WO2015130203A1

WO2015130203A1 - Method and device for predicting roaming

Info

Publication number: WO2015130203A1
Application number: PCT/SE2014/050242
Authority: WO
Inventors: Azedeh BARARSANI; Tor Kvernvik; Tony Larsson; Yu Wang
Original assignee: Telefonaktiebolaget L M Ericsson (Publ)
Priority date: 2014-02-27
Filing date: 2014-02-27
Publication date: 2015-09-03

Abstract

It is presented a methodfor providing a network roaming prediction for a wireless communication device in a cellular network (5). The method is performed by a network device (20) of the cellular network and comprises the steps of: obtaining (43) live data location features for a wireless communication device(1) in the cellular network; and obtaining (45) a network roaming prediction class, from a trained network roaming prediction classifier (80), based on the live data location features, to predict if the wireless communication device(1) is about to network roam or not, wherein the network roaming prediction classifier (80) has been trained using historical data location features. A corresponding wireless communication device, computer program and computer program product are also presented.

Description

METHOD AND DEVICE FOR PREDICTING ROAMING

TECHNICAL FIELD

The invention relates to a method, network device, computer program and computer program product for prediction of network roaming. BACKGROUND

While mobile users move from one country to another, they typically need to switch to another network operator for mobile services, a roaming operator. During this transition, they may experience service interruption or service level downgrading, until they have selected and accept the costs to one of sometimes several roaming operators and get connected to the roaming network of the selected roaming operator. This leads to inconvenient user experience and therefore unhappy users.

US patent US 8,538,454 describes a method for determining a customized location area (CLA) for a subscriber in a cellular network, including determining which CLA a mobile station currently is located within.

SUMMARY

It is an object of the invention to enable facilitated network roaming in cellular networks.

According to a first aspect, it is presented a method for providing a network roaming prediction for a wireless communication device in a cellular network. The method is performed by a network device of the cellular network and comprises the steps of: obtaining live data location features for a wireless communication device in the cellular network; and obtaining a network roaming prediction class, from a trained network roaming prediction classifier, based on the live data location features, to predict if the wireless communication device is about to network roam or not, wherein the network roaming prediction classifier has been trained using historical data location features. By predicting that a wireless communication device is about to roam, preparation for roaming may be performed, which will facilitate roaming between operators, particularly from the end-user's perspective if the wireless communication device is executing a service which the end-user would like to seamlessly continue even when roaming.

The method may further comprise the step of triggering a network roaming preparation when a probability for network roaming of the wireless communication device is determined to be above a threshold, based on the network roaming prediction class. To avoid too many false alarms that a wireless communication device is about to roam, utilization of a threshold may be used to adjust the likelihood that a wireless communication device is predicted to change operator.

The network roaming preparation may comprise one or more of the following: notifying a user of the wireless communication device of the predicted network roaming, notifying the wireless communication device of the predicted network roaming, notifying an application in the wireless communication device of the predicted network roaming, notifying a network roaming operator that the wireless communication device is predicted to network roam thereto, notifying a cellular roaming network that the wireless communication device is predicted to network roam thereto, notifying an application of the cellular network that the wireless communication device is predicted to network roam, and notifying the cellular network that the wireless communication device is predicted to network roam. One or more of these preparations maybe utilized to ease change of operator.

The method may further comprise the step of triggering a network roaming prediction for the wireless communication device. The step of triggering a network roaming prediction may comprise the steps of: detecting a location of the wireless communication device; and determining the location of the wireless communication device to be in a network boarder cell or a neighbouring network boarder cell. The step of determining may comprise determining the location of the wireless communication device to be in a network boarder cell. To limit unnecessary use of network recourses, triggering of predictions may be based on proximity to network boarders. The step of determining the location of the wireless communication device may be performed by a satellite-aided positioning system, cell identity based positioning, or radio signalling based positioning, or a combination thereof. Localization may be used e.g. in dependence on capabilities of the wireless communication device.

The step of obtaining a network roaming prediction class may comprise the step of inputting the live data location features to the trained network roaming prediction classifier. The method may further comprise the step of training the network roaming prediction classifier using historical data of wireless communication devices, about to network roam or not, and resulting in roaming and not roaming classes. Machine learning is automatized i.e. if a travel habit changes the system will be automatically updated.

The historical data location features may comprise information about network boarder cell, neighbouring network boarder cell, time of entry of the neighbouring network boarder cell and time of entry of the network boarder cell. The historical model is trained resulting in if and when a WCD has roamed.

The live data location features may comprise information about network boarder cell, neighbouring network boarder cell, time of entry of the neighbouring network boarder cell and time of entry of the network boarder cell. With these data the classifier may estimate likelihood of roaming as well as time to roaming.

The live and historical data location features may comprise information about geographical position and/or cell identity and/or radio signalling data. The classifier may be a Support Vector Machine (SVM), a decision tree, a neural network or a Bayesian network.

According to a second aspect, it is provided a network device for a cellular network. The network device is arranged to predict if a wireless

communication device is about to network roam. The network device comprises: a processor; and a computer program product storing instructions that, when executed by the processor, causes the network device to: obtain live data location features for a wireless communication device in the cellular network; and obtain a network roaming prediction class, from a trained network roaming prediction classifier, based on the live data location features, resulting in a prediction if the wireless communication device is about to network roam or not.

