DE202023101924U1 - A modified naive Bayes classifier for mode switching and mobility management in cellular networks - Google Patents

Abstract

A system (100) using a modified Naive Bayes classifier for mode switching and mobility management using a cellular network, comprising:
an integration module that enables the integration of mode switching and mobility management in the network includes:
a switching model is used to allow user mobility during communication;
a naive Bayesian algorithm is used for user mode or cellular mode evaluation;
mobility management for route maintenance is carried out.

Description

The present invention relates to the field of machine learning. The system uses a novel system switching model using the Naive Bayes classifier for mode switching and mobility management over cellular networks.

5G has the "Big Five" features of device-centric communications, massive multiple input multiple output (M-MIMO), native support for heterogeneous communications, and smarter. Evolutionary algorithms (EAs) and conventional mobility management techniques are used to achieve device-centric communication while DMM approaches are used for mobility management, but these techniques face challenges that include fairness, high signaling costs, and irregular topologies.

CN110431911 - The present disclosure relates to a communication method and system for convergence of a 5G communication system to support higher data rates than a 4G system with an loT technology. The present disclosure can be applied to intelligent services based on 5G communication technology and loT-related technology, such as smart home, smart building, smart city, smart car, connected car, healthcare, digital education, smart retail, security and security services. The present disclosure provides a scheme for the efficient operation of a session UP connection in the case that a terminal has a plurality of sessions in a mobile communication system, such as a 5G system, with a network structure in which an AMF for the mobility management and an SMF for session management are separate. Through the present disclosure, a user equipment (UE) can optimize a non-access layer (NAS) signaling message and perform data transmission/reception with low latency.

CN112534869 - The document describes mobility improvements for user equipment (UE) (110) between cellular and wireless local area networks (WLAN) (170) in fifth generation wireless networks (5G NR), as well as in 3rd Generation Partnership Project Long-Term Evolution (3GPP LTE) or Evolved Universal Terrestrial Radio Access (E-UTRA) networks. A cellular WLAN network interface is introduced to monitor and manage WLAN networks (170) and access points (160) and to handover UE (110) between WLAN APs (160), between WLAN networks (170) and between WLAN networks (170) and cellular networks (140). The cellular WLAN network interface allows an access and mobility function (220) in a 5G network or a mobility manager (330) in an E-UTRA network to request information from UE (110) and WLAN access points (160). to manage the operational configuration of WLAN access points (160) and to initiate handoffs of UE (110).

To solve the problem, the present invention provides a modified Naive Bayes classifier for mode switching and mobility management in cellular networks.

The system improves the performance of mobility management in the network.

The system improves user mobility during communication.

The system improves available spectrum reuse and minimizes power consumption by improving network throughput.

The system offers superior performance with a minimum delay of 0.164s, a maximum performance of 58.786 bpm, a maximum link utilization of 0.727 and a maximum throughput of 1,641,723.

The system achieves minimal network latency by solving the problem of limited capacity of network caches.

The system reduces training data and is easy to implement.

In one embodiment, the system proposes mode switching using Naive Bayes for mobility management, where networks switch the communication mode from cellular network mode to user mode for reliable communication. The switching factor is determined based on signal strength, bandwidth usage, energy, delay and connection utilization. The Naive Bayes classifier is designed to allow the user to switch between cellular mode and user-centric mode processing. The switching factors are used in this case by Naive Bayes to change modes for better QoS and travel in each direction. When using Naive Bayes, the proposed approaches to mode switching were outperformed by the classifier.

In one embodiment, the system is used to generate real-time predictions and processes both discrete and continuous data.

In one embodiment, the system is used to provide better services to mobile users.

In one embodiment, the system is used to improve signal strength, bandwidth, power, delay, and link utilization.

• : zeigt das Blockdiagramm eines modifizierten Naive-Bayes-Klassifikators für Moduswechsel und Mobilitätsmanagement in zellularen Netzen.

