METHOD AND SYSTEM FOR THE EXPERIMENTAL EVALUATION OF OPTIMIZATION ACTIONS ON A MOBILE COMMUNICATIONS NETWORK

Description

DESCRIPTION Technical field This description concerns the mobile communications networks sector. In particular, this description concerns the optimization of networks of 5 mobile radio communications, particularly but not limited to 5G or future 5G networks generations. Specifically, this description concerns a method, and a related system, to evaluate the effectiveness of optimization actions on communication networks mobile radios (mobile radio networks). Technical context 10 In order to optimize the Quality of Service (QoS) experienced by users in mobile radio networks, with particular reference to user throughput (the amount of data received - bits received correctly - from a user of a user equipment within a network cell in a given period of time), it is useful to compensate asymmetries in the distribution of network traffic on the cells of the mobile radio network. 15 Such asymmetries in the distribution of network traffic can be compensated for by acting on modifiable parameters of the Radio Access Network (RAN). The actions on the RAN editable parameters include actions on configuration parameters of the network cells, in order to obtain the right compromise between the signal-to- interference-plus-noise (SINR) and user traffic distribution per network cell. 20 Network cells include radio transceiver stations comprising one or Multiple antennas. Optimization actions on cell configuration parameters network may involve an optimization of the cell antenna parameters network (transmission power, electrical inclination and azimuth of the antennas and any other available parameter that affects the radiation pattern and the 25 spatial distribution of power, including parameters that control the diagrams of radiation for active antennas, for beamforming techniques, typical of networks 5G), as well as an optimization of the parameters that influence the selection procedures and handover of network cells, in order to make traffic volumes and the number of users served per cell (traffic distribution in the cells) as much as possible 30 homogeneous possible. The throughput of the users depends on the SINR guaranteed by the mobile radio network in a specific pixel (one of the elementary territorial units in which it is conventionally divided a territory covered by a mobile radio network), but it is also inversely proportional to the number of active users in each single cell of the network. An action 5 on the coverage area of ​​a network cell, carried out through modifications of the pointing of the antenna(s) of the network cell(s) and / or the transmission power and / or the parameters of the network cell selection and handover procedure, can then allow for a better distribution of users between cells, even if at potential at the expense of optimal planning of the SINR offered, in order to avoid critical areas 10 in terms of low user throughput due to heavy concentrations of users in the individual cells of the network. The same applies to stability in terms of latency, which increases as the number of users increases. In fact, a more homogeneous distribution of users in the various cells of the network allows you to avoid areas of greater QoS degradation, in terms of throughput 15 users. In network planning there is a tendency to try to improve the most ° criticisms, for example maximizing the throughput of users of the worst fifth (5 ) ° percentile (in general, the 5th percentile is the value of a certain parameter – the throughput of users in the case considered here – in which only 5% of the population considered – 20 users of user equipment in this case – finds lower values ​​than this parameter (i.e. lower user throughput values) refers to more areas congested, since in areas with good QoS the throughput deteriorates of user is not perceived by users as critically as it is instead perceived by users in congested areas. 25 At the same time, optimization may include shutdown based on the traffic of network sites (base stations) to reduce energy consumption and therefore operating costs for telecommunications network operators (Telcos). The right trade-off between QoS and energy consumption can be a goal of network optimization. 30 To redistribute user traffic across the network cells, it is possible to adopt two different network solutions: coverage and capacity optimization (Coverage and Capacity Optimization (CCO), which involves actions on the antenna parameters of the cells network such as pointing and power - and Mobile Load Balancing (MLB), which operates on parameters of the cell selection and handover procedures. 5 To proceed with an optimization of the RAN parameters, a map is required of traffic sources in the territory. This map of traffic sources can be obtained with the Minimization of Drive Test (MDT) traffic source profiles, or from pixel-mapped cell-by-cell traffic measurements, using information on the building density and road conditions in the area covered by the network 10 mobile radio. Using this map of traffic sources distribution on the territory, it is possible to estimate the optimal configuration of the mobile radio network for the generic traffic profile and the generic user profile considered. MLB-based network optimization and site on / off 15 network can be static, that is, an optimal fixed network configuration is calculated for a reference distribution of traffic sources, or dynamic, that is a different network configuration can be expected: a network configuration for the peak hours, with user asymmetry that tends to rebalance between cells, and a off-peak network configuration, where the network configuration tends to 20 maximize network capacity, as there are no off-peak hours criticality, given the low number of users and their homogeneous distribution. The network will choose one or the other network configuration based on the user traffic profile of a or more cells. A redistribution of users across network cells can be detrimental in 25 terms of network capacity but, as mentioned, it can lead to a mitigation of significant criticalities in some areas, improving the QoS that the mobile network provides to the most disadvantaged users, a criterion according to which the network is planned. The action of modifying RAN parameters for network optimization purposes it is studied and defined with simulation tools. 30 For example, WO 2022 / 207402 describes a method, implemented by a data processing system, to adjust modifiable parameters of network cells of a self-organizing cellular mobile communications network, comprising: recover a current configuration of network cells currently deployed in the field, including editable parameters; explore different network cell configurations, each 5 differing from the recovered current configuration and other different configurations for a change in the value of at least one of the modifiable parameters of at least one network cell; evaluate the different network cell configurations explored. When an adequate degree of goodness is evaluated for a given cell configuration of network, said new network cell configuration is automatically deployed 10 in the field by modifying one or more modifiable parameters of the network cells. Another example is given in Marco Skocaj et al., “Cellular Network Capacity and Coverage Enhancement with MDT Data and Deep Reinforcement Learning”, Computer Communications Volume 195, November 1, 2022, pages 403–415. The authors study a MDT-based Deep Reinforcement Learning (DRL) algorithm for 15 optimize coverage and capacity by tuning the antenna inclinations to a aggregate of cells from a cellular network of an Italian telecommunications operator. MDT data, electromagnetic simulations and key performance indicators of the network (Key Performance Indicator, KPI) are used together to define a simulated network environment for training a Deep Q-Network agent 20 (DQN). EP 4175370 describes that to maximize energy savings in a RAN comprising cells, using a trained model an action is determined optimal among the actions including the switching on of one or more cells, the switching off of one or more cells and doing nothing, which maximizes a long-term reward on the 25 tradeoff between throughput and power. The trained model accepts as input a load estimate. The trained model can be updated online using the load, throughput and energy consumption measurement results. Summary As mentioned above, the action of modifying RAN parameters for the purpose of 30 of the optimization of the mobile radio network is studied and defined with tools of simulation. From the tests carried out on the mobile radio network in operation in the field it is relatively simple to carry out analytical evaluations to define the best configuration of net. 5 However, the Applicant understood that the actual operating conditions of the network may not be properly considered in the process of design, due to the partial randomness of some aspects. Among these aspects the evaluation of simulated and / or real network performance includes: distribution of users on the territory, their mobility conditions and profiles 10 path loss: in the network optimization process a validation experimental investigation of the correct modelling of these effects is therefore necessary. The Applicant felt that a field assessment was necessary of the effectiveness of the design process and changes in the configuration of network applied to the mobile radio network. This evaluation can be carried out after having 15 implemented the network configuration change in the field, analyzing statistical counters and comparing them with those acquired before implementation of network optimization. The Applicant noted that known solutions, such as those described in WO 2022 / 207402, in the article by Marco Skocaj et al., and in EP 4175370, do not concern 20 such experimental verification / validation of the effectiveness of the network configuration modified implemented. One purpose of such verification should be to demonstrate that the phase of optimization has achieved the expected results, particularly in terms of improvement of QoS, for example as a result of an improvement in the throughput of 25 users. The Applicant observed that when changes to the configuration of network are implemented on the mobile radio network in the field, it is particularly challenging to define a metric to evaluate the positive effects, or rather the ineffectiveness, on the QoS of the implemented optimized network configuration, in 30 particular on user throughput. In particular, the distribution of users (and therefore of the sources of traffic) on the territory represents the most evident and important source of uncertainty present in the optimization process procedure. It is not in fact it is simple to compare, by analyzing the network statistics counters, the QoS that the network mobile is able to provide users with operating conditions before and after the 5 network configuration change, because traffic distribution profiles offered on the territory changes significantly and the distribution of the sources of traffic has a direct impact on the detected user throughput, which, as mentioned, plays an important role in quantifying QoS. One purpose of this description is to define a methodology for 10 mitigate the effects of changes in user distribution and volumes of traffic on the territory covered by a mobile radio network, in order to be able to carry out a verification process capable of reliably assessing the effectiveness of the modified network configuration implemented under normal operating conditions of the mobile radio network. 15 In essence, this description proposes a method, and a related system, able to make the verification process independent of such variations in the distribution of users and traffic volumes on the territory, thanks to the introduction - in the comparison between experimental data collected before and after the implementation, in the mobile network, network configuration modification - appropriate criteria for 20 homogenize the distribution of users, and therefore of network traffic, before and after implementing the network configuration change. This way it is possible ensure correct modelling of path losses, mobility and other aspects of modeling, regardless of user distribution. According to one aspect of the present invention, a method is proposed 25 implemented on computers to evaluate the variation of Quality of Service, QoS, offered by a mobile communications network, resulting from a change in the configuration parameters of one or more network cells of the communications network mobile. The method includes the definition of a network cell aggregate 30 comprising one or more network cells and neighboring network cells adjacent to said one or more network cells. Advantageously, the network cell aggregate is defined in such a way that correspond to a mobile communications network coverage area that remains substantially constant before and after changing the configuration parameters of one or more network cells. 5 The method comprises the acquisition of network measurements over an area of the mobile communications network corresponding to the network cell cluster. Acquisition of network measurements includes: acquiring, during a first interval of acquisition time before said configuration parameter modification, first indicators of network traffic quantity and early QoS indicators, and acquire, during 10 a second acquisition time interval after said parameter modification configuration, second network traffic quantity indicators and second QoS indicators. The method includes: deriving from the first indicators of the amount of network traffic acquired a first distribution of the amount of network traffic during the first acquisition time interval, and derive from the second quantity indicators of 15 network traffic acquired a second distribution of the amount of network traffic during the second acquisition time interval. A reference distribution of the amount of network traffic is obtained, suitable to provide a baseline distribution of the amount of network traffic in the cluster network cells during a reference time interval. 20 The method also includes: calculating a first reward based on said first QoS indicators and calculate a second reward based on said second QoS indicators, and evaluate the change in QoS based on a comparison between the first reward and second reward. The calculation of the first reward and the second reward includes the 25 calculation of first weights to be applied to the first QoS indicators and of second weights to be applied to the first QoS indicators apply to the second QoS indicators, called first weights and second weights being calculated so as to make the first distribution of the amount of network traffic and the second distribution of the amount of network traffic corresponding to the distribution of reference to the amount of network traffic. 