According to a third aspect, it is provided a network device for a cellular network. The network device is arranged to predict if a wireless

communication device is about to network roam. The network device comprises: a live data manager configured to obtain live data location features for a wireless communication device in the cellular network; and a prediction class generator configured to obtain a network roaming prediction class, from a trained network roaming prediction classifier, based on the live data location features, resulting in a prediction if the wireless communication device is about to network roam or not.

According to a fourth aspect, it is presented a computer program for predicting if a wireless communication device is about to network roam. The computer program comprises computer program code which, when run on a network device, causes the network device to: obtain live data location features for a wireless communication device in a cellular network; and obtain a network roaming prediction class, from a trained network roaming prediction classifier, based on the live data location features, resulting in a prediction if the wireless communication device is about to network roam or not. According to a fifth aspect, it is presented a computer program product comprising a computer program and a computer readable storage means on which the computer program is stored.

Generally, all terms used in the claims are to be interpreted according to their ordinary meaning in the technical field, unless explicitly defined otherwise herein. All references to "a/an/the element, apparatus, component, means, step, etc." are to be interpreted openly as referring to at least one instance of the element, apparatus, component, means, step, etc., unless explicitly stated otherwise. The steps of any method disclosed herein do not have to be performed in the exact order disclosed, unless explicitly stated. BRIEF DESCRIPTION OF THE DRAWINGS

The invention is now described, by way of example, with reference to the accompanying drawings, in which:

Fig. l is a schematic diagram illustrating an environment where

embodiments presented herein can be applied; Fig. 2 is a schematic diagram illustrating details in a wireless communication device and a core network for an embodiment presented herein;

Fig. 3 is a schematic diagram illustrating details in a wireless communication device and a core network for another embodiment presented herein;

Figs. 4A-4B are flow charts illustrating methods for embodiments presented herein;

Fig. 5 is a schematic diagram illustrating some components of a network device;

Figs. 6A-6B are schematic diagrams illustrating various locations where the network device of Fig. 5 can be implemented; Fig. 7 is a schematic diagram illustrating a network boarder of a cellular network;

Fig. 8 is a schematic diagram illustrating a classifier; and

Fig. 9 is a schematic diagram showing functional modules of a network device. DETAILED DESCRIPTION

The invention will now be described more fully hereinafter with reference to the accompanying drawings, in which certain embodiments of the invention are shown. This invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided by way of example so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art. Like numbers refer to like elements throughout the description. A way of preventing a bad service experience for network roaming users is to estimate in advance when those users are likely to roam. In other words predicting a future location of a wireless communication device to be in a roaming network of the user, to e.g. be used as an input for offered service to those specific users who are about to roam to a new operator, typically across a national boarder. As a consequence such users may be informed about the need for switching to a new operator in advance of, or very early in, a roaming process. This will facilitate users' switching to a new network when roaming to another country.

Knowing in advance about roaming users can result in offering a better service. For example, informing a user of a wireless communication device about the fact that they will likely roam soon will give them the chance to prepare for it and stop their usage consciously, e.g. they can end a phone call or finish downloading before roaming to a high cost network, instead of causing an interruption with no prior notice. Depending on the operators' agreements, the operator in the visiting network may offer a few minutes of service for free, especially noticeable in video streaming use cases, so there will be no interruption in user's experience.

The visiting network may take advantage from knowing about likely roaming users. For example, getting to know about a high number of new users, or users/wireless communication devices which consume a relatively high degree of a certain service or bandwidth, coming to the visiting network will give the visiting network the chance for some preparation of the network for such a high number of new users joining the network, e.g. increasing the capacity of some cells, etc.

By notifying the wireless communication device of the predicted network roaming, the wireless communication device may notify the user that roaming may happen. The user may thus avoid or end expensive downloads.

By notifying an application program in the wireless communication device of the predicted network roaming, the application program may e.g. preload data. One example maybe to buffer streamed video at an accelerated rate before roaming and/ or by enlarging the buffer size itself before roaming.

This solution is based on identifying users' mobility pattern and correlating the pattern with historical mobility and roaming records. Applying machine learning prediction algorithms may help out with predicting future location of the users. Fig. 1 is a schematic diagram illustrating a cellular network 5 providing an environment where embodiments presented herein can be applied. The cellular network 5 is connected to a wireless communication device (WCD) 1 in connectivity with a base station 2, such as an eNodeB in a Long Term Evolution (LTE) access network connected to a core network 3. The core network 3 is in turn connected to a roaming core network 4 connected to a roaming base station 6. The term wireless communication device maybe or alternatively be termed as a mobile communication terminal, user

equipment, mobile terminal, user terminal, user agent, machine-to-machine device etc., and can be, for example, what today are commonly known as a smartphone or a tablet/laptop with wireless connectivity. Moreover, the

WCD 1 may, but do not need to, be associated with a particular end user. The WCD 1 may also be a telematics unit embedded in a vehicle such as a car, bus and truck, and be connected to a vehicle-internal network for exchange of e.g. vehicle or driver data with a fleet management system connected to the vehicle via the cellular network 5. The WCD may also be a unit mounted in a dashboard of a vehicle for displaying information and communicating with the driver or passengers of the vehicle and being connected to the telematics unit embedded in the vehicle.