The invention is explained again below with reference to the figure. show:

• : shows the block diagram of a modified Naive Bayes classifier for mode switching and mobility management in cellular networks.

: shows the block diagram of a modified Naive Bayes classifier for mode switching and mobility management in cellular networks. The system (100) includes an integration module. The integration module is used to enable the integration of mode switching and mobility management in the network, and further includes a switching model to enable user mobility during communication and a Naive Bayes algorithm used to evaluate the user mode or the cellular mode is used and mobility management performed to maintain the route. The modified Naive Bayes classifier for mode switching and mobility management in cellular networks was studied to develop a mode switching scheme in cellular networks for mobility management.

Proposed Naïve Bayes classifier for mobility management in cellular networks.

The aim of the system (100) is to develop a model for mode switching in order to implement mobility management in cellular networks. First, the cellular network is initialized to create a route for further communication. Once the route is determined, route finding is performed using the optimal routing algorithm model. Once the optimal route is selected, communication is initialized, followed by dynamic mode switching, where the network switches the communication mode for each user to operate in either CM or UM mode to achieve reliable communication. Mode switching is determined based on the switching factor, which takes into account factors such as signal strength from UE or CN, bandwidth utilization from UE or CN, energy from UM or CM, delay from UM or CM, and link utilization from UM or CM. These performance metrics are used as input to the Naive Bayes classifier to select UM or CM. After mode switching, mobility management is performed to maintain the route.

The flow of the proposed mobility management model in cellular networks using the Naive Bayes classifier.

Network initialization

wobei myx die Anforderungsrate für das Objekt Ny am Knoten x angibt, tx, r, y die Entfernung angibt, die der Knoten x benötigt, um das Inhaltsobjekt Ny vom Knoten r anzufordern. nx, r, y ist hier ein Element, das den Wert 1 annimmt, wenn der Knoten x eine Kopie des Inhaltsobjekts y vom Knoten r herunterlädt, was folgendermaßen formuliert wird, $${\displaystyle \sum _{r=1}^s{n}_{x,r,y}}=\mathrm{1,}\forall x,y$$

$$n_{x,r,y}\le W_{r,y},\forall x,r,y$$

$${\displaystyle \sum _{y=1}^T{z}^y{W}_{x,y}\le X_x,\forall x}$$

$$n_{x,r,y}\in \left\{\mathrm{0,1}\right\}\text{ and }{W}_{x,y}\in \left\{\mathrm{0,1}\right\},\forall x,r,y$$

The network is modeled as a connected graph M = (K, L), where K = (K1, K2...Ks) represents a set of content routers in the network and L ⊆ (K × K) a set of bidirectional links connecting devices in the network. Suppose N = (N1N2...NT) denotes a set of content objects available on the network. These objects are primarily distributed on the network servers, which are tightly connected to the edge content routers. The goal is to achieve minimal network latency by solving the problem of limited cache capacity in the network. Therefore, the optimal routing problem is modeled as an integer linear problem (ILP) and formulated as follows, $$\underset{\mathrm{and}.v}{\text{at least}}:{\displaystyle \sum _{x=1}^S{\displaystyle \sum _{y=1}^T{m}_x^v}{\displaystyle \sum _{\mathrm{right}=1}^S{t}_{x,\mathrm{right},y}{n}_{x,\mathrm{right},y|}}}$$

where myx indicates the request rate for object Ny at node x, tx,r,y indicates the distance node x needs to request content object Ny from node r. Here nx,r,y is an element that takes the value 1 if node x downloads a copy of content object y from node r, which is formulated as follows, $${\displaystyle \sum _{\mathrm{right}=1}^s{n}_{x,\mathrm{right},y}}=\mathrm{1,}\forall x,y$$

$$n_{x,\mathrm{right},y}\le W_{\mathrm{right},y},\forall x,\mathrm{right},y$$

$${\displaystyle \sum _{y=1}^T{\mathrm{e.g}}^y{W}_{x,y}\le X_x,\forall x}$$

$$n_{x,\mathrm{right},y}\in \left\{0.1\right\}\text{other}{W}_{x,y}\in \left\{0.1\right\},\forall x,\mathrm{right},y$$