30 Network traffic quantity indicators can be any of, or one or more of: network traffic volume indicators (e.g. expressed in kbit), indicators of user presence in network cells (i.e. users for whom the mobile network records a last event in the network cells, derived from location cues of the user at the time of recording the last event), indicators of the number 5 of active users in the network cells. The first and second QoS indicators can include indicators of a distribution of network traffic quantities as a function of the user throughput that the mobile communications network is able to provide to the users. In embodiments of the present invention, the first range of 10 acquisition time includes a first succession of first time intervals of elementary acquisition, and the second acquisition time interval includes a second succession of elementary second acquisition time intervals. first indicators of network traffic quantity and first QoS indicators are acquired during each first elementary acquisition time interval of the 15 first succession, and the second indicators of network traffic quantity and the second QoS indicators are acquired during each second time interval of elementary acquisition of the second sequence. In embodiments of the present invention, the first distribution of the amount of network traffic during the first acquisition time interval is a 20 distribution of the first elementary acquisition time intervals as a function of the first indicators of network traffic quantity, and the second distribution of quantity of network traffic during the second acquisition time interval is a distribution of the second elementary acquisition time intervals as a function of the early indicators of network traffic volume. 25 In embodiments of the present invention, obtaining a distribution reference amount of network traffic includes: - define a reference acquisition time interval, in which said reference acquisition time interval includes a third succession of third elementary acquisition time intervals; 30 - acquire, during each third elementary time interval of acquisition of the third succession, third indicators of network traffic quantity and third QoS indicators, and - derive from the third indicators of network traffic quantity the distribution of reference of the amount of network traffic, the reference distribution of the 5 amount of network traffic being a distribution of the third time intervals of elementary acquisition based on third party indicators of network traffic quantity. In other embodiments of the present invention, the distribution of reference of the amount of network traffic is the first distribution of the amount of network traffic during the first acquisition time interval, and obtaining 10 a reference distribution of the amount of network traffic includes derived from the first indicators of network traffic quantity the first distribution of the quantity of network traffic during the first acquisition time interval. The first weights to be apply to the first QoS indicators are unit weights and the second weights are calculated so as to match the second distribution of the amount of network traffic 15 at the first distribution of the amount of network traffic. In embodiments of the present invention, the weights are calculated in so as to obtain, from the distribution of the second elementary time intervals of acquisition based on the first indicators of network traffic quantity, a reshaped distribution having a shape corresponding to one of: the distribution 20 of the third elementary acquisition time intervals as a function of the third indicators of the amount of network traffic, the distribution of the first elementary time intervals acquisition based on the first indicators of network traffic quantity. In embodiments of the present invention, the calculation of the weights to be apply to second network measurements includes: 25 - define a plurality of classes of network traffic quantity with respect to the distribution of the amount of reference network traffic, each quantity class of network traffic of said plurality of classes of network traffic quantity corresponding to a respective amount of reference network traffic; - classify each of the first elementary acquisition time intervals 30 and each of the second elementary acquisition time intervals in a respective class of said plurality of classes according to the respective first and second indicators of amount of network traffic; - determine a number of first elementary acquisition time intervals and second elementary acquisition time intervals in each class of said 5 plurality of classes, and - calculate the weights to be applied to the second network measurements includes be based on the given number of first elementary acquisition time intervals and second elementary acquisition time intervals in each class of said plurality of classes. 10 In embodiments of the present invention, the method comprises aggregate, across all network cells of the cell cluster, the first indicators of quantity of network traffic and the first acquired QoS indicators, and the second indicators of quantity of network traffic and the second acquired QoS indicators. In embodiments of the present invention, the first and second indicators 15 of the amount of network traffic, and the first and second QoS indicators are acquired from the mobile communications network as aggregated data at the level of each cell cell cluster network. In other embodiments of the present invention, the first and second indicators of the amount of network traffic, and the first and second QoS indicators 20 are acquired as georeferenced data, in particular MDT data. The method can also include the aggregation of georeferenced data acquired at the sub-level coverage area areas of each network cell of the cell cluster. According to another aspect of the present invention, a system is proposed to evaluate the variation of the Quality of Service, QoS, offered by a network 25 mobile communication, resulting from a change in configuration parameters of one or more network cells of the mobile communications network. The system is configured to acquire network measurements over an area of the mobile communications network corresponding to a cluster of network cells comprising one or more network cells and neighboring network cells adjacent to the one or more cells 30 network. Advantageously, the network cell cluster corresponds to a coverage area of the mobile communications network which remains substantially constant before and after the variation of the configuration parameters of one or more network cells. The system is configured to acquire said network measurements by: acquisition, during a first acquisition time interval prior to said 5. Changing configuration parameters, first indicators of the amount of network traffic and first QoS indicators, and acquiring, during a second time interval of acquisition after said configuration modification parameters, second indicators of the amount of network traffic and second QoS indicators. The system is also configured to obtain a reference distribution 10 of the amount of network traffic required to provide a reference distribution of the amount of network traffic in the network cell cluster during a time interval of reference. The system is also configured to: - derive from the first indicators of network traffic quantity a first 15 distribution of the amount of network traffic during the first time interval of acquisition; - derive from the second indicators of the quantity of network traffic acquired a second distribution of the amount of network traffic during the second interval acquisition time; 20 - calculate a first reward based on said first QoS indicators and calculate a second reward based on said second QoS indicators; - evaluate the QoS variation based on the comparison between the first reward and the second reward. The system is configured to calculate the first reward and the second 25 reward by calculating first weights to apply to the first and second QoS indicators weights to be applied to the second QoS indicators, the first weights and the second weights being calculated to make the first distribution of the amount of network traffic and the second distribution of the amount of network traffic corresponding to the reference distribution of the amount of network traffic. The method and system of this disclosure are intended to validate experimental mobile network optimization actions involving changes in mobile network configuration in the field. An action of optimization, i.e. a change in the mobile network configuration, which is 5 experimentally evaluated to be such as to bring about an improvement in QoS is maintained, while an optimization action that does not bring an improvement significant QoS degradation (not to mention a QoS deterioration) can be discarded. After the implementation of a change in the mobile network network configuration (e.g., changes in electrical inclination and / or 10 in the antenna azimuth, in the transmitted power, in the selection / reselection thresholds of network cells), the values ​​of network performance indicators (KPIs) are evaluated which, before the network configuration change, were unsatisfactory and which led to the execution of the network optimization action: if the new KPI values ​​are worse than before (or possibly even if the new KPI values 15 are not significantly better than before), mobile network configuration can be prudently restored to the previous network configuration. As is known, the network optimization process is based at least in part on simulations (since it is not practical to continuously act on the mobile network just for try to see if a certain change in network configuration works or 20 less). Thanks to the method and system of this disclosure, if one evaluates experimentally that the QoS variation after a change in the configuration network is not satisfactory (because the QoS after the change has not changed significantly) significant, not to mention the QoS after the change is worse than before), then yes 25 may take one or more of the following actions. One possible action is to rerun the algorithm optimization / simulation by modifying: - the variability ranges of the modifiable parameters (e.g. inclinations electrical and / or azimuth of the antennas, transmitted power, selection / reselection thresholds 30 of the network cells), so as to increase the degrees of freedom of the algorithm network optimization / simulation; - the set of modifiable parameters that the algorithm network optimization / simulation can consider (for example in the previous execution of the optimization / simulation algorithm were not taken into account 5 consideration of all possible modifiable parameters); - the set of network cells whose parameters can be modified; - the importance (weights) assigned to the different elements (e.g. volume of traffic, user throughput) considered in the calculation of the reward / cost function for network configurations before and after the change, thus modifying the 10 goals set for the optimization / simulation algorithm; - as a last resort, modify the optimization / simulation algorithm (e.g. example, in case of machine-based optimization / simulation algorithms learning / artificial intelligence, this may involve retraining of the algorithm). 15 Brief description of the drawings Features and benefits of the solution described here, including those above mentioned, will appear clearer by reading the following detailed description of exemplary and non-limiting embodiments. For a better For understandability, it is recommended to read the following description with reference to the 20 attached drawings, in which: - Fig. 1 is a pictorial view of an exemplary geographical area of ​​interest covered by network cells of a mobile radio network involved in a procedure network optimization; - Fig. 2 summarizes, in the form of an activity diagram, the activities of a 25 method according to an embodiment of the present invention; - Fig. 3 shows the first and second acquisition time intervals during which experimental data are acquired (network measurements and key indicators of performance - Key Performance Indicator, KPI), before and after the implementation of a changing the network configuration in the mobile radio network in Fig. 1; - Fig. 4 schematizes a generic acquisition time interval elementary (each elementary acquisition time interval being a “Record Of Period”, in short “ROP”) of the first and second time interval of acquisition of Fig. 3; 5 - Fig. 5 shows an example distribution of traffic volume (TrV (in kB, on the ordinate) as a function of the user throughput (ThrP in kbit / s, on the abscissa); - Fig. 6A illustrates a classification of the measurements acquired in the intervals of elementary acquisition time (ROP) before implementing the change of network configuration in classes of number of users (traffic volume), and a 10 corresponding traffic distribution profile; - Fig. 6B illustrates a classification of the measurements acquired in the intervals of Elementary acquisition time (ROP) after the implementation of the change of network configuration, in traffic volume classes, and a corresponding profile traffic distribution; 15 - Fig. 6C represents the classification of the measures of Fig. 6B after a weighting action to make the traffic distribution profile in the acquisition time interval following the change implementation of the network configuration comparable with the traffic distribution profile in the acquisition time interval before the change is implemented 20 of the network configuration; - Fig. 7 shows schematically, in terms of functional blocks, some modules of a system according to an embodiment of the present invention for evaluate the effectiveness of a network configuration change; - Fig. 8 shows, in addition to the acquisition time intervals of Fig. 3, a 25 reference acquisition time interval during which data is acquired experimental data (network measurements and KPIs), according to another embodiment of the present invention; - Fig. 9A shows a classification in classes of number of users (volume of traffic) of the measurements acquired in the elementary acquisition time intervals of 30 reference (ROP) of the reference acquisition time interval of Fig. 8; - Fig. 9B and Fig. 9C show classifications into traffic volume classes of the measurements acquired in the elementary acquisition time intervals (ROP) before and, respectively, after implementing the configuration change network in terms of traffic volume, and a corresponding distribution profile 5 of traffic; - Fig. 10 schematizes the areas of the coverage area of ​​a network cell before and after a change in network configuration, and - Fig. 11 outlines a definition of classes to classify measures georeferenced network traffic (average number of users) before and after a change 10 of the network configuration. Detailed description of exemplary embodiments The disclosure in this paper proposes a method, and a related system, to evaluate the effectiveness of implemented network configuration changes in a mobile radio network as a consequence of network optimization actions. 