A network device 20 for a cellular network 5 is presented with reference to Fig. 5, which network device 20 is arranged to determine network roaming predictions if a WCD 1 is about to network roam or not. The network device 20 comprises: a processor 30; and a computer program product 62 storing a computer program 64 with instructions that, when executed by the processor 30, causes the network device 20 to: obtain live data location features for a WCD 1 in the cellular network; and obtain a network roaming prediction class, from a trained network roaming prediction classifier, based on the live data location features, resulting in a prediction if the WCD is about to network roam or not.

Fig. 5 is a schematic diagram showing some components of the network device 20. The processor 30 may be provided using any combination of one or more of a suitable central processing unit (CPU), multiprocessor, microcontroller, digital signal processor (DSP), application specific integrated circuit etc., capable of executing software instructions if a computer program 64 stored in a memory. The memory can thus be considered to be or form part of the computer program product 62. The processor 30 maybe configured to execute methods described herein with reference to Figs. 4A-4B.

The memory may be any combination of read and write memory (RAM) and read only memory (ROM). The memory may also comprise persistent storage, which, for example, can be any single one or combination of magnetic memory, optical memory, solid state memory or even remotely mounted memory.

A second computer program product in the form of a data memory 63 may also be provided, e.g. for reading and/or storing data during execution of software instructions in the processor 30. The data memory 63 can be any combination of read and write memory (RAM) and read only memory (ROM) and may also comprise persistent storage, which, for example, can be any single one or combination of magnetic memory, optical memory, solid state memory or even remotely mounted memory. The data memory 63 may e.g. hold other software instructions 65, to improve functionality for the network device 20.

The network device 20 may further comprise an I/O interface 61 including e.g. a user interface. Other components of the network device 20 are omitted in order not to obscure the concepts presented herein. The network device 20 is in an embodiment implemented in the WCD 1, which is illustrated in Fig. 6A.

The network device 20 is in an embodiment implemented in the core network 3, such as in or by an SGSN (Serving GPRS (General Packet Radio Service) Support Node), a GGSN (Gateway GPRS Support Node), a Serving Gateway, or a Packet Data Network Gateway, which is illustrated in Fig. 6B. The network device 20 may in other embodiments be implemented in a Business Support System (BSS) device and/or in an Operational Support System (OSS) device, typically owned by the network operator owning the core network 3.

Fig. 2 is a schematic diagram illustrating, in more detail, a network roaming predictor implemented in the core network 3.

As shown in Fig. 2, a historical model 11 is stored in the core network 3 and is fed into a network roaming predictor 12. The WCD 1 reports real-time mobility related data, i.e. live data 10, such as location or radio fingerprints, to the roaming predictor 12 located in the core network 3. The core network 3 takes the historical model 11 and the live data 10 as input to predict a network roaming of the WCD 1. If a future network roaming is determined, the roaming predictor 12 sends a trigger for roaming preparation. A roaming preparation executor 13 maybe provided in the core network 13 and/or a roaming preparation executor 14 maybe provided in the WCD 1, and/or in other locations such as in the roaming core network 4. The roaming predictor 12 is implemented by means of the network roaming predictor classifier.

When the core network 3 does not have access to live data 10 from a WCD 1, e.g. when the WCD 1 is in the idle mode, the roaming predictor may be implemented in the WCD 1. When the WCD is in idle mode the location updates are less frequent compared to when the WCD is in active mode. Implementation in the WCD 1 may be an alternative implementation, or an additional implementation, to the implementation in the core network 3.

Fig. 3 is a schematic diagram illustrating, in more detail, network roaming predictor implemented in the WCD 1.

As shown in Fig. 3, the historical model 11 is stored in the core network 3 and is downloaded in the roaming predictor 12 in the WCD 1. The WCD 1 takes the historical model 11 and real-time mobility data, i.e. the live data 10, such as location or radio fmgerprints, as input to perform a roaming prediction. If future roaming is determined, the roaming predictor 12 sends a trigger for roaming preparation. The roaming preparation executor 13 may be provided in the core network 13 and/or the roaming preparation executor 14 maybe provided in the WCD 1, and/ or in other locations such as the roaming core network 4. Live data is data relevant for the WCD for which a network roaming prediction is made, which is in the classifier compared to historical data earlier collected from other WCDs. The live data is typically real-time data, but may also be recent data collected by the WCD e.g. in idle mode. Live data 10 may be obtained by e.g. receiving a report from a WCD or downloading in a WCD. A radio network and a core network get location information continuously. This data may be sent in Communication Data Records (CDRs) or other types of events. The WCD itself always knows which cell it is located in.

The instructions of the computer program 64 may comprise a further instruction to trigger a network roaming preparation when a probability for network roaming of the WCD 1 is determined to be above a threshold, based on the network roaming prediction class. This probability may also be coupled to a time, e.g. 80% likelihood that the WCD will roam in the next 10 minutes. Illustration of a situation of possible network roaming is shown in Fig. 7. A network boarder is a boarder between two network operators, wherein a user typically have subscription to only one of them, a home network, and thus needs to network roam in the other network, a visiting network. A cell of the home network adjacent a cell of the visiting network is a border cell and a cell of the home network adjacent a border cell is a neighbouring boarder cell. A border cell can thus also be a neighbouring boarder cell at the same time. Two different wireless communication devices 1 are shown in a boarder cell to illustrate that one WCD in a boarder cell may be predicted to be about to network roam whereas the other one in the same boarder cell may be predicted to be about to not network roam.