Mode switching using performance metrics with Naive Bayes classifier

Here, the dynamic mode switching process is illustrated using the Naive Bayes classifier, with the switching factors serving as input to the classifier to select UM or CM to achieve reliable communication. The user-centric mode is less dependent on network infrastructure and the associated enhanced node represents control over performance measures. Naive Bayes is a clustering and classification model based on Bayes' theorem. This model estimates the probability per class by assuming the attributes are unconditionally independent. It uses the same method to predict the likelihood of different classes based on different attributes. It is mainly used to solve multi-class problems. The Naive Bayes model has several advantages, some of which are listed here. It doesn't require that much training data and is easy to implement. It can be used for real-time predictions. Also, it can handle both discrete and continuous data.

• Schritte der Modusumschaltung mit Hilfe von Umschaltfaktoren
• Schritt 1 Das UE löst einen Ruf aus und wird von dem verbündeten erweiterten Knoten übernommen.
• Schritt 2 Der erweiterte Knoten berechnet den Standort der Zelle und des Anrufers und bestimmt den Standort des Zielknotens.
• Schritt 3 Der erweiterte Knoten berechnet Schaltfaktoren wie Signalstärke, Bandbreite, Energieverbrauch, Verzögerung und Verbindungsauslastung für jedes UE und jeden CN, der der mobile Zielnutzer ist.
• Schritt 4 Die erhaltenen Umschaltfaktoren werden in den Naive Bayes-Klassifikator eingespeist, der feststellt, ob der Benutzer in UM oder CM arbeitet.
• Schritt 5 In UM wertet der erweiterte Knoten die QoS-Parameter für jeden mobilen Benutzer aus. Wenn die QoS-Werte unter dem Schwellenwert liegen, wird ein Relaisknoten zur Verbesserung der Leistung des Netzwerks angebracht.
• Schritt 6 Der erweiterte Knoten kontrolliert die Leistung des Systems mit benutzerorientierter Kopplung, um eine zufriedenstellende QoS bereitzustellen. Das Netzwerk überwacht die Leistung und passt bestimmte Maßnahmen an, um den mobilen Nutzern bessere Dienste zu bieten. Dieses Modul besteht aus zwei untergeordneten Phasen. Zunächst nimmt das Netz die Anruferinformationen auf und führt eine Analyse durch. Dann wird mit Hilfe des Naïve Bayes-Klassifikators eine Entscheidung für die Umschaltung der Modi auf der Grundlage von Leistungsmessungen getroffen, während zwischen UM und CM umgeschaltet wird. Um eine bessere Leistung bei der Modusumschaltung zu erreichen, werden die Signalstärke, die Bandbreite, die Energie, die Verzögerung und die Verbindungsauslastung berechnet.

Switching the operating mode is shown below:

• Mode switching steps using switching factors
• Step 1 The UE initiates a call and is taken over by the federated extended node.
• Step 2 The extended node calculates the location of the cell and the caller and determines the location of the destination node.
• Step 3 The advanced node calculates switching factors such as signal strength, bandwidth, power consumption, delay and link utilization for each UE and each CN that is the target mobile user.
• Step 4 The switching factors obtained are fed into the Naive Bayes classifier, which determines whether the user is working in UM or CM.
• Step 5 In UM, the advanced node evaluates the QoS parameters for each mobile user. If the QoS values are below the threshold, a relay node is placed to improve the performance of the network.
• Step 6 The advanced node controls the performance of the system with user-centric coupling to provide satisfactory QoS. The network monitors performance and adjusts certain measures to provide better services to mobile users. This module consists of two sub-phases. First, the network takes the caller information and performs an analysis. Then, using the Naïve Bayes classifier, a mode switching decision is made based on power measurements while switching between UM and CM. To achieve better mode switching performance, the signal strength, bandwidth, power, delay and link utilization are calculated.