15 The method and system described can be applied in the context of improving or optimizing the QoS of a mobile radio network in an area geographic area of ​​interest covered by the mobile radio network. Referring to Fig. 1, a (portion of) is schematically illustrated a) mobile radio network, in particular a cellular mobile network (for short, "mobile network"), 20 globally referred to as 100. The mobile network 100 comprises a plurality of equipment cellular communication 110 or base transceiver stations (for short, "base stations"), for example eNodeB or gNodeB, each of which provides radio coverage on a respective portion of area 115 of a geographical area covered by the mobile network. In the 25 exemplary and simplified scenario considered here, each base station 110 is associated with a respective network cell, which represents the portion of the coverage area base station radio 115, however, in practical scenarios, each base station 110 may be associated with a plurality of network cells, for example three network cells. According to one embodiment of the present invention, illustrated by way of example For example, each network cell 115 has a hexagonal shape. In practice, however, the shape of the network cells can differ significantly from an ideal hexagonal shape, for example due to geographical and / or propagation characteristics or constraints of the area in which the cell is located. 5 According to one embodiment of the present invention, each base station 110 includes one or more electronic devices (not shown). Examples of electronic devices include, but are not limited to, transceivers and computers of digital signals. According to one embodiment of the present invention, Each base station 110 includes one or more antennas. 10 According to one embodiment of the present invention, a station base 110 allows UE user equipment of users located inside of the respective network cell 115 (and that are connecting / connected to the mobile network 100) to exchange data traffic (for example for web browsing, for sending e-mail, for voice or multimedia data traffic). The EU user equipment can 15 include for example personal communication devices owned by users of the mobile network 100 (users are for example subscribed to services offered by the network mobile 100). Examples of EU user equipment include, but are not limited to a, cell phones, smartphones, tablets, personal digital assistants, and computers. According to one embodiment of the present invention, the base stations 20 110 and their corresponding network cells 115 are part of a radio access network (Radio Access Network, RAN) of the mobile network 100. In the following a base station 110 and its corresponding network cell 115 will be referred to as the network cell 115 for short. network 115. The RAN may be based on any suitable radio access technology 25 (Radio Access Technology, RAT). Examples of RAT include, but are not limited to, UTRA (UMTS Terrestrial Radio Access), WCDMA (Wideband Code Division Multiple Access, code division multiple access) broadband), CDMA2000, LTE (Long Term Evolution), LTE-A (LTE-Advanced) and NR (New Radio). 30 According to one embodiment of the present invention, the RAN is coupled in communication with one or more core networks, such as the core network 120. The core network 120 can be any type of network configured to provide functionality aggregation, authentication, call control / switching, charging, invocation of services, gateways and subscriber databases, or at least a subset (i.e. 5 one or more) of them. According to one embodiment of the present invention, the core network 120 includes a 4G / LTE core network, or a 5G core network, or a core network 6G. According to one embodiment of the present invention, the core network 10 120 is coupled in communication with other communication and / or data networks, such as Internet and / or public switched telephone networks (not shown). The mobile network 100 may in particular be a self-organizing network (Self- Organizing Network (SON). The mobile radio network can for example be a network 5G mobile or a 6G mobile network or a next-generation mobile network, which adopts the 15 SON paradigm. A SON implements SON 125 functionality that allows you to perform a CCO - MLB of the mobile network 100, or features that allow you to set (i.e. tune or adjust) one or more operating parameters of the network cells 115 (of followed, cell parameters) to maximize network performance towards the users of the 20 mobile network on the field. According to one embodiment of the present invention, the parameters of cell of each network cell define a configuration of that network cell (of followed, cell configuration), and the cell configurations of the network cells of the mobile network define as a whole a mobile network configuration (of 25 followed, mobile network configuration or network configuration). According to one embodiment of the present invention, the parameters of cell include, but are not limited to, one or more antenna parameters and parameters cell selection / transfer. Examples of antenna parameters include, but are not are limited to, transmitted power, electrical inclination of the antenna, azimuth, antenna gain and radiation pattern (e.g. pointing direction, directivity and width of one or more lobes of the lobe pattern shown by the model antenna radiation). Examples of cell selection / transfer parameters include, but are not limited to, layer priority, minimum signal level for 5 layer, intra-layer cell offset. According to the present invention, and as described in detail further below, the SON features include a 127 system (QoS Improvement Assessment, QoS improvement evaluator) configured and usable to perform a method to evaluate the effectiveness of changes in the implemented network configuration 10 in the mobile network 100 following the CCO and / or MLB procedures carried out by the functions SON. Reference 105 indicates a specific geographic area that is a portion of the territory covered by the mobile network 100, which is assumed to be considered for the purposes of the improving or optimizing QoS with MLB and CCO. 15 According to one embodiment of the present invention, the network cells 115 covering the geographical area 105 considered are appropriately divided into three sets, as schematized in Fig. 1: - a first set of network cells (“Target cells”) 115a: the Target cells 115a They are network cells susceptible to changes in their configuration, i.e. 20 likely to have cell parameters modified for optimization purposes (e.g., one or more antenna parameters, e.g. electrical tilt antenna, antenna azimuth, transmission power, parameters beamforming or cell selection / transfer parameters). Preferably, the number of Target 115a cells considered is limited for example to a maximum 25 of about twenty cells; - a second set of cells (“Adjacent Cells” or “Corona-1 Cells” or “Cells Tier-1”) 115b, geographically contiguous, i.e. in proximity and adjacent to the Target 115a cells; Corona-1 115b cell configuration parameters do not can be changed, or should not be changed, but these cells of 30 network are still affected (due to radio signal interference) from Target 115a cells and can influence Target 115a cells (i.e. interfere with the relevant radio signals), so that the configuration change of Target 115a cells may imply a variation of the useful coverage area (best serving area) of the Corona-1 115b cells; the Corona-1 115b cells 5 are taken into consideration to consider the boundary effects of the network configuration optimization actions attempted by the features SON 125 for CCO and MLB; - in embodiments of the present invention, it is also considered a third set of cells ("Corona-2 cells" or "Tier-2 cells") 115c; the cells 10 Corona-2 115c cells are located in the vicinity of the Corona-1 115b cells, whose cell parameters cannot be changed / must not be changed, and which are not influenced by the Target 115a cells, i.e. the areas of best Corona-2 115c cell servers do not undergo significant changes change the configuration (i.e. cell parameters) of the Target cells 15 115a. Corona-2 115c cells are used to delimit appropriately the overall best serving area of ​​the cells of the first and of the second set (Target cells 115a and Corona-1 cells 115b). In embodiments of the present invention, a "cell cluster" 130 is defined to include Target 115a cells and Corona-1 115b cells, 20 considered as a whole. In other embodiments of the present invention, the "cell cluster" can be defined to include, in addition to the Target 115a cells and to the Corona-1 115b cells, also the Corona-2 115c cells. In both cases, the area of overall coverage of the 130 cell cluster remains constant (at least in good approximation) after a change in the configuration has been implemented 25 of the mobile network. Considering the 130 cell cluster as a whole, rather than individual cells of the network, it is possible to introduce, in the comparison between the experimental data (measurements, counters network statistics, KPIs) collected before and after the implementation of the change of network configuration, a constraint on the traffic carried by the mobile network (and hence 30 on the distribution of users) on the territory (while this constraint cannot be imposed directly on the traffic carried in the individual cells of the network, which changes, after the implementation of the network configuration change, precisely because the change of the network configuration moves portions of the traffic offered by a cell of the network to the other; an alternative embodiment will be described below in which the cell cluster 130 is not considered as a whole). 5 The following is an overview of a method according to one form of embodiment of the present invention. The method according to an embodiment of the present invention includes the following activities, actions or phases (not necessarily performed in chronological order in which they are presented below), schematically represented in the diagram of 10 activities of Fig. 2; reference is also made to the functional block diagram of Fig. 7. 1. Activity 205 A is identified and selected (e.g. by the network operator) network optimization method or algorithm to be used for CCO and MLB of the mobile network 100. For this purpose any of the algorithms can be chosen 15 known network optimizations, e.g. the algorithm described in WO 2022 / 207402 Al, or the algorithm presented in the paper by Skocaj et al. "Cellular Network Capacity and Coverage Enhancement with MDT Data and Deep Reinforcement Learning". 2. Activity 210 Always preliminarily, a metric for the evaluation of the 20 Mobile network performance, especially a metric for QoS evaluation experienced by mobile network users (QoS Evaluation Metric, QoS evaluation). As mentioned above, in one embodiment of the present invention, in the QoS evaluation is taken into account user throughput (amount of data received - bits received correctly - by a 25 users of a user equipment within a network cell in a certain period of time; for example, user throughput can be expressed in kbits per second, [kbit / s]), so the QoS evaluation metric is based on user throughput. In particular, the QoS evaluation metric is based on the distribution of the user throughput, or the distribution of data transfer speeds 30 experienced by users, or, otherwise said, on the distribution of traffic offered to the mobile network by users as a function of user throughput. For example, as will be described in more detail below, the QoS evaluation metric can ° be based on an appropriately weighted combination of the 5th percentile of the ° user throughput distribution and 50th percentile of user throughput distribution 5 user throughput. In embodiments of the present invention the evaluation metric QoS can also take into account, in addition to user throughput, other parameters, such as parameters that provide an indication of the number of transfers (handover), and / or parameters that provide an indication of how many 10 mobile network base stations are “active” or “on”, i.e. not “off”: in fact, the greater the number of base stations “on”, the greater the user throughput, but this comes at the expense of increased energy consumption of the mobile network; metric can then be defined in order to apply increasing penalties for a number increasing number of active base stations). In any case, the specific metric defined for the 15 Network performance evaluation does not constitute a limitation of this invention. 3. Activity 215 During the operation of the mobile network 100, at a certain time in the area geographic 105 emerges and is detected (for example by monitoring systems of the 20 mobile network operator) a situation that requires adequate Network optimization action (MLB and CCO). For example, the situation detected it could be an asymmetry in the distribution of traffic offered that persists over time, causing some cells in the network to be more heavily loaded than the cells nearby / adjacent; in this case there is a degradation in network performance 25 mobile, particularly in the QoS experienced in terms of user throughput, due to the presence of congested network cells. The network operator may then decide to take action network optimization (MLB and CCO), for example to distribute part of the traffic from the congested network cell to one or more adjacent network cells (despite this 30 action may negatively impact the overall capacity of the mobile network, a benefit could be derived in terms of improved user throughput, and then of the QoS). The selected optimization algorithm is then executed and, as a result, a changed network configuration is identified, which affects the cell parameters of one or more Target 115a cells in area 105, which improves / optimizes network performance; at a point in time the network configuration changed c 5 identified is deployed and implemented in the mobile network in the field, modifying the cell parameters of one or more of the Target 115a cells. 