The speed and direction of the movement of the wireless communication devices 1 may be captured by time stamps for entering and leaving the neighbouring boarder cells. The likelihood that each of the subscribers is soon going to network roam may be predicted and preparation for network roaming may be triggered if the likelihood is above a certain threshold.

The network roaming preparation may comprise one or more of the following instructions to: notify a user of the WCD of the predicted network roaming, notify the WCD of the predicted network roaming, notify an application in the WCD of the predicted network roaming, notify a network roaming operator that the WCD is predicted to network roam thereto, notify a cellular roaming network that the WCD is predicted to network roam thereto, notify an application of the cellular network that the WCD is predicted to network roam, and notify the cellular network that the WCD is predicted to network roam. Knowing in advance about roaming users can result in offering a better service. For example, informing a user of a WCD about the fact that they will likely roam soon will give them the chance to prepare for it and stop their usage consciously, e.g. they can end a phone call or finish downloading before roaming to a high cost network, instead of causing an interruption with no prior notice. Depending on the operators' agreements, the operator in the visiting network may offer a few minutes of service for free, especially noticeable in video streaming use cases, so there will be no interruption in user's experience. In an embodiment having more than one visiting network operator to select among, offers of a few minutes of service for free may be useful.

By notifying the WCD of the predicted network roaming, the WCD may notify the user that roaming may happen. The user may thus avoid or end expensive downloads.

By notifying an application program in the WCD of the predicted network roaming, the application program may e.g. preload data. One example may be to buffer streamed video at an accelerated rate before roaming and/ or by enlarging the buffer size itself before roaming.

The computer program product may comprise a further instruction to trigger a network roaming prediction for the WCD 1. By not having the network device continuously predicting roaming, but only after being triggered, processing resources may be saved. The instruction to trigger a network roaming prediction may more limited comprise the instructions to determining the location of the WCD 1 to be in a network boarder cell. By continuously storing live date for e.g. the last two or three cells, utilization of a trigger only at a boarder cell maybe provided and still have live data from a neighbouring boarder cell as well as boarder cell to provide to the network roaming prediction classifier. This is a good way to reduce signalling. Another option may be to trigger a network roaming prediction when the WCD is in a Location Area close to a network border.

The instruction to determine the location of the WCD 1 may be performed by a satellite-aided positioning system, cell identity based positioning, or radio signalling based positioning, or a combination thereof. Also, location and routing area based positioning may be used when the wireless

communication terminal is in idle mode. Examples of satellite-aided positioning systems are Global Positioning System (GPS), Glonass, Beidou and Galileo systems.

Further to the prediction that a WCD 1 is about to roam, a predicted time to roam may also be provided, based on speed and direction of the wireless communication device. A prediction will be a combination of probability and time e.g. 80% in 5 minutes. An idle WCD may be activated if in an idle mode e.g. if a location area (LA) has a roaming boarder.

A historical model database maybe provided for each type of localization that the network roaming prediction classifier can use, such as e.g. one database for satellite-aided positioning, one database for cell identity based

positioning, one database for signalling based positioning and one database for mobile positioning system (MPS) that collects signalling based

positioning and provides a more detailed location. Utilization of more than one database for a network roaming prediction classifier maybe made, either sequentially or in parallel. According to an embodiment of a method by which a WCD can know its location is to use a receiver for satellite positioning signals, e.g a GPS receiver operating independently of the cellular network. Alternatively, the cellular network itself can establish the approximate location of a connected WCD using knowledge of the coverage area of each cell.

In Release 8 LTE, a WCD can be located by means of Assisted Global

Navigation Satellite System (A-GNSS) and Enhanced Cell-ID-based techniques in conjunction with a general-purpose positioning protocol known as Secure User Plane Location (SUPL), defined by the Open Mobile Alliance (OMA). SUPL operates as a service in the application layer and requires only a normal User Plane (UP) connection between a server in the network (known in OMA as a SUPL Location Platform (SLP) and in LTE as an Evolved Serving Mobile Location Centre (E-SMLC)) and the SUPL client application in the wireless communication device. A-GNSS-based positioning refers in general to any satellite-based positioning system, in conjunction with assistance provided over a terrestrial network to improve the sensitivity and/or speed of detection of the satellites.

Basic GNSS-based positioning relies on accurate knowledge of the locations of the satellites and the transmission times of their signals. With some simplification it can be said that a GNSS receiver measures the exact time at which it receives the signal of each satellite it can detect. Using this information, it is possible to calculate the location of a wireless

communication device.

The instruction to obtain a network roaming prediction class may comprise the instruction to input the live data location features to the trained network roaming prediction classifier.

