNN	0.12791	0.00490	1.24
RF	0.11987	0.00524	49.34
GBM	0.12452	0.00727	32.98
XGBoost	0.12343	0.00714	25.67
LightGBM	0.11553	0.00317	14.81

TABLE VI: Regression results for Scenario 1.

Method	$MAE$	Std. Dev.	Time (s)
MLP	6.32415	0.07108	8.98
KNN	6.00046	0.07525	1.68
RF	5.99726	0.05373	53.05
GBM	6.19593	0.07841	34.15
XGBoost	5.88485	0.10150	17.53
LightGBM	5.08845	0.04832	13.76

TABLE VII: Regression results for Scenario 2.

${G}_{1}$	${G}_{2}$	$p$ -value	Low. b.	Upp. b.	${H}$
XGB	Light	6.85E-7	0.0052	0.0106	1
XGB	GBM	0.5785	-0.0047	0.0027	0
XGB	MLP	8.11E-40	-0.1530	-0.1406	1
XGB	RF	0.0493	-0.0014	0.0063	1
XGB	KNN	0.0049	-0.0076	-0.0014	1
Light	GBM	1.88E-7	-0.0118	-0.0061	1

TABLE VIII:

T

-test results for regression in Scenario 1.

${G}_{1}$	${G}_{2}$	$p$ -value	Low. b.	Upp. b.	${H}$
XGB	Light	1.99E-37	0.7565	0.8363	1
XGB	GBM	3.00E-20	-0.3563	-0.2658	1
XGB	MLP	1.13E-27	-0.4830	-0.3956	1
XGB	RF	1.19E-6	-0.1576	-0.0731	1
XGB	KNN	2.71E-6	-0.1602	-0.0710	1
Light	GBM	1.95E-53	-1.1401	-1.0748	1

TABLE IX:

T

-test results for regression in Scenario 2.

ML-Based Real-Time Downlink Performance Prediction in Standalone 5G NR Using Smartphones
Mar 19, 2026Md Mahfuzur Rahman, Jareen Shuva, Nishith Tripathi +2Wireless Communications
Dueling DDQN-Based Adaptive Multi-Objective Handover Optimization for LEO Satellite Networks
May 4, 2026Po-Heng Chou, Chiapin Wang, Chung-Chi Huang +1Multi-Objective Reinforcement Learning Deep Q-Learning
Research Center for Information Technology Innovation (CITI), Academia Sinica (AS), Taipei 11529, Taiwan · Department of Electrical Engineering, National Taiwan Normal University (NTNU), Taipei 106308 Taiwan
Energy and Latency First: How Deployment Constraints Reverse Time-Series Model Selection
Oct 22, 2026Shruti Bothe, Illyyne Saffar, Ayush K. VarshneyModel Selection Energy-Efficient ML
Ericsson Research, Santa Clara, USA · Ericsson Research, Massy, France · Ericsson Research, Kista, Sweden

Mar 19, 2026J/K move · Enter open · S save

ML-Based Real-Time Downlink Performance Prediction in Standalone 5G NR Using Smartphones

Md Mahfuzur Rahman, Jareen Shuva, Nishith Tripathi, Jeffrey H. Reed+1

Wireless Communications

Open paper AskPDF arXiv

We propose a machine learning (ML)-based framework for downlink performance prediction in 5G networks using real-time measurements from commercial off-the-shelf (COTS) user equipment (UE). Our experimental platform integrates the srsRAN 5G New Radio (NR) stack deployed on a Dell desktop serving as the 5G next generation nodeB (gNB), operating at 3.4 GHz. Two Google Pixel 7a smartphones are used to collect physical layer characteristics such as channel quality indicator (CQI), modulation and coding scheme (MCS), bit rate, transmission time interval (TTI), and block error rate (BLER), which are leveraged as predictors in model training. We use commercial-grade traffic generation tools, including Ookla, for stationary and mobility measurements under line-of-sight (LOS) and non-line-of-sight (nLOS) conditions. Test data includes global Ookla servers (e.g., USA, Portugal, Ghana, Egypt, Japan), iperf TCP/UDP data, and video streaming sessions from YouTube. To analyze inter-user interference, we also include scenarios with multiple UEs at the same location. We evaluate the predictive performance of five supervised regression models - linear regression, decision tree regression, random forest regression, extreme gradient boosting (XGBoost), light gradient boosting machine (LightGBM). Our results demonstrate that throughput and BLER can be accurately predicted using COTS hardware and standard ML techniques in diverse real-world 5G scenarios.

Abstract

Explore similar work

ML-Based Real-Time Downlink Performance Prediction in Standalone 5G NR Using Smartphones

Dueling DDQN-Based Adaptive Multi-Objective Handover Optimization for LEO Satellite Networks

Energy and Latency First: How Deployment Constraints Reverse Time-Series Model Selection