4. Activity 220 – Block 705 As schematized in Fig. 3, two time intervals are defined acquisition, INT and INT , a first acquisition time interval INT ABA 10 from time t to time t before time t of implementation of the A1 A2 c changed network configuration (change of network configuration) and a second acquisition time interval INT from time t to time t B B1 B2 subsequent to the instant t of the network configuration change. Each of the two c INT and INT acquisition time intervals are relatively long, lasting AB 15 For example, a few days or a few weeks. It should be noted that it is not relevant how long before time t of the network configuration change, the first one is placed c acquisition time interval INT , nor how long after time t of the A c network configuration change is placed in the second time interval of INT acquisition; the two acquisition time intervals INT and INT must be chosen BAB 20 to represent a generic condition of the mobile network before the change of network configuration and, respectively, after changing network configuration (where “generic condition” means a mobile network condition not influenced by particular events, such as national or local holiday periods, and / or periods of time including events that cause exceptional gatherings of people in an area, 25 such as concerts or sports matches, and / or exceptional natural events). During the two acquisition time intervals network measurements, statistical counters are acquired network and key performance indicators (KPI). The KPIs that It may be useful to acquire during the two acquisition time intervals INT and INT AB include: number of users (in Erlang), user distribution, traffic volumes, 30 throughput that the mobile network is able to provide to users (user throughput, in particular in uplink, or in downlink, or overall, i.e. in uplink and in downlink), distribution of traffic volumes based on user throughput. In a embodiment of the present invention, the network statistical counters and KPIs They are aggregated data at the level of individual network cells. In particular, in the example of realization of the invention considered here, the acquired KPIs are or include the 5 distribution of traffic volumes based on user throughput and average number of active users in the network cells. In particular, in one embodiment of the present invention, there are defined elementary acquisition time intervals, called “Records Of Period” (“period registers”), briefly “ROPs”, which are shorter than the first and second 10 acquisition time interval INT and INT , and which have for example a duration AB of a few minutes or tens of minutes each, for example 15 minutes. Each of the two INT acquisition time intervals and INT consists of a set of ROPs AB subsequent: the first acquisition time interval INT consists of the TO sequence of a number P of ROP ROP , ROP , …, ROP ; the second A,1 A,2 A,P 15 acquisition time interval INT is made up of a succession of a number B Q of ROP ROP , ROP , …, ROP . It is not necessary that the number of ROPs in the B,1 B,2 B,Q two acquisition time intervals INT and INT (and their time duration) are AB the same. In each ROP ROP , ROP , …, ROP of the first time interval of A,1 A,2 A,P 20 INT acquisition and in each ROP ROP , ROP , …, ROP of the second interval AB,1 B,2 B,Q of acquisition time INT the above KPIs are acquired (distribution B of traffic volumes per user and media throughput, in the ROP, of the number of users active in the network cells). 5. Activity 225 25 In order to carry out a statistical characterization of network traffic, defined a traffic homogeneity metric so that ROPs can then be classified during which the KPIs will be acquired in homogeneity classes, which are classes of homogeneity of the amount of network traffic. The traffic homogeneity metric can be based on measurements of active users or traffic volumes; this information 30 can then be aggregated at the cell set level, in particular at the level of 130 cell cluster. An example of traffic homogeneity metric for the ROP classification will be described further on. 6. Activity 230 Traffic homogeneity classes are defined based on the metric of defined traffic homogeneity. 5 7. Activity 235 – Block 710 KPIs are acquired in the ROP ROP , ROP , …, ROP of the first A,1 A,2 A,P INTA acquisition time interval. In particular, the acquired KPIs that are collected during the generic ROP ROP , ROP , …, ROP of the first acquisition time interval INT , a A,1 A,2 A,PA 10 level of the individual network cells 115a and 115b of the cell cluster 130, are aggregated to obtain corresponding aggregate KPIs at the cell cluster level 130: for example in the generic ROP ROP , ROP , …, ROP of the first interval of A,1 A,2 A,P INT acquisition time, the acquired KPIs relating to the individual network cells 115a and TO 115b of cell cluster 130 are summed for all network cells of the cluster 15 cells. 8. Activity 240 The modified network configuration (135 in Fig. 1) determined from the execution of the network optimization algorithm is implemented (140 in Fig. 1) in the mobile network, by changing the cell parameters of one or more of the Target cells 20 115a. 9. Activity 245 – Block 710 Similar to Activity 235, KPIs are captured in the ROPs, B,1 ROP , …, ROP of the second acquisition time interval INT after the B,2 B,QB change network configuration. 25 As previously, the KPIs acquired and collected, during the generic ROP ROP , ROP , …, ROP , at the level of the individual network cells 115a and 115b of the B,1 B,2 B,Q clusters of 130 cells, are aggregated (e.g. added together) to obtain corresponding KPIs aggregated at the cell cluster level 130. 10. Activity 250 – Blocks 715 and 720 As schematized in Fig. 4, the generic ROP (with X = A or B ei = 1, 2, …, X,i P oi = 1, 2, …, Q) includes an aggregation, 405, at the cell cluster level 130 and of the ROP, of KPIs collected, during the ROP ROP time interval, at the X,i of the individual network cells 115a and 115b of the cell cluster 130. The KPIs collected at the 5 of the individual network cells of the cell cluster include ThrP user throughput and the number of active Nu users, averaged in the ROP. The 405 aggregation results in an aggregate average number of users Nu[ROPX,i] in the cell cluster 130 in the ROP and an aggregate distribution of user throughput in the cell cluster 130 in the ROP, in terms of traffic volume transported TrV (e.g. in kB) for the value of 10 ThrP throughput (e.g., in kbit / s). The aggregate average number of users Nu[ROP ] in ROP is for example calculated as the sum, over the network cells of the X,i 130 cell cluster, of the average, in the ROP, of the number of active users. Fig. 5 illustrates schematically an exemplary distribution of the aggregate user throughput (For purely illustrative purposes, a Gaussian distribution is shown). From the ° 15 aggregate user throughput distribution, 5 percentile are obtained ° ThrP [ROP ] and the 50th percentile ThrP [ROP ] (i.e. the median) of the 5% X,i 50% X,i distribution of aggregate user throughput in the ROP. Each ROP (of the first and second INT acquisition time interval TO and INT ) is assigned to a respective traffic homogeneity class, such as B 20 will be described in detail later. 11. Activity 255 – Block 725 Based on the result of the classification of the first and second ROPs acquisition time interval INT and INT according to the homogeneity classes of the AB traffic, the weights to be assigned to the ROPs of the second time interval are calculated 25 of INT acquisition in calculating network performance based on the metric of B network performance evaluation. The calculated weights are used to weigh differently the ROPs of the second INTB acquisition time interval in the different traffic homogeneity classes. 12. Activity 260 – Blocks 730 and 735 30 Finally, the network performance evaluation metric is applied to the ROP of the first acquisition time interval INT and weighted ROPs (ROP TO weighted with the calculated weights) of the second acquisition time interval INT , for B obtain respective reward values. Compare the two metric results evaluation of the network performance, i.e. the two reward values, and it is evaluated so if the new network configuration identified by the optimization algorithm 5 and implemented in mobile network is effective in improving network performance mobile, in particular the QoS (output 740 in Fig. 7). As mentioned in the activity overview presented above, the method according to the embodiments of the present invention provides for the definition of a metric for evaluating network performance, specifically a metric 10 useful for evaluating the effectiveness of a network configuration change. In In particular, the network performance evaluation metric is based on QoS (QoS Evaluation Metric (QoS evaluation metric) and can take into account throughput of user, of the number of transfers between cells (handover) and of the percentage of success, energy consumption. The QoS evaluation metric is based on 15 a statistical aggregation, over an extended time interval (the first and second acquisition time interval INT and INT , lasting a few days or a few AB weeks, before and after implementing the configuration change network), measurements, network statistical counters, KPIs acquired in the intervals elementary acquisition time, of the ROPs which, in succession, compose 20 the extended time interval. More specifically, in embodiments of the present invention the QoS evaluation metric is applied to a set of network cells. In a embodiment of the present invention, the QoS evaluation metric can for example be defined by considering all the Target 115a cells and the cells 25 Corona-1 115b, or cell cluster 130, whose overall area remains constant, as a first approximation, by varying the network configuration parameters (network cell parameters) as a result of the implementation of the modified (optimized) network configuration. In another embodiment of the present invention, the Corona-2 115c cells can also be considered in the 30 cell cluster definition and QoS evaluation metric. An illustrative and suitable metric for QoS evaluation, when the user throughput is the relevant quantity, it is a reward function (or metric of reward) which is a combination of two different percentiles of the distribution of the user throughput (see for example Fig. 5), where one of the two percentiles reflects the 5 QoS (in terms of user throughput) experienced by the largest user population penalized and the other of the two percentiles reflects the QoS of a population of users ° major. For example, the 5th percentile of the distribution is considered user throughput, to reflect the QoS experienced by the user population ° most penalized, and the 50th percentile of the throughput distribution is considered 10 user, to reflect the QoS experienced by half the user population. For example, the reward function W can be expressed as: W = p * ThrP + p * ThrP 1 5% 2 50% which represents the weighted average of the average user throughput of the worst 5% ° of users (5th percentile of user throughput distribution; but it is possible 15 adopt other percentile values), ThrP , weighted with a first weight p , and the 5% 1 average user throughput of 50% of users (50th percentile of the user distribution) user throughput; in this case too another value can be adopted percentile), Thr_ut , weighted with a second weight p . The first and second weights 50% 2 pep can for example be chosen equal to 0.5, although other values ​​are possible 1 2 20 (in any case p + p = 1). This reward function W is intended to favor the 1 2 population of the most penalised users without having too much of a negative impact on the average conditions. User throughput can be calculated using several approaches. According to one embodiment of the present invention, a method for 25 calculate the user throughput (uplink, or downlink, or overall, i.e. in (uplink and downlink) is the measurement, performed by the mobile network, of the throughput offered, or the throughput that the mobile network is able to provide to users: the network provides, among the monitoring parameters at the network cell level, the density of User throughput distribution (percentage and traffic volumes per interval) ° ° 30 of throughput values); from these measurements it is possible to estimate the 5 and 50 percentile of user throughput distribution density. These percentiles are made available for each single ROP (of the first and second time interval of acquisition) at the level of individual cells 115 of the cell cluster 130 (Target cells 115a more Corona-1 cells 115b). 5 To evaluate the reward function W for each ROP (ROP ROP , ROP , A,1 A,2 …, ROP of the INT acquisition time interval prior to the change of A,PA network configuration, and ROP ROPB,1, ROPB ,2, …, ROPB,Q of the time interval (INT acquisition after changing the network configuration), it is considered the B aggregate user throughput distribution density across all cells 10 cell cluster network 130 and in the considered ROP, obtained by aggregating, for all the 130 cell cluster network cells, the distribution of traffic volumes as a function of the throughput that the mobile network is able to provide to the user, thus obtaining the values Thr_ut and Thr_ut for each ROP (as shown graphically in Fig. 4). 5% 50% According to another embodiment of the present invention, another 15 method to estimate user throughput is to start from SINR estimates per pixel. The capacity of the mobile network in the generic pixel is estimated, therefore the estimate of the mobile network capacity is divided by the number of users included in the area best server obtained from the measurements of each cell. The percentile values they are evaluated after having assigned to each pixel a respective weight which is given 20 from the intensity of the traffic source in the pixel (e.g. the number of active users in the pixel considered). Compensations may be provided for users with SINR worse at the expense of those with higher SINR. The entire range is therefore considered. cell cluster area 130 and Thr_ut and Thr_ut values ​​are estimated 5% 50% . ° The number of users per cell plays an important role in the evaluation of 5 ° 25th and 50th percentile of user throughput; in fact, a change in traffic offered per cell will substantially disrupt these user throughput estimates. estimation of the QoS that the mobile network is able to provide to users is in fact a function of the number of active users distributed across the served territory (aggregated) in the individual cells: as the number of active users decreases, the 5th and 50th percentile of throughput 30 of user increases, which means that the network performance rating is very sensitive to the distribution of the number of active users in the network cells and, therefore, it is sensitive to the distribution of traffic volumes in the cell cluster areas 130. This makes it very difficult to compare measurement acquisitions of the throughputs performed in different time intervals. QoS degrades in the presence of inhomogeneity in the number of users 5 active from network cell to network cell, and the reward function W is able to ° detect it. With greater asymmetry, the 5th percentile throughput and the ° 50th percentile throughputs will tend to degrade, to the advantage of network cells with a low number of users with the highest percentiles. Consistency between traffic volumes (number of users) at cell level 10 is normally not maintained with network configuration changes: the acquisition of network measurements in two different time periods relating to time before and after the network