The instructions to obtain a network roaming prediction class may comprise a further instruction to train the network roaming prediction classifier using historical data of wireless communication devices, about to network roam or not, and resulting in roaming and not roaming classes. Historical model data that may be used to build the historical model is user mobility data represented by either user location or radio fingerprints, e.g. Cell ID (CID) and Reference Signal Received Power (RSRP), within a period of time before the user roams to another public land mobile network

(PLMN). In order to decide if a user roams or not a time limit may be used. E.g. if a WCD has not roamed in x minutes after entering a border cell it is regarded as not roaming.

In order to obtain historical data from wireless communication devices, the ID of a roaming user and the time the user leaves the home network and registers to the visiting network may be used. Based on the ID and time, the mobility data used for roaming prediction may be extracted from the home network.

When moving to a visiting network, the WCD will try to register to that network. The visiting location register (VLR) of the visiting network will contact the home location register (HLR) of the home network to request service information about the roaming user using IMSI (International Mobile Subscriber Identity) as a user ID. Therefore by checking query records in the HLR, the roaming user ID and registration time to the visiting network may be obtained. The mobility data, e.g. user ID, time, geographical position (longitude, latitude and optionally altitude), serving cell ID, RSRP, may be stored in an OSS or BSS as PM (Performance Management) data. Alternatively the mobility data will be stored in a BSS device such as a BSS server. It should be noticed that the core network does not always possess the user mobility data in cell level, such as when a WCD is in idle mode before roaming and have no active connection to the home network. In this case, the network may ask the WCD to store the mobility data locally and report it to the network when the WCD is connected to the network in the future. Such mechanism exists in 3GPP known as Minimization of Drive Tests (MDT) which is specified in 3 GPP TS 37.320 V11.3.0. l6

By correlating the user roaming data, e.g. user ID and roaming time, and the mobility data, e.g. user ID, time, geographical position, serving cell ID and RSRP, the historical data may be constructed and the historical model may be built upon it. For initialisation of a historical model, i.e. bootstrapping, an initial location value maybe added instead of starting with an empty database. Geographical data, such as where roads/bridges exist between countries maybe utilized to add initial historical data to the model. Border cell id from operators may alternatively be used as initial historical data. The live and historical data location features may comprise information about network boarder cell, neighbouring network boarder cell, time of entry of the neighbouring network boarder cell and time of entry of the network boarder cell. A classifier 80 having such features is illustrated in Fig. 8.

Fig. 8 is a schematic diagram of a network roaming preparation classifier 80. Historical data about subscribers' movement prior to roaming, or not roaming, is used to train the network roaming preparation classifier 80. Data for all wireless communication devices entering a neighbouring cell to a bordering cell may be logged with time stamps for when the subscriber is entering and leaving the cells. This data may be used to estimate the speed that the WCD is moving towards the border cell. An assumption may be that it is more likely that wireless communication devices moving in higher speed than wireless communication devices moving in lower speed is about to roam soon, and when they are about to roam.

The live and historical data location features may comprise information about geographical position and/ or cell identity and/ or radio signalling data.

As a result of having the historical model with historical data location features for the users' mobility pattern and the roaming records, the network roaming prediction classifier may be trained using Machine Learning techniques for those users who will roam and who will not roam. Such a network roaming prediction classifier identifies roaming user mobility patterns, and given input from other users, it can determine the probability of roaming for a specific user.

A simple classification is digital in its output, classes e.g. classified into roaming or not roaming. A more complex classification may be provided having classes with likelihood for roaming. High confidence probability may e.g. be at 0.9 likelihood of roaming and a low confidence probability may e.g. be at 0.55 likelihood of roaming.

The network roaming prediction classifier may e.g. be a Support Vector Machine (SVM), a decision tree, a neural network or a Bayesian network. Machine learning based on SVM, decision tree and neural networks support may be used to support e.g. two classes, roaming and not roaming. Machine learning based on Bayesian networks and neural networks may be used to support more advanced classifiers e.g. providing probability classes between 0-1. A network roaming prediction class may be obtained by training a SVM and feeding live data to the trained SVM.

An embodiment of a method for providing network roaming predictions for wireless communication devices in a cellular network is shown in Fig. 4A.

The method is performed by the network device 20 of the cellular network and comprises the steps of: obtaining 43 live data location features for a WCD 1 in the cellular network; and obtaining 45 a network roaming prediction class, from a trained network roaming prediction classifier 80, based on the live data location features, to predict if the WCD 1 is about to network roam or not, wherein the network roaming prediction classifier 80 has been trained using historical data location features.

In one embodiment a method for providing network roaming predictions for wireless communication devices in a cellular network is shown in Fig. 4B.

The method may further comprise the step of triggering 47 a network roaming preparation when 46 a probability for network roaming of the WCD l8

1 is determined to be above a threshold, based on the network roaming prediction class.

The network roaming preparation may comprise one or more of the following: notifying a user of the WCD of the predicted network roaming, notifying the WCD of the predicted network roaming, notifying the WCD of the predicted network roaming, notifying a network roaming operator that the WCD is predicted to network roam thereto, notifying a cellular network that the WCD is predicted to network roam thereto, and notifying the cellular network that the WCD is predicted to network roam. The method may further comprise the step of triggering a network roaming prediction for the WCD 1.

The step of triggering a network roaming prediction may comprise the steps of: detecting 41 a location of the WCD 1; and determining 42 the location of the WCD 1 to be in a network boarder cell or a neighbouring network boarder cell.