Mode switching using a Naive Bayes classifier to switch modes.

Here, dynamic mode switching is illustrated using Naïve Bayes and a set of performance metrics. Here the network switches the communication mode from network-centric mode to user-centric mode to enable realistic communication. These five performance metrics serve as input to the Naïve Bayes classifier for evaluating user mode or cellular mode. Naïve Bayes is a classifier used for classification based on Bayes' theorem. This Bayesian classifier uses Bayes' theorem to determine the probability of data fitting into a given class associated with an observation. This model assesses the probability per class by considering the attributes that are conditionally independent. This classifier is easy to create and does not involve complex iterative parameters and evaluations, making it advantageous for a hardware implementation. Naïve Bayes is known as supervisory machine learning, commonly used for dataset classification. This classifier predicts things based on its prior knowledge and sovereign assumptions. In addition, this classifier is effective in making decisions using anonymous datasets. The functioning of the naive Bayes classifier is examined using performance measures. Suppose H denotes the class variable, Fa the conditional attribute a used for classification, and P (H¬F1...Fb) represents the conditional probability of class H given the proof F1,...,Fb occured. The probability model of the Naive Bayes classifier is expressed as follows, $$P\left(H|f_1,...,f_b\right)=\frac{P\left(H\right)\times P\left(f_1,...,f_b|H\right)}{|P\left(f_1,...f_b\right)}$$

As an alternative to assessing the conditional probability for the merging of class H attributes Fa, a conjecture was considered, formulated as follows, $$P\left(H|f_1,...f_b\right)=\frac{P\left(H\right)\times P\left(f_1|H\right)\times ...\times P\left(f_b|H\right)}{P\left(f_1,...f_b\right)}$$

The best class for assigning data is the one that maximizes this conditional probability among all classes. For classification, the above equation is known as likelihood functions formulated as a function of H with F. For the class variables H1......H¬ one can categorize the evidence in ¬ likelihood values. In addition, the proofs are assigned to the class with a high probability. Whenever the likelihood function is compared, one can effectively filter P(F1...,Fb). Thus, the likelihood function is expressed as follows, $$LikeliHOO\mathrm{i.e}\left(H_a\right)=P\left(H_a\right)\times P\left(f_1|H_a\right)\times ...P\left(f_b|H_a\right)$$

The likelihood calculation is essential because it does not require a large amount of training due to the evidence showing the division into different aspects by the presumption of sovereignty. Furthermore, the conditional densities of the classes for each attribute can be calculated independently and the multi-dimensional task is reduced to a one-dimensional task. The output produced by the Naïve Bayes classifier is either UM or CM.

Reference List

100

system

QUOTES INCLUDED IN DESCRIPTION

This list of documents cited by the applicant was generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Patent Literature Cited

• CN110431911 [0003]
• CN112534869 [0004]

Claims (5)

A system (100) using a modified Naive Bayes classifier for mode switching and mobility management using a cellular network, comprising: an integration module that enables the integration of mode switching and mobility management in the network includes: a switching model is used to allow user mobility during communication; a naive Bayesian algorithm is used for user mode or cellular mode evaluation; mobility management for route maintenance is carried out. system (100) after claim 1 , where the naïve Bajes use the switching factors as input for switching modes. The switching factors include certain performance-based parameters, such as link utilization, bandwidth, delay, power consumption, and signal strength for each mobile user. system (100) after claim 1 , where mobility management between nodes is a fundamental factor of cellular communication, allowing the user to move in any direction to improve QoS. system (100) after claim 1 , where mobility management is responsible for managing user-mode and cellular model-based communication, and communication reverts to traditional modes when the threshold is breached. system (100) after claim 1 , where the modified naive Bayes classifier for mode switching and mobility management using a cellular network was studied to design a mode switching scheme in cellular networks for mobility management.

Images