configuration change will be influenced by the change in the coverage area of ​​the network cells (resulting from the change in the network configuration) and therefore from the variation in the number of active users per cell 15 network, even with the same traffic offered. However, the Applicant observed that it is it is possible to ensure at least the consistency of traffic volumes at the area level 130 cell cluster coverage, since the cluster coverage area (area of overall coverage of Target 115a cells and Corona-1 115b cells) remains, in first approximation, constant as the parameters of the network cells vary. It is therefore 20 It is possible to impose a constraint on the total traffic at the cell cluster level in the generic ROP in order to compare consistent ROPs with respect to traffic volume / quantity of traffic. According to one embodiment of the present invention, a comparison on network measurements, network statistics counters, KPIs, in particular KPIs relating to 25 user throughput distribution, is done considering time periods with homogeneous traffic volumes / quantities at the cell cluster level 130. Considering for example the measurements of network cells with ROP of 15 minutes, the constraint of homogeneous traffic volumes / quantities translates into a change of the reward function W applied in the processing of the acquired KPIs (in 30 in particular, the distribution of user throughput) in the ROPs ROP , ROP , …, B,1 B,2 ROP of INT acquisition time interval after changing the B,QB network configuration, so as to compare ROP ROP , ROP , …, ROP A,1 A,2 A,Q of the acquisition time interval INT and the ROP ROP , ROP , ..., ROP AB,1 B,2 B,Q of the INT acquisition time interval that present traffic sources B 5 homogeneous in the cell cluster 130. According to one embodiment of the present invention, for example, the 130 cell cluster network cell KPIs are captured in intervals of unit time ROP ROP , ROP , …, ROP of the time interval of A,1 A,2 A,P INT acquisition before the time t of the network configuration change and in the ROPs A c 10 ROP , ROP , …, ROP of the acquisition time interval INT after the B,1 B,2 B,QB network configuration change. In one embodiment of the present invention, the acquired KPIs include measurements of traffic volumes as a function of user throughput that the mobile network is able to provide users (user throughput distribution), 15 and measurements of the number of active users (average) in the ROPs. These KPIs can be acquired at the level of individual network cells of the cell cluster 130, and can then be aggregated to obtain values ​​relating to the 130 cell cluster as a whole. As mentioned above, the first and second time intervals of INT and INT acquisition can last a few days or a few weeks and can AB 20 include a different number of ROPs. Let P be the number of ROPs that compose the INT acquisition time interval before the configuration change TO network and Q the number of ROPs that make up the acquisition time interval INT B after changing the network configuration, with P  Q. For each ROP ROP (with X = A or B ei = 1, 2, …, P oi = 1, 2, …, Q), it is X,i 25 evaluated the total number of active Nu users in the cell cluster 130 and in the ROP and ° ° the 5th and 50th percentiles of the user throughput distribution are calculated, aggregates on cell cluster 130. According to an embodiment of the present invention, the following are defined: traffic homogeneity classes and each ROP ROP is then assigned to a X,i 30 of the defined traffic homogeneity classes. According to one embodiment of the present invention, the definition of traffic homogeneity classes includes considering as an indication of the traffic volume / quantity the number of active users in the cell cluster 130, and identify the minimum of the total number of active users N in the cell cluster u,min 5 130 and the maximum total number of active users N in the cell cluster 130 u,max in the first acquisition time interval INT . The range of values ​​[N , A u,min Nu,max] is divided into a predetermined number n of sub-intervals or bins (“bin”). For example, the number n of bins can be 20, with a variability of the number of active users (i.e. traffic volume) less than approximately 5% within each 10 bins. In another example, referring to Fig. 6A – Fig. 6C, the range of values [N , N ] is divided into five bins Bin , – Bin , , by the minimum of the number u,min u,max A 1 A 5 total number of active users N measured, in the first time interval of u,min INT acquisition, in the cell cluster 130 at most of the total number of TO active users N measured, in the acquisition time interval INT before the u,max A 15 network configuration change, in cell cluster 130. Each ROP ROP (i = 1, 2, …, P) of the first acquisition time interval To the INT is assigned to one of the traffic homogeneity classes as above TO defined, that is, to one of the bins Bin , – Bin , , based on the value of the parameter A 1 A 5 Nu[ROP ] for that ROP. To the 20 Fig. 6A graphically shows the classification of ROP ROP (i = 1, 2, To the …, P) of the first acquisition time interval INT: in the example three ROPs are TO classified in the bin Bin , , four ROPs are classified in the bin Bin six ROPs are A 1 A,2, classified in the bin Bin , , five ROPs are classified in the bin Bin , , and two ROPs are A 3 A 4 classified in the bin Bin , (in a practical scenario, the number of ROPs that make up A 5 25 the first INT acquisition time interval will be greater: with ROP of 15 minutes TO each, one day comprises 96 ROPs, so a first time interval of INTA acquisition of one week includes 672 ROPs; possibly also the number of bins – the number of traffic homogeneity classes – will be greater than 5). The piecewise linear curve shown with dashed line and identified as 605 in Fig. TO 30 6A represents the traffic distribution profile in the cell cluster 130, in the acquisition time interval INT TO. Similarly, each ROP ROP (j = 1, 2, …, Q) of the second interval of B,j acquisition time INT is assigned to a respective of the homogeneity classes of B traffic as defined above, or to one of the bins Bin , – Bin , , (which correspond, B 1 B 5 in terms of ranges of values ​​of the number of active users in the interval [N , N ], u,min u,max 5 to bins Bin , – Bin , ) based on the value of the Nu[ROP ] parameter for that ROP. A 1 A 5 B,j Fig. 6B graphically shows the classification of ROP ROP (j = 1, 2, B,j …, Q) of the second INTB acquisition time interval: in the example two ROPs are classified in the bin Bin , , one ROP is classified in the bin Bin , , four ROPs are B 1 B 2 classified in the bin Bin , , six ROPs are classified in the bin Bin , , and five ROPs are B 3 B 4 10 ranked in the bin Bin , . The piecewise linear curve shown with dashed line and B 5 identified as 605 in Fig. 6B represents the traffic distribution profile in the B 130 cell cluster, in the second acquisition time interval INT B. In order to evaluate the effectiveness, or “goodness”, of the modified network configuration which has been identified by the network optimization algorithm and which, at time t, is c 15 has been implemented in the mobile network (by modifying the cell parameters of one or more of the Target cells 115a of the cell cluster 130), a reward is calculated and This is assigned to network performance, i.e. QoS (in terms of throughput of user) compared to the network configuration before the configuration change of network. This can be done for example by calculating the value W of a function of TO 20 reward, for example with the reward function formula W presented in precedence, on the ROP ROP (i = 1, 2, …, P) of the first time interval of To the ° INT acquisition, in particular on ThrP values ​​(5th percentile of the distribution At 5% ° of throughput) and ThrP (50th percentile of throughput distribution) of ROPs 50% ROP (i = 1, 2, …, P): To the 25 W = p * ThrP + p * ThrP A 1 5% 2 50% One possibility to calculate the reward function W for the TO network configuration before changing the network configuration is to perform the calculating the reward W for each ROP ROP individually, and then A,i A,i calculate the average of the calculated rewards W . Another way to calculate the To the 30 W reward function for network configuration before changing the network configuration is to obtain an aggregate distribution of user throughput, ° ° aggregated over all ROP ROP , then calculate the 5th and 50th percentile of a To the aggregate user throughput distribution and apply to the calculated percentiles formula W = p * ThrP + p * ThrP . A 1 5% 2 50% 5 As regards network performance, i.e. QoS (in terms of user throughput) after changing the network configuration, you can see that the traffic distribution profile 605B in the cell cluster 130 in the second INT acquisition time interval is different from the distribution profile of the B 605 traffic in cell cluster 130 in the first acquisition time interval TO 10 INT , this being the consequence of random factors such as the mobility of the TO users (i.e. traffic sources) across the cell cluster coverage area 130. The Applicant understood that such differences in the distribution profiles of the Traffic before and after the network configuration change make a direct comparison of network measurements (KPIs) before and after the network configuration change 15 reliable in evaluating the effectiveness, or ineffectiveness, of the optimization action of the network identified by the network optimization algorithm and implemented on the field. In other words, by calculating the reward function W directly on the measures / KPIs acquired in the ROP ROP (j = 1, 2, …, Q) of the second time interval B,j of INT acquisition without considering the variations in traffic distribution B 20 would provide a value that is not very significant for comparison with the W value calculated for TO i ROP ROP (i = 1, 2, …, P) of the first acquisition time interval INT . A,i A According to the present invention, for the calculation of the reward to be awarded to the modified network configuration detected by the optimization algorithm network and implemented in the mobile network, the traffic distribution profile, calculated 25 compared to the measurements acquired during the second acquisition time interval INT, exemplified by curve 605 in Fig. 6B, is “reshaped” to BB to correspond, that is, to be made similar, to the traffic distribution profile calculated with respect to the measurements acquired during the first time interval of INT acquisition, exemplified by curve 605 in Fig. 6A. AA 30 According to one embodiment of the present invention, the reshaping of the traffic distribution profile for the second interval INT acquisition time, curve 605 in Figure 6B, is based on a processing BB statistics of network measurements in the ROP ROP (j = 1, 2, …, Q) of the second B,j acquisition time interval INT . B According to one embodiment of the present invention, the processing 5 statistics of network measurements in the ROP ROP (j = 1, 2, ..., Q) of the second interval B,j For example, the INT acquisition time is performed as described below. B The aggregated measures in the ROPs ROPB,j (j = 1, 2, …, Q) of the time interval of INT acquisition after changing the network configuration, in particular the B ° ° 5th percentile ThrP [ROP ] and 50th percentile ThrP [ROP ] values ​​of the 5% X,i 50% X,i 10 aggregate distribution of user throughput in ROPs ROPs (j = 1, 2, …, Q) are B,j weighted with their respective weights. According to one embodiment of the present invention, the weights Peso da B,j assign to the ROPs ROPs ( j = 1, 2, …, Q) of the acquisition time interval B,j INT for the reshaping of the traffic distribution profile for the second B 15 acquisition time interval INT and therefore in the calculation of the function of B reward W are determined as: B Weight [Bin ] = #Bin / (#Bin * P ) B,j B,k A,k B,k Where: - Weight [Bin ] is the weight to assign to the generic ROP ROP (j = 1, 2, …, Q) B,j B,k B,j 20 of the second INT acquisition time interval that was classified B in the bin Bin (with k = 1, …, Max , Max = 5 in the example of Fig. 6B, which B,k Bin Bin identifies one of the bins Bin – Bin , that is one of the homogeneity classes of the B,1 B,5 traffic defined above); - #Bin (with k = 1, …, Max , Max = 5 in the example of Fig. 6A, which identifies A,k Bin Bin 25 one of the bins Bin – Bin , that is one of the bins of the homogeneity classes of A,1 A,5 traffic defined above) is the cardinality of the bin Bin in the classification of the A,k ROP ROPA,i (i = 1, 2, …, P) of the first acquisition time interval INT , i.e. the number of ROP ROP (i = 1, 2, …, Q) that have been AA,i classified in the bin Bin ; A,k - #Bin is the cardinality of the bin Bin , and B,k B,k - P is the total number of ROPs in the first time interval of INT acquisition. TO Returning to the example in Fig. 6A and Fig. 6B, the total number P of ROPs 5 in the first acquisition time interval INT is 20, P = 20, and the weights are: TO #Bin = 3 #Bin = 2 Weight [Bin ] = 3 / (2*20) = 0.075 A,1 B,1 B,i B,1 #Bin = 4 #Bin = 1 Weight [Bin ] = 4 / (1*20) = 0.2 A,2 B,2 B,i B,2 #Bin = 6 #Bin = 4 Weight [Bin ] = 6 / (4*20) = 0.075 A,3 B,3 B,i B,3 #Bin = 5 #Bin = 6 Weight [Bin ] = 5 / (6*20) = 0.042 A,4 B,4 B,i B,4 #Bin = 2 #Bin = 5 Weight [Bin ] = 2 / (5*20) = 0.02 A,5 B,5 B,i B,5 In case no ROP ROP (j = 1, 2, …, Q) of the second interval of B,j INT acquisition time has been classified into one or more of the bins Bin - Bin , BB,1 B,5 said bin(s) is / are not taken into consideration, i.e. the number of ROPs of the 10 first INT acquisition time interval that were classified into those bins TO Bin – Bins corresponding to the empty bin or empty bins Bin – Bin is subtracted A,1 A,5 B,1 B,5 from the total number of ROP P. The value of the reward function W is then calculated based on the B ϯ ϯ ° [ThrP ] and [ThrP ] values, which are the ThrP values ​​(5th percentile of the 5% 50% 5% ° 15 throughput distribution) and ThrP (50 percentile of the throughput distribution) 50% throughput) of the ROPs ROP (j = 1, 2, …, Q), weighted with the weights Weight [Bin ]: B,j B,j B,k ϯ ϯ W = p * [ThrP ] + p * [ThrP ] B 1 5 % 2 50% Again, there are two possibilities to calculate the function of W reward for network configuration after network configuration change B 20 network: one possibility is to perform the calculation of the reward W for each ROP ROP B,j B,j (weighted by their respective weight Weight [Bin ]) individually, and then calculating the B,j B,k average of the calculated rewards W ; another possibility is to obtain a B,j aggregate user throughput distribution, aggregated across all ROPs B,j ° ° (weighted by the respective weight Weight [Bin ]) and then calculate the 5th and 50th percentile B,j B,k of the aggregate (weighted) distribution of user throughput and apply to the ϯ ϯ percentiles calculated using the formula W = p * [ThrP ] + p * [ThrP ]. In particular, the B 1 5% 2 50% aggregate distribution of user throughput can be