The step of determining 42 may comprise determining the location of the WCD 1 to be in a network boarder cell.

The step of determining the location of the WCD 1 may be performed by satellite-aided positioning, cell identity based positioning, or radio signalling based positioning, or a combination thereof.

The step of obtaining 45 a network roaming prediction class may comprise the step of inputting 44 the live data location features to the trained network roaming prediction classifier.

The method may further comprise the step of training 40 the network roaming prediction classifier using historical data of wireless communication devices, about to network roam or not, and resulting in roaming and not roaming classes. The live and historical data location features may comprise information about network boarder cell, neighbouring network boarder cell, time of entry of the neighbouring network boarder cell and time of entry of the network boarder cell. The live and historical data location features may comprise information about geographical position and/or cell identity and/or radio signalling data.

The network roaming prediction classifier may e.g. be a Support Vector Machine (SVM), a decision tree, a neural network or a Bayesian network.

The computer program 64, 65, for determining network roaming predictions if a WCD 1 is about to network roam or not, comprises computer program code which, when run on the network device, causes the network device to: obtain live data location features for a WCD 1 in a cellular network; and obtain a network roaming prediction class, from a trained network roaming prediction classifier, based on the live data location features, resulting in a prediction if the WCD is about to network roam or not.

The computer program product 62, 63 comprises the computer program 64, 65 and the computer readable storage means on which the computer program 64, 65 is stored.

The computer program can cause the processor to execute a method according to embodiments described herein. The computer program product may further be an optical disc, such as a CD (compact disc) or a DVD (digital versatile disc) or a Blu-Ray disc.

Fig. 9 is a schematic diagram showing functional blocks of the network device 20. The modules maybe implemented as only software instructions such as a computer program executing in the network device or only hardware, such as application specific integrated circuits, field programmable gate arrays, discrete logical components, transceivers, etc. or as a combination thereof. In an alternative embodiment, some of the functional blocks may be

implemented by software and other by hardware. The modules correspond to the steps in the methods illustrated in Figs. 4A-4B, comprising a model manager 70, a detection manager 71, a trigger selector 72, a live data manager 73, a classifier manager 74, a prediction class generator 75, a probability selector 76 and a preparation manager 77. In the embodiments where one or more of the modules are implemented by a computer program, then it shall be understood that these modules do not have to correspond to programming modules, but can be written as instructions according to the programming language in which they would be implemented, since some programming languages do not typically contain programming modules. The model manager 70 is arranged to train the network roaming prediction classifier using historical data of WCDs, about to network roam or not, and resulting in roaming and not roaming classes. This module corresponds to the training step 40 of Fig 4B. This module can e.g. be implemented by the processor 30 of Fig. 5, when running the computer program. The detection manager 71 is arranged to detect a location of the WCD 1. This module corresponds to the detecting step 41 of Fig. 4B. This module can e.g. be implemented by the processor 30 of Fig. 5, when running the computer program.

The trigger selector 72 is arranged to determine if the WCD is close to a network boarder. This module corresponds to the determining step 42 of Fig. 4B. This module can e.g. be implemented by the processor 30 of Fig. 5, when running the computer program.

The live data manager 73 is arranged to obtain live date from the WCD. This module corresponds to the obtaining step 43 of Figs. 4A and 4B. This module can e.g. be implemented by the processor 30 of Fig. 5, when running the computer program.

The classifier manager 74 is arranged to classify if a WCD is about to network roam or not. This module corresponds to the inputting step 44 of Fig. 4B. This module can e.g. be implemented by the processor 30 of Fig. 5, when running the computer program. The prediction class generator 75 is arranged to predict if a WCD is about to network roam or not. This module corresponds to the obtaining step 45 of Figs. 4A and 4B. This module can e.g. be implemented by the processor 30 of Fig. 5, when running the computer program. The probability selector 76 is arranged to determine when the probability of roaming is above a threshold. This module corresponds to the condition 46 of Fig. 4B. This module can e.g. be implemented by the processor 30 of Fig. 5, when running the computer program.

The preparation manager 77 is arranged to trigger preparation for network roaming. This module corresponds to the triggering step 47 of Fig. 4B. This module can e.g. be implemented by the processor 30 of Fig. 5, when running the computer program.

A network device 20 for a cellular network 5, the network device arranged to determine network roaming predictions if a WCD 1 is about to network roam or not, the network device comprising: a live data manager 73 configured to obtain live data location features for a WCD 1 in the cellular network; and a prediction class generator 75 configured to obtain a network roaming prediction class, from a trained network roaming prediction classifier, based on the live data location features, resulting in a prediction if the WCD is about to network roam or not.

The network device may further comprise a preparation manager 77 configured to trigger a network roaming preparation when a probability for network roaming of the WCD 1 is determined to be above a threshold, based on the network roaming prediction class.

The preparation manager 77 maybe configured to perform one or more of the following: notify a user of the WCD of the predicted network roaming, notify the WCD of the predicted network roaming, notify an application program in the WCD of the predicted network roaming, notify a network roaming operator that the WCD is predicted to network roam thereto, notify a cellular roaming network that the WCD is predicted to network roam thereto, notify an application program in the cellular network of the predicted network roaming, and notify the cellular network that the WCD is predicted to network roam.