calculated by weighting the 5 Distribution of user throughput of each ROP ROP based on its weight B,j Weight [Bin ] = #Bin / (#Bin * P). B,j B,k A,k B,k The calculated value of the WB reward function is then compared with the calculated value of the reward function W , to evaluate whether the configuration TO of modified network implemented in the field brought benefits, in terms of QoS, 10 compared to the previous network configuration. For example, a value of W B sufficiently higher than the W value ensures that the network configuration TO modified improved QoS. In this way a homogeneous statistical characterization is given to the acquired network measurements, network statistics counters, KPIs, such as throughput 15 user, acquired at different times (before and after the configuration change of the network), so that the acquired network measurements can be compared with each other in significantly. The homogeneous statistical characterization can be given on the basis of measurements of the number of active users in the 130 cell cluster, attributing to the individual ROP appropriate weights, suitable to ensure such homogeneity. The weights assigned to the ROPs are 20 aimed at allowing a comparison between ROPs with homogeneous numbers of active users in the cluster of cells. Alternative formulations of the reward function W are possible. If, from the comparison of the values ​​of the reward function W and W , it turns out that AB the modified network configuration implemented in the field did not bring any benefits, 25 in terms of QoS, compared to the previous network configuration (because the QoS after the change did not improve significantly, not to mention the QoS after the change is worse than before), you can take one or more of the following actions. One possible action is to rerun the algorithm optimization / simulation by modifying: 30 - the variability ranges of the modifiable parameters (e.g. inclinations electrical and / or azimuth of the antennas, transmitted power, selection / reselection thresholds of network cells), in order to increase the degrees of freedom of the algorithm network optimization / simulation; - the set of modifiable parameters that the algorithm 5 network optimization / simulation can be considered (for example in the execution previous optimization / simulation algorithm were not taken into account taking into account all possible modifiable parameters); - the set of network cells (Target 115a cells) whose parameters can be modified; 10 - the importance (weights p , p ) assigned to the different items (e.g. throughput 1 2 of user ThrP , ThrP ) considered in the calculation of the reward function for 5% 50% network configurations before and after the change, thus modifying the objectives fixed for the optimization / simulation algorithm; - as an additional resource, the modification of the algorithm of 15 optimization / simulation (e.g. in case of algorithms optimization / simulation based on machine learning / artificial intelligence, this can lead to retraining of the algorithm). In accordance with another embodiment of the present invention, in addition at the first and second acquisition time interval INT and INT AB 20 previously described, a further time interval is defined INT acquisition, called reference acquisition time interval R INT, during which network measurements and KPIs are acquired. The time interval R INT reference acquisition is a time interval chosen sufficiently R long (lasting a few days or a few weeks) so as to represent a 25 generic condition of the mobile network (which is taken as a reference condition of the mobile network), or a condition not influenced by particular events that can occur at particular moments, before and / or after the change of network configuration. As illustrated in Fig. 8, the time interval of INT reference acquisition can for example be a time interval that R 30 starts before (time t) the instant t of the implementation of the change of the R1 c network configuration and ends after (time t) the instant t of the change of the R2 c network configuration; for example (but not limited to) the time interval of INT reference acquisition spans the first and second time intervals R of INT and INT acquisition (in the example of Fig. 8 the beginning (time t ) of the interval AB R1 5 of reference acquisition time INT is before the start (time t ) of the R A1 first acquisition time interval INT and the end (time t ) of the interval To R2 reference acquisition time INTR is after the end (time tB2) of the second acquisition time interval INT ; however, the time interval of B INT reference acquisition could also start at the beginning of the first interval R 10 of INT acquisition time and end at the end of the second time interval of TO INT acquisition An appropriate selection of the acquisition time interval B. of reference INT , for example a relatively long time interval, R ensures that it reliably represents a generic network condition mobile, that is, a robust condition of the reference mobile network. 15 Like the two acquisition time intervals INT and INT , the acquisition time interval AB reference acquisition time INT is made up of a succession of ROPs of R reference ROP , ROP , …, ROP , for example lasting 15 minutes R,1 R,2 R,N each. Network measurements, network statistics counters and KPIs, in particular network traffic 20 network / number of active users and user throughput distribution, acquired during the reference acquisition time interval INT are used to obtain R a reference statistical characterization of network traffic (i.e., otherwise from the embodiment of the invention described above, the statistical characterization of network traffic is not derived from the KPIs acquired during 25 the first acquisition time interval INT ). TO Traffic homogeneity classes for statistical characterization of the network traffic are defined with respect to the acquisition time interval of reference INT . As in the embodiment of the invention previously R described, the number of active users in the cell cluster 130 during the time interval 30 of INT reference acquisition is considered indicative of traffic volume, and R the minimum of the total number of active users N is identified in uR,min cell cluster 130 and the maximum total number of active users N in the uR,max 130 cell clusters in the reference acquisition time interval INT R. The range of values ​​[N, N] is divided into a predetermined number n of uR,min uR,max sub-intervals or bins, e.g. 20, with a variability in the number of active users 5 (i.e., traffic) of less than about 5% within each bin. In an example differently, with reference to Fig. 9A, the range of values ​​[N , N ] is divided uR,min uR,max in five bins BinR,1 – BinR,5, from the minimum of the total number of active users N measured, in the reference acquisition time interval INT , in uR,min R cell cluster 130 at most of the total number of active users N uR,max 10 measured, in the reference acquisition time interval INT , in the cluster of R cells 130. Each reference ROP ROP (h = 1, 2, …, N) of the time interval of R,h INT reference acquisition is assigned to a respective of the classes of R traffic homogeneity as defined above, i.e. to one of the bins Bin , – Bin , , in R 1 R 5 15 based on the value of the Nu[ROP ] parameter (average number of active users in the ROP of R,h reference) for that reference ROP. Fig. 9A graphically shows a Example classification of reference ROPs ROP (h = 1, 2, ..., N) R,h of the reference acquisition time interval INT . In the example, three ROPs of R reference are classified in the bin Bin , , four reference ROPs are R 1 20 are classified in the bin Bin , , six reference ROPs are classified in the bin Bin , , R 2 R 3 five reference ROPs are classified into the bin Bin , , and two reference ROPs R 4 are classified in the bin Bin , (as mentioned in relation to the shape of R 5 realization of the invention described above, in a practical scenario the number of reference ROPs ROP , h = 1, 2, ..., N, which make up the interval of R,h 25 reference acquisition time INT will be longer: with reference ROP of R 15 minutes each, one day includes 96 reference ROPs, so one interval of one week's INTR reference acquisition time includes 672 ROPs of reference). The piecewise linear curve shown with dashed line and identified as 905 in Fig. 9A represents the reference traffic distribution profile in the R 30 clusters of 130 cells, in the reference acquisition time interval INT . R Similarly, each ROP ROP (i = 1, 2, …, P) of the first interval of To the acquisition time INT , and each ROP ROP (j = 1, 2, …, Q) of the second AB,j acquisition time interval INT is assigned to one of the respective classes of B traffic homogeneity as defined above, or to one of the bins Bin , – Bin , (which R 1 R 5 correspond to ranges of values ​​of the number of active users in the interval [N , uR,min 5 N ]), based on the value of the parameter Nu[ROP ] (X = A or B, i = 1, …, P or 1, …, uR,max X,i Q) for that ROP. Fig. 9B and Fig. 9C graphically show the classification of ROPA ROPs,i (i = 1, 2, …, P) of the acquisition time interval INT before the change of TO Network configuration and classification of ROPs ROP (j = 1, 2, …, Q) B, j 10 of the INT acquisition time interval after the configuration change of the B network by exploiting the traffic homogeneity classes Bin , – Bin , defined on the basis of R 1 R 5 to the reference acquisition time interval INT : in the example: R - for the acquisition time interval INT before the change of TO network configuration, four ROPs are classified into the bin Bin , (corresponding A 1 15 to the bin Bin , ), two ROPs are classified into the bin Bin , (corresponding to the bin Bin , ), R 1 A 2 R 2 three ROPs are classified in the bin Bin , (corresponding to the bin Bin , ), five ROPs A 3 R 3 are classified in the bin Bin , (corresponding to the bin Bin , ), and four ROPs are A 4 R 4 classified in the bin Bin , (corresponding to the bin Bin , ). The piecewise linear curve A 5 R 5 shown with dashed line and identified as 905 in Fig. 9B represents the TO 20 traffic distribution profile in the cell cluster 130, in the time interval of INT acquisition ; TO - for the INT acquisition time interval after the change of the B network configuration, two ROPs are classified into the bin Bin , (corresponding to the bin B 1 Bin , ), a ROP is classified into the bin Bin , (corresponding to the bin Bin , ), four R 1 B 2 R 2 25 ROPs are classified in the bin Bin , (corresponding to the bin Bin , ), six ROPs are R 3 R 3 classified in the bin Bin , (corresponding to the bin Bin , ), and five ROPs are classified B 4 R 4 in the bin BinB,5 (corresponding to the bin BinR,5). The piecewise linear curve shown with dotted line and identified as 905 in Fig. 9C represents the profile of B traffic distribution in the cell cluster 130, in the time interval of 30 INT acquisition B. You can see that the traffic distribution profiles 905 and 905 in the cluster AB of cells 130 in the acquisition time intervals INT and INT are different from the AB Reference traffic distribution profile 905 in cell cluster 130 R in the reference acquisition time interval INT , this being the R 5 consequence of random factors such as the mobility of users over time (i.e. traffic sources) across the coverage area of ​​cell cluster 130. As in the embodiment of the invention previously described, according to the present invention, for the calculation of the reward to be assigned to the modified network configuration that has been identified by the algorithm 10 network optimization and implemented in the mobile network, the distribution profiles of the traffic 905 and 905, calculated on the measurements acquired during the first interval of AB acquisition time INT and, respectively, during the second time interval TO of INT acquisition, are “remodeled” to be made similar to the profile of B distribution of reference traffic 905 calculated on the measurements acquired during R 15 the reference acquisition time interval INT . R According to one embodiment of the present invention, the remodeling of the traffic distribution profiles 905 and 905 relating to the first and AB at the second acquisition time interval INT and INT is based on a processing AB Network measurement statistics, network statistics counters, KPIs collected in ROPs 20 ROPs (i = 1, 2, …, P) of the first acquisition time interval INT and in the ROPs A,i A ROP (j = 1, 2, …, Q) of the second acquisition time interval INT . B,j B According to one embodiment of the present invention, the processing statistics on KPIs in ROPs ROP (i = 1, 2, …, P) and ROP (j = 1, 2, …, Q) is A,i B,j example performed as described below. 25 The KPIs aggregated in the ROPs ROP (i = 1, 2, …, P) and ROP (j = 1, 2, …, Q), in A,i B,j ° ° in particular the values ​​of the 5th percentile ThrP [ROP ] and the 50th percentile ThrP 5% Y,i 50% [ROPY,i] of the aggregate user throughput distribution in the ROPs are weighted with their respective weights. The weights Weight (Y = A or B; w = 1, …, P for Y = A, w = 1, …, Q for Y = B ) from Y,w 30 assign to the ROPs ROP (i = 1, 2, …, P) and ROP (j = 1, 2, …, Q) in the calculation of the A,i B,j reward functions W and W are determined as: AB Weight [Bin ] = #Bin / (#Bin * N) Y,w Y,k R,k Y,k Where: - Weight [Bin ] is the weight to assign to the generic ROP ROP (i = 1, 2, …, P) Y,w Y,k A,i 5 or ROP (j = 1, 2, …, Q) which was classified into the bin Bin (with k = 1, …, B,j Y,k Max , Max = 5 in the example of Fig. 9A, which identifies one of the bins Bin Bin Bin Y,1 – BinY,5, i.e. one of the traffic homogeneity classes defined above); - #Bin (with k = 1, …, Max , Max = 5 in the example of Fig. 9A, which identifies Y,k Bin Bin one of the bins Bin – Bin , that is one of the bins of the homogeneity classes of Y,1 Y,5 10 traffic defined above) is the cardinality of the bin Bin in the classification of the Y,k ROP ROP (i = 1, 2, …, P) and ROP (j = 1, 2, …, Q), that is, the number of A,i B,j ROP ROP (i = 1, 2, …, P) and ROP ROP (j = 1, 2, …, Q) which have been A,i B,j classified in the bin Bin ; Y ,K - #Bin is the cardinality of the bin Bin , and R,k R,k 15 - N is the total number of reference ROPs in the time interval of INT reference acquisition. R Similarly to the previously described embodiment, in the case where no ROP ROP (i = 1, 2, …, P) or ROP ROP (j = 1, 2, …, Q) has been A,i B,j classified in one or more of the bins Bin – Bin or Bin - Bin called / s bins not A,1 A,5 B,1 B,5 20 is / are taken into account, i.e. the number of ROPs of the interval reference acquisition time INT that have been classified into the bin(s) R corresponding to the empty bin(s) Bin – Bin or Bin - Bin is subtracted from the A,1 A,5 B,1 B,5 total number of ROP N. The values ​​of the reward functions W and W are then calculated based on AB ϯ ϯ ϯ ϯ 25 to the values ​​[ThrP ] and [ThrP ] , [ThrP ] and [ThrP ] , which are the values 5% A 50% A 5% B 50% B ° ° ThrP (5th percentile of the throughput distribution) and ThrP (50th percentile of the 5% 50% throughput distribution) of the ROPs ROPA,i (i = 1, 2, …, P) and ROPB ,j (j = 1, 2, …, Q), weighted with the weights Weight [Bin ]: Y,w Y,k ϯ ϯ W = p * [ThrP ] + p * [ThrP ] A 1 5% A 2 50% A ϯ ϯ W = p * [ThrP ] + p * [ThrP ] B 1 5% B 2 50% B As before, you can use any of the methods described to the calculation of the reward W and W A B. 5 The reward W calculated for the network configuration after the change of B network configuration is compared with the reward W calculated for the TO network configuration before changing network configuration, to evaluate whether the modified network configuration implemented in the field brought benefits, in terms of QoS, compared to the previous network configuration. For example, a value 10W sufficiently higher than the W value ensures that the network configuration BA modified improved QoS. In essence, a difference of this embodiment from the form of the embodiment described above is that in the embodiment previously described is taken as the acquisition time interval of 15 reference the first acquisition time interval INT , and the weights Weight (w = AA,w 1, … , P) are unit weights.    In the embodiment of the invention described above, the traffic overall (e.g., expressed in terms of number of users, or average number of 20 active users) in the cell cluster 130 as a whole was considered in the definition of traffic volume classes. The acquired KPIs to be used for homogenize network traffic (distribution of traffic sources, i.e. users) before and after the network configuration change are given (number of active users, user throughput distribution) that the mobile network provides aggregated at the user level 25 of the network cells of the 130 cell cluster and which are then aggregated at the level of the 130 cell cluster, and the network traffic classes are network traffic classes at the of the cell cluster 130 . An alternative method for traffic homogenization will now be described. network, which is based on the geographical distribution of traffic volumes (e.g. 30 number of active users) in the territory covered by the 130 cell cluster. As before, to evaluate the benefits, in terms of QoS (and, in particular, QoS in terms of user throughput) resulting from the change of the network configuration, user throughput before and after changing network configuration to be compared should be referred to homogeneous conditions 5 of network traffic volume (e.g. expressed in terms of average number of users active), so that the user throughput measured before and after is actually comparable. While in the previously described embodiment it was considering the overall network traffic in the 130 cell cluster, in this method alternatively a finer geographical distribution of the volume of the 10 network traffic, at pixel level. As better described below, in this alternative mode of homogenization of network traffic, traffic quantity classes are defined homogeneous based on the distribution of users in portions, sub-areas of the areas of the network cells, for example in pixels of the geographical territory, using measurements 15 georeferenced. Network traffic quantity indicators can be any one, or one or more of: network traffic volume indicators (e.g. expressed in kbit), indicators of user presence in network cells (i.e. users for whom the mobile network records a final event in the network cells, derived from location cues 20 of the user at the time of recording the last event), number indicators of active users in the network cells. In particular, this alternative method of homogenizing traffic network provides an alternative implementation of Activity 225 and Activity 230 described above. 25 First, using for example a simulation tool (e.g. for example an electromagnetic field propagation simulator, of the often type used by network operators in the design phase of a communications network mobile, capable of simulating the propagation of radio signals across a territory taking into account the data describing the territory, such as natural orography, presence 30 of human artifacts, i.e. buildings, etc., presence of trees, etc.), or from measurements georeferenced in the field, theoretical and hypothetical coverage areas are evaluated of the cells of the cell cluster network 130 (i.e., the Target cells 115a plus the cells Corona-1 115b), for network configuration before changing configuration network (first network configuration A) and for the new network configuration after the 5 network configuration change (second network configuration B). For each network cell of the cell cluster 130, three are defined as follows: areas or zones Z1, Z2 and Z3 of the cellular coverage area (as shown in Fig. 10): - a first zone Z1: includes those pixels of the cell coverage area that 10 fall within the best serving area of ​​the cell both in the first network configuration A and in the second network configuration B, that is both before and after changing network configuration; - a second zone Z2: includes those pixels that fall within the area of ​​best cell server only in the first network configuration A (i.e. only before the 15 network configuration change: these are lost pixels, that is, pixels that are lost from the network cell after the network configuration is changed from the first network configuration A to the second network configuration B); - a third zone Z3: includes those pixels that fall within the area of ​​best cell server only in the second network configuration B (i.e. only after the 20 network configuration change: these are acquired pixels, i.e. pixels that are acquired from the network cell after the network configuration is changed by the first network configuration A to second network configuration B). For each network cell of the cell cluster 130, during the first and second ROPs second acquisition time interval INT and INT , respectively before and AB 25 after the implementation of the modified network configuration (and, where defined, also for the reference acquisition time interval INT described above), R georeferenced measurements of network traffic volumes in the cell are obtained network (network traffic volume measurements may be or include indications of network traffic volumes, such as user attendance measurements 30 in the network cell – i.e. those users for whom the mobile network records a last event in the network cell - and / or the number of active users in the network cell). Network traffic volume measurements can be derived from MDT measurement campaigns (or other georeferenced measurements). As known to industry experts, MDT is a 3GPP standardized mechanism designed for 5 network operators to leverage user equipment in a mobile network to collect mobile network data: the user equipment collects measurements in the field, including radio measurements; MDT measurements are linked to information that allows to obtain the position of the user equipment and a timestamp, when the measurements are collected 10 on the field. Network traffic volume measurements can also be derived from the aggregation of network statistical counters, e.g. radius counters (the networks 5G mobiles, having base stations equipped with smart antennas – that is, highly directives – make individual counters available for each antenna beam 15 directional, providing geographic detail in the network cells). The measurements are acquired during the ROP ROP (i = 1, 2, …, P) of the first To the INT acquisition time interval, before the implementation of the TO modified network configuration, and the ROP ROP (i = 1, 2, …, Q) of the second Bi acquisition time interval INT , after implementation of the B 20 network configuration changed (and, where defined, also for the time interval of reference acquisition INT ). R For each network cell of the cell cluster 130 the area is then considered of the union zone Z2 Z3 which is the union of the respective second zone Z2 and third zone Z3 zone, and the volume of network traffic / user presence / number of active users in the 25 network cell is calculated as the difference between network traffic volume / presence users / number of active users in the second zone Z2e the traffic volume of network / user presence / number of active users in the third zone Z3: the traffic volume of network / user presence / number of active users thus calculated represent the volume of network traffic / user presence / number of active users that the network cell has handed over, 30 lost (if the difference is a negative number) or gained (if the difference is a negative number) positive). In this way, for each network cell of the cell cluster 130 is provided a quantification of the change in traffic volume before and after the change network configuration. This will allow you to make a comparison of the measurements before and after network configuration change in cell areas 5 network where network traffic remains substantially constant before and after the network configuration change. Subsequently, for each network cell, ranges of values ​​are defined (bin) of network traffic volume / user presence / number of active users measured in first acquisition time interval INT (before the configuration change TO 10 network) for classification purposes, for the first zone Z1 and for the union zone Z2 Z3. For example, two bins are defined for the first zone Z1 and two bins for the union zone Z2 Z3: for example one of the two bins may correspond to “higher” volume values of network traffic / user presence / number of active users, the other bin can match at “lower” values ​​of network traffic volume / user presence / number of active users, 15 where “upper” and “lower” can refer to a selected threshold of network traffic volume / user presence / number of active users). Fig. 11 schematizes the definition of two bins, Bin and Bin , for the generic zone Z1 and the LH generic union zone Z2 Z3 of the generic network cell of the cell cluster 130; bins are defined taking into account the number of active users N; N uu,Th 20 denotes a threshold for the number of active users used to discriminate numbers “higher” numbers of active users from “lower” numbers of active users. In principle, nothing prevents you from defining more than two bins for classification. of the measured network traffic volume / user presence / number of active users, however since in the embodiment of the invention the classification is 25 performed at the zone level of each network cell (rather than at the cell cluster level 130), it might be preferable to keep the number of bins relatively low, for a lower computational load. The bins are used to define classes of traffic volume homogeneity. network / user presence / number of active users. For zones Z1 and union zones Z2 Z3, the 30 ROPs of the first INT acquisition time interval and the ROPs of the second TO acquisition time interval INT (and, where defined, also the ROPs ROP , k = BR,k 1, 2, …, N, of the reference acquisition time interval INT ) will be R classified in the appropriate bin Bin or Bin , depending, for example, on the average number of LH active users in ROPs. 5 Considering all Target 115a cells and, optionally, Corona-1 cells 115b (or at least a number v of more significant Corona-1 115b cells that are more significant in terms of traffic volumes / user presence / number of active users in the respective zone Z1 and / or connection zone Z2 Z3: for example, they are considered Corona-1 115b cells with a traffic volume greater than 3 Erlangs, or with a number 10 active users higher than a selected number of active users), are identified all possible combinations of the two bins Bin , Bin of the respective zone Z1 and of the LH Z2 Z3 connection zones. The total number of such combinations is: ( [ ] ) # ( ) # where #bin is the number of bins (two in the example considered, Bin or Bin ), 2 is the number LH 15 of zones considered for each cell of the network (the first zone Z1 and the union zone Z2 Z3), #Targetcells is the number of Target cells 115a in cell cluster 130, and v[Corona-1cells] is the number v of most significant Corona-1 115b cells in the cluster cells 130. The total number of possible combinations of network cell bins 20 considered of the 130 cell cluster can easily become relatively high. For limit the total number of possible bin combinations, the number of cells in network can be reduced, for example by merging network cells, in particular by merging the first zones Z1 and / or the union zones Z2 Z3 of different network cells, for example on the basis of the network traffic volume (merging together high-traffic network cells 25 and low traffic volume cells). In particular, network traffic can be considered average peak in peak network traffic time: the first Z1 zones of the cells of networks having high peak network traffic are merged, and the first Z1 zones of the Low peak network traffic network cells are merged together; the merge zone Z2 Z3 of each network cell is divided into sub-areas that give pixels to cells of 30 network with high peak network traffic and sub-areas that acquire pixels from cells of network with low peak network traffic, so sub-areas are aggregated based on the peak traffic volume. In relation to the reduction of the overall number of possible combinations of bin, it is observed that the union zones Z2 Z3 of the network cells can be 5 divided into sub-zones considering the traffic volume (number of users) of the cells of networks that acquire / lose network traffic. Considering a network cell, the The Z2 Z3 connection zone can be divided into sub-zones, each of which represents an aggregate of pixels that “migrate” from (second zone Z2) and to (third zone Z3) of the network cell considered, respectively to, and from, network cells with classes 10 of homogeneous network traffic (i.e. pixels of the second zone Z2 of the network cell considering that they migrate to other low-traffic network cells, and pixels of the third zone Z3 of the network cell considered that come from other network cells at low traffic volume; and pixels of the second zone Z2 of the network cell considered that migrate to other network cells with high traffic volume, and pixels of the 15 third zone Z3 of the network cell considered that come from other network cells. cells high traffic volume). Based on MDT measurements (in ROPs) in the first Z1 zones and in the zones of union Z2 Z3 of the network cells of the cell cluster 130 are then defined as network traffic classes (user presence, or average number of active users). ROPs 20 ROP (i = 1, 2, …, P) of the first acquisition time interval INT ei ROP A,i A ROP (j = 1, 2, …, Q) of the second acquisition time interval INT (e, where B,j B defined, also the ROP ROP , k = 1, 2, …, N, of the acquisition time interval R,k (of reference INT ) are classified in the appropriate class, i.e. in the bins Bin or Bin . RLH Subsequently, as described in relation to the first embodiment 25 of the invention, the weights to be assigned to the ROPs ROP (j = 1, 2, …, Q) are calculated B,j of the second acquisition time interval INT (and, when used B the reference acquisition time interval INT , also the weights to be assigned to the R ROP ROP , i = 1, 2, …, P, of the first acquisition time interval INT ) A,i A . The reward functions W and W for the two network configurations, first and AB 30 after the change, they are calculated and compared with each other, to evaluate whether the new network configuration achieved improved QoS. Combinations of the two homogenization methods are also possible. network traffic described above, where for each class determined according to the first network traffic homogenization method subclasses are identified in 5 agreement with the second method of network traffic homogenization.    