The network device may further comprise a trigger selector 72 configured to trigger a network roaming prediction for the WCD 1.

The trigger selector 72 may further be configured to: detect a location of the WCD 1; and determine the location of the WCD 1 to be in a network boarder cell or a neighbouring network boarder cell.

The trigger selector 72 may be configured to determining the location of the WCD 1 to be in a network boarder cell.

The trigger selector 72 may be configured to determine the location of the WCD 1 by satellite-aided positioning, cell identity based positioning, or radio signalling based positioning, or a combination thereof.

The network device may further comprise a classifier manager 74 configured to input the live data location features to the trained network roaming prediction classifier. The network device 20 may further comprise a model manager 70 configured to train the network roaming prediction classifier using historical data of wireless communication devices, about to network roam or not, and resulting in roaming and not roaming classes.

The live and historical data location features may comprise information about network boarder cell, neighbouring network boarder cell, time of entry of the neighbouring network boarder cell and time of entry of the network boarder cell. The live and historical data location features may comprise information about geographical position and/or cell identity and/or radio signalling data.

The network roaming prediction classifier may e.g. be a Support Vector Machine (SVM), a decision tree, a neural network or a Bayesian network. The invention has mainly been described above with reference to a few embodiments. However, as is readily appreciated by a person skilled in the art, other embodiments than the ones disclosed above are equally possible within the scope of the invention, as defined by the appended patent claims.

Claims

1. A method for providing a network roaming prediction for a wireless

communication device in a cellular network (5), the method being performed by a network device (20) of said cellular network and comprising the steps of: obtaining (43) live data location features for a wireless communication device (1) connected to said cellular network; and obtaining (45) a network roaming prediction class, from a trained network roaming prediction classifier (80), based on said live data location features, to predict if said wireless communication device (1) is about to network roam or not, wherein the network roaming prediction classifier (80) has been trained using historical data location features.

2. The method according to claim 1, further comprising the step of: triggering (47) a network roaming preparation when (46) a probability for network roaming of said wireless communication device (1) is determined to be above a threshold, based on said network roaming prediction class.

3. The method according to claim 2, wherein said network roaming preparation comprises one or more of the following: notifying a user of said wireless communication device of the predicted network roaming, notifying said wireless communication device of the predicted network roaming, notifying an application of said wireless

communication device of the predicted network roaming, notifying a network roaming operator that said wireless communication device is predicted to network roam thereto, notifying a cellular roaming network that said wireless communication device is predicted to network roam thereto, notifying an application of said cellular network that said wireless communication device is predicted to network roam, and notifying said cellular network that said wireless communication device is predicted to network roam.

4. The method according to any one of the preceding claims, further comprising the step of: triggering a network roaming prediction for said wireless communication device

(i).

5. The method according to claim 4, wherein said step of triggering a network roaming prediction comprises the steps of: detecting (41) a location of said wireless communication device (1); and determining (42) said location of said wireless communication device (1) to be in a network boarder cell or a neighbouring network boarder cell.

6. The method according to claim 5, wherein said step of determining (42) comprises determining said location of said wireless communication device (1) to be in a network boarder cell.

7. The method according to claim 5 or 6, wherein said step of determining said location of said wireless communication device (1) is performed by satellite-aided positioning, cell identity based positioning, or radio signalling based positioning, or a combination thereof.

8. The method according to any one of the preceding claims, wherein said step of obtaining a network roaming prediction class comprises the step of: inputting (44) said live data location features to said trained network roaming prediction classifier.

9. The method according to any one of the preceding claims, further comprising the step of: training (40) said network roaming prediction classifier using historical data of wireless communication devices, about to network roam or not, and resulting in roaming and not roaming classes.

10. The method according to any one of the preceding claims, wherein said live and historical data location features comprise information about network boarder cell, neighbouring network boarder cell, time of entry of said neighbouring network boarder cell and time of entry of said network boarder cell.

11. The method according to any one of the preceding claims, wherein said live and historical data location features comprise information about a geographical position and/or cell identity and/or radio signalling data.

12. The method according to any one of the preceding claims, wherein said classifier is a Support Vector Machine (SVM), a decision tree, a neural network or a Bayesian network.

13. A network device (20) for a cellular network (5), said network device arranged to predict if a wireless communication device (1) is about to network roam, the network device comprising: a processor (30); and a computer program product (62) storing instructions that, when executed by the processor, causes the network device (20) to: obtain live data location features for a wireless communication device (1) in said cellular network; and obtain a network roaming prediction class, from a trained network roaming prediction classifier, based on said live data location features, resulting in a prediction if said wireless communication device is about to network roam or not.

14. The network device (20) according to claim 13, wherein the instructions comprise a further instruction to: trigger a network roaming preparation when a probability for network roaming of said wireless communication device (1) is determined to be above a threshold, based on said network roaming prediction class.

15. The network device (20) according to claim 13, wherein said network roaming preparation comprises one or more of the following instructions to: notify a user of said wireless communication device of the predicted network roaming, notify said wireless communication device of the predicted network roaming, notify an application in said wireless communication device of the predicted network roaming, notify a network roaming operator that said wireless communication device is predicted to network roam thereto, notify a cellular roaming network that said wireless communication device is predicted to network roam thereto, notify an application of said cellular network that said wireless communication device is predicted to network roam, and notify said cellular network that said wireless communication device is predicted to network roam.

16. The network device (20) according to any one of claims 13-15, wherein the instructions comprise a further instruction to: trigger a network roaming prediction for said wireless communication device (1).

17. The network device (20) according to claim 16, wherein said instruction to trigger a network roaming prediction comprises the instructions to: detect a location of said wireless communication device (1); and determine said location of said wireless communication device (1) to be in a network boarder cell or a neighbouring network boarder cell.

18. The network device (20) according to claim 17, wherein said instruction to trigger a network roaming prediction comprises the instructions to determining said location of said wireless communication device (1) to be in a network boarder cell.

19. The network device (20) according to claim 17 or 18, wherein said instruction to determine said location of said wireless communication device (1) is performed by satellite-aided positioning, cell identity based positioning, or radio signalling based positioning, or a combination thereof.

20. The network device (20) according to any one of claims 13-19, wherein said instruction to obtain a network roaming prediction class comprises the instruction to: input said live data location features to said trained network roaming prediction classifier.

21. The network device (20) according to any one of claims 13-20, wherein the instructions comprise a further instruction to: train said network roaming prediction classifier using historical data of wireless communication devices, about to network roam or not, and resulting in roaming and not roaming classes.

22. The network device (20) according to any one of claims 13-21, wherein said live and historical data location features comprise information about network boarder cell, neighbouring network boarder cell, time of entry of said neighbouring network boarder cell and time of entry of said network boarder cell.

23. The network device (20) according to any one of claims 13-22, wherein said live and historical data location features comprise information about geographical position and/or cell identity and/or radio signalling data.

24. The network device (20) according to any one of claims 13-23, wherein said classifier is a Support Vector Machine (SVM), a decision tree, a neural network or a Bayesian network.

25. A network device (20) for a cellular network (5), said network device arranged to predict if a wireless communication device (1) is about to network roam, the network device comprising: a live data manager (73) configured to obtain live data location features for a wireless communication device (1) in said cellular network; and a prediction class generator (75) configured to obtain a network roaming prediction class, from a trained network roaming prediction classifier, based on said live data location features, resulting in a prediction if said wireless communication device is about to network roam or not.

26. The network device (20) according to claim 25, further comprising: a preparation manager (77) configured to trigger a network roaming preparation when a probability for network roaming of said wireless communication device (1) is determined to be above a threshold, based on said network roaming prediction class.

27. The network device (20) according to claim 26, wherein said preparation manager is configured to perform one or more of the following: notify a user of said wireless communication device of the predicted network roaming, notify said wireless communication device of the predicted network roaming, notify an application in said wireless communication device of the predicted network roaming, notify a network roaming operator that said wireless communication device is predicted to network roam thereto, notify a cellular roaming network that said wireless communication device is predicted to network roam thereto, notify an application of said cellular network that said wireless communication device is predicted to network roam, and notify said cellular network that said wireless communication device is predicted to network roam.

28. The network device (20) according to any one of claims 25-27, further comprising: a trigger selector (72) configured to trigger a network roaming prediction for said wireless communication device (1).

29. The network device (20) according to claim 28, wherein said trigger selector (72) further is configured to: detect a location of said wireless communication device (1); and determine said location of said wireless communication device (1) to be in a network boarder cell or a neighbouring network boarder cell.

30. The network device (20) according to claim 29, wherein trigger selector (72) is configured to determining said location of said wireless communication device (1) to be in a network boarder cell.

31. The network device (20) according to claim 29 or 30, wherein said trigger selector (72) is configured to determine said location of said wireless communication device (1) by satellite-aided positioning, cell identity based positioning, or radio signalling based positioning, or a combination thereof.

32. The network device (20) according to any one of claims 25-30, further comprising: a classifier manager (74) configured to input said live data location features to said trained network roaming prediction classifier.

33. The network device (20) according to any one of claims 25-32, further comprising: a model manager (70) configured to train said network roaming prediction classifier using historical data of wireless communication devices, about to network roam or not, and resulting in roaming and not roaming classes.

34. The network device (20) according to any one of claims 25-33, wherein said live and historical data location features comprise information about network boarder cell, neighbouring network boarder cell, time of entry of said neighbouring network boarder cell and time of entry of said network boarder cell.

35. The network device (20) according to any one of claims 25-34, wherein said live and historical data location features comprise information about geographical position and/ or cell identity and/ or radio signalling data.

36. The network device (20) according to any one of claims 25-34, wherein said classifier is a Support Vector Machine (SVM), a decision tree, a neural network or a Bayesian network.

37. A computer program (64, 65) for predicting if a wireless communication device (1) is about to network roam, the computer program comprising computer program code which, when run on a network device, causes the network device to: obtain live data location features for a wireless communication device (1) in a cellular network; and obtain a network roaming prediction class, from a trained network roaming prediction classifier, based on said live data location features, resulting in a prediction if said wireless communication device is about to network roam or not.

38. A computer program product (62, 63) comprising a computer program (64, 65) according to claim 37 and a computer readable storage means on which the computer program (64, 65) is stored.