While in the embodiments of the invention described above the network performance evaluation metric, or QoS evaluation metric, took into account user throughput, other indicators can be used for the 10 QoS evaluation, in addition to user throughput. For example, the metric of QoS evaluation may also take into account, in addition to the throughput of user, parameters that provide an indication of the number of handovers and / or parameters which provide an indication of how many mobile network base stations are "active" or "on", that is, not "off": in fact, the greater the number of base stations 15 “on” the greater the user throughput, but this comes at the cost of a greater energy consumption of the mobile network; the metric can then be defined in order to apply increasing penalties for an increasing number of active base stations). In each case, the specific metric defined for evaluating network performance does not constitutes a limitation of the present invention. 20 Naturally, in order to meet specific local needs, a technician of the branch will be able to make numerous modifications and alterations to the invention described above of a logical and / or physical nature. More specifically, although the present invention is has been described with some degree of particularity with reference to its forms of preferred implementation, it is understood that various omissions, substitutions and 25 changes in form and details as well as other embodiments. In In particular, different embodiments of the invention may be implemented even without the specific details given in the previous description to provide a deeper understanding; on the contrary, well-known features can be have been omitted or simplified so as not to weigh down the description with unnecessary details. 30 necessary. Furthermore, it is expressly provided that specific elements and / or phases of the method described in connection with any disclosed embodiment of the invention may be incorporated into any other embodiment. * * * * * 5 TRANSLATION OF THE WORDS IN THE FIGURES Fig. 1: Core Network QoS Improvement Assessment Fig. 2: Select Optimization Algorithm Select Optimization Algorithm Define QoS Evaluation Metric Determine Optimization Needed 15 Define INT, INT, ROPs Define INT, INT, ROPs ABAB Define Traffic Homoc. Metric Define Traffic Homog. Classes Get KPIs in INT Get KPIs in INT AA Deploy Optimization Deploy Optimization 20 Get KPIs in INT Get KPIs in INT BB Classify ROPs In Traffic Classify ROPs Into Homogenous Classes. Homog. Classes Traffic Calculate ROPs Weights Apply QoS Metric to ROPs Apply QoS Metric to ROPs 25 Assess Optimization Effectiveness Fig. 7: QoS Improvement Assessment Definition of Acquisition Definition of t & ROP Acquisition ts & ROPs 5 KPIs Acquisition & Aggregation for ROP for ROPs Statistical Processing ROPs Classification ROP Classification ROPs Weights Calculator ROP Weights Calculator 10 Reward Calculator Reward Calculator Reward Comparator QoS Improvement Y / N QoS Improvement Y / N

Claims

1. A computer-implemented method for evaluating the change in the quality of service, QoS, offered by a mobile communications network, resulting from a change in the configuration parameters of one or more network cells (115a) of the mobile communications network (100), the method comprising: - defining a network cell aggregate (130) comprising said one or more network cells (115a) and neighbouring network cells (115b) adjacent to said one or more network cells (115a), the network cell aggregate being defined to correspond to a coverage area of ​​the mobile communications network which remains substantially constant before and after the change in the configuration parameters of the one or more network cells (115a), - acquiring network measurements with respect to an area of ​​the mobile communications network corresponding to the network cell aggregate (130), wherein said acquiring network measurements comprises: - acquiring,during a first acquisition time interval (INTa) before said configuration parameter change, first network traffic quantity indicators and first QoS indicators, and - acquire, during a second acquisition time interval (INTb) after said configuration parameter change, second network traffic quantity indicators and second QoS indicators,- deriving from the acquired first network traffic quantity indicators a first distribution of the network traffic quantity (605a) during the first acquisition time interval (INTa); - deriving from the acquired second network traffic quantity indicators a second distribution of the network traffic quantity (605b) during the second acquisition time interval (INTb); - obtaining a reference distribution of the network traffic quantity (605a; 905r) in the network cell aggregate (130) during a reference time interval (INTr; INTa); - calculating a first reward based on said first QoS indicators and calculating a second reward based on said second QoS indicators; - evaluating the QoS variation based on the comparison between the first reward and the second reward,wherein said calculating the first reward and the second reward comprises: calculating first weights to be applied to the first QoS indicators and second weights to be applied to the second QoS indicators, said first weights and second weights being calculated so as to make the first distribution of the amount of network traffic (605a) and the second distribution of the amount of network traffic (605b) correspond to the reference distribution of the amount of network traffic (605a; 905r)., 2. The method according to claim 1, wherein said first acquisition time interval (INTa) comprises a first succession of first elementary acquisition time intervals (ROPa) and said second acquisition time interval (INTb) comprises a second succession of second elementary acquisition time intervals (ROPb), the first network traffic quantity indicators and the first QoS indicators being acquired during each first elementary acquisition time interval (ROPa) of the first succession, and the second network traffic quantity indicators and the second QoS indicators being acquired during each second elementary acquisition time interval (ROPb) of the second succession.

3. The method according to claim 2, wherein said first distribution of the amount of network traffic (605a) during the first acquisition time interval (INTa) is a distribution of the first elementary acquisition time intervals (ROPa) as a function of the first indicators of the amount of network traffic, and said second distribution of the amount of network traffic (605b) during the second acquisition time interval (INTb) is a distribution of the second elementary acquisition time intervals (ROPb) as a function of the first indicators of the amount of network traffic.

4. The method according to any preceding claim, wherein said obtaining a reference distribution of the amount of network traffic (905r) comprises: - defining a reference acquisition time interval (INTr), wherein said reference acquisition time interval comprises a third succession of third elementary acquisition time intervals (ROPr); - acquiring, during each third elementary acquisition time interval (ROPr) of the third succession, third network traffic amount indicators and third QoS indicators; - deriving from the third network traffic amount indicators the reference distribution of the amount of network traffic (605r), the reference distribution of the amount of network traffic being a distribution of the third elementary acquisition time intervals (ROPr) as a function of the third network traffic amount indicators.

5. The method according to any of claims 1-3, wherein said reference distribution of the amount of network traffic (605a) is the first distribution of the amount of network traffic (605a) during the first acquisition time interval (INTa), said obtaining a reference distribution of the amount of network traffic including said deriving from the first indicators of the amount of network traffic the first distribution of the amount of network traffic (605a) during the first acquisition time interval (INTa), and wherein said first weights to be applied to the first QoS indicators are unit weights and said second weights are calculated so as to make the second distribution of the amount of network traffic (605b) correspond to the first distribution of the amount of network traffic (605a). I24021-IT / MM 6. The method according to claim 4 or 5, wherein said weights are calculated so as to obtain, from the distribution of the second elementary acquisition time intervals (ROPb) as a function of the first indicators of the amount of network traffic, a remodeled distribution having a shape corresponding to one of: the distribution of the third elementary acquisition time intervals (ROPr) as a function of the third indicators of the amount of network traffic, the distribution of the first elementary acquisition time intervals (ROPa) as a function of the first indicators of the amount of network traffic.

7. The method according to any of claims 4-6, wherein said calculating the weights to be applied to the second network measurements comprises: - defining a plurality of network traffic quantity classes (BinA,i - BinA,5; BinR,i - BinR,5) with respect to the distribution of the reference network traffic quantity, each network traffic quantity class of said plurality of network traffic quantity classes corresponding to a respective reference network traffic quantity; - classifying each of the first elementary acquisition time intervals (ROPa) and each of the second elementary acquisition time intervals (ROPb) into a respective class of said plurality of classes according to the respective first and second network traffic quantity indicators;- determine a number of first elementary acquisition time intervals (ROPa) and second elementary acquisition time intervals (ROPb) in each class of said plurality of classes, and - calculate the weights to be applied to the second network measurements based on the determined number of first elementary acquisition time intervals (ROPa) and second elementary acquisition time intervals (ROPb) in each class of said plurality of classes.; 8. The method of any of the preceding claims, comprising I24021-IT / MM aggregating, over all network cells of the cell aggregate (130), the first network traffic quantity indicators and the first acquired QoS indicators, and the second network traffic quantity indicators and the second acquired QoS indicators.

9. The method of any preceding claim, wherein: - said first and second network traffic quantity indicators comprise at least one of measurements of traffic volume, number of active users, user presences, and - said first and second QoS indicators comprise indicators of a distribution of network traffic volumes as a function of the user throughput that the mobile communications network is capable of providing to users.

10. The method according to any preceding claim, wherein said first and second network traffic quantity indicators and said first and second QoS indicators are acquired from the mobile communications network as aggregated data at the level of each network cell of the cell aggregate (130).

11. A method according to any of claims 1-9, wherein said first and second network traffic quantity indicators and said first and second QoS indicators are acquired as geo-referenced data, in particular MDT data, the method further comprising aggregating the acquired geo-referenced data at the level of sub-areas of the coverage area of ​​each network cell of the cell aggregate (130).

12. A system for evaluating the change in the Quality of Service, QoS, offered by a mobile communications network, resulting from a change in the configuration parameters of one or more network cells (115a) of the mobile communications network (100), the system being configured to: - acquire network measurements with respect to an area of ​​the mobile communications network corresponding to an aggregate of network cells (130) comprising said one or more network cells (115a) and neighbouring network cells (115b) adjacent to said one or more network cells (115a), the aggregate of network cells corresponding to a coverage area of ​​the mobile communications network which remains substantially constant before and after the change in the configuration parameters of the one or more network cells (115a), wherein the system is configured to acquire said network measurements by: - ​​acquiring, during a first acquisition time interval (INTa) prior to said change in the configuration parameters of the one or more network cells (115a),configuration parameters, first network traffic quantity indicators and first QoS indicators, and - acquiring, during a second acquisition time interval (INTb) after said modification of the configuration parameters, second network traffic quantity indicators and second QoS indicators; - obtaining a reference distribution of the amount of network traffic (605a; 905r) suitable for providing a reference distribution of the amount of network traffic in the aggregate of network cells (130) during a reference time interval (INTr; INTa); - deriving from the first network traffic quantity indicators a first distribution of the amount of network traffic (605a) during the first acquisition time interval (INTa); - deriving from the acquired second network traffic quantity indicators a second distribution of the amount of network traffic (605B) during the second acquisition time interval (INTb);- calculating a first reward based on said first QoS indicators and calculating a second reward based on said second QoS indicators; - evaluating the QoS variation based on a comparison between the first reward and the second reward, wherein the system is configured to calculate the first reward and the second reward by: calculating first weights to be applied to the first QoS indicators and second weights to be applied to the second QoS indicators, said first weights and second weights being calculated so as to make the first distribution of the amount of network traffic (605a) and the second distribution of the amount of network traffic (605b) corresponding to the reference distribution of the amount of network traffic (605a; 905r).