OBJECTIVE: To train and test a model predicting chronic kidney disease (CKD) using the Generalized Additive2 Model (GA2M), and compare it with other models being obtained with traditional or machine learning approaches. MATERIALS: We adopted the Health Search Database (HSD) which is a representative longitudinal database containing electronic healthcare records of approximately 2 million adults. METHODS: We selected all patients aged 15 years or older being active in HSD between January 1, 2018 and December 31, 2020 with no prior diagnosis of CKD. The following models were trained and tested using 20 candidate determinants for incident CKD: logistic regression, Random Forest, Gradient Boosting Machines (GBMs), GAM, and GA2M. Their prediction performances were compared by calculating Area Under Curve (AUC) and Average Precision (AP). RESULTS: Comparing the predictive performances of the 7 models, the AUC and AP for GBM and GA2M showed the highest values which were equal to 88.9%, 88.8% and 21.8%, 21.1%, respectively. These 2 models outperformed the others including logistic regression. In contrast to GBMs, GA2M kept the interpretability of variable combinations, including interactions and nonlinearities assessment. DISCUSSION: Although GA2M is slightly less performant than light GBM, it is not "black-box" algorithm, so being simply interpretable using shape and heatmap functions. This evidence supports the fact machine learning techniques should be adopted in case of complex algorithms such as those predicting the risk of CKD. CONCLUSION: The GA2M was reliably performant in predicting CKD in primary care. A related decision support system might be therefore implemented.

Base station (BS) densification is increasingly adopted by mobile operators in order to support increasing traffic demand. However, a large number of BSs requires many backhaul connections, which may be very expensive. For this reason, provisioning high speed backhaul connections to BSs in a cost-effective way is challenging, and it is important to efficiently utilize an existing fixed network infrastructure if possible. This paper proposes two optimized infrastructure-aware planning strategies for small cells and fiber backhaul. The first strategy is referred to as joint design (JD) and is based on the joint cost minimization of small cells and fiber backhaul. The JD strategy is compared to a traditional design (TD) solution based on a two-step optimization approach. In the latter a cost-optimal small cells placement is identified first, then the corresponding minimum cost fiber backhaul deployment is determined accordingly. A comparison between these two approaches in dense 5G urban scenarios shows that the relative performance of JD and TD largely depends on the expected traffic demand and on the existing infrastructure. In dense urban areas with the average traffic levels expected for the year 2020 and beyond, JD returns up to 50% lower deployment cost in a greenfield scenario, and up to 70% lower deployment cost in a brownfield scenario. Only in areas with extremely high traffic demand (e.g., open-air festivals and stadiums) JD returns deployment costs very similar to TD.

Providing high capacity to the end users is one of the main challenges for the fifth generation (5G) of mobile networks. The users' habit to consume online contents indoor makes the outdoor-to-indoor capacity provisioning impractical, especially when the high-frequency bands proposed for 5G are employed. The centralized radio architecture (CRA) is an in-building solution, which relies on the centralization of baseband processing functions, fully or partly allowing for centralized cell management while providing signals directly inside the buildings. On the other hand, the massive deployment of CRAs in urban areas may yield to unacceptably high installation costs, due to the radio network equipment to be activated. To make CRAs appealing to mobile operators, we propose different deployment strategies to minimize the CRA deployment cost. We define the remote radio unit placement (RRUP) problem and formulate it as an integer linear program, obtaining optimal deployment solutions in small urban residential scenarios. We prove the RRUP problem to be NP-hard, requiring heuristic approaches to solve large problem instances. To this end, we propose an effective and scalable heuristic for minimizing the amount of radio equipment required to deploy CRAs in large urban areas.

To efficiently handle the fast growing traffic inside data centers, several optical interconnect architectures have been recently proposed. However, most of them are targeting the aggregation and core tiers of the data center network, while relying on conventional electronic top-of-rack (ToR) switches to connect the servers inside the rack. The electronic ToR switches pose serious limitations on the data center network in terms of high cost and power consumption. To address this problem, we recently proposed a passive optical top-of-rack interconnect architecture, where we focused on the data plane design utilizing simple passive optical components to interconnect the servers within the rack. However, an appropriate control plane tailored for this architecture is needed to be able to analyze the network performance, e.g., packet delay, drop rate, etc., and also obtain a holistic network design for our passive optical top-of-rack interconnect, which we refer to as POTORI. To fill in this gap, this paper proposes the POTORI control plane design which relies on a centralized rack controller to manage the communications inside the rack. To achieve high network performance in POTORI, we also propose a centralized medium access control (MAC) protocol and two dynamic bandwidth allocation (DBA) algorithms, namely Largest First (LF)and Largest First with Void Filling (LFVF). Simulation results show that POTORI achieves packet delays in the order of microseconds and negligible packet loss probability under realistic data center traffic scenarios.

This paper deals with resource management in data centre networks that feature optical interconnects. It first proposes an optical resource management framework as a platform to develop different solutions for multipoint-to-multipoint optical communication systems with a centralized controller. The paper takes optical intra-rack communications as an example application scenario of the framework and studies the intra-rack scheduling (IRS) problem using a theoretical approach. The problem is mapped into the classical open-shop scheduling problem, in which the optical interfaces are viewed as the jobs and the wavelengths used for communication among the interfaces are viewed as the machines. The IRS problem can therefore be solved by adopting the existing preemptive and/or non-preemptive open-shop scheduling algorithms. In a realistic intra-rack communication scenario with non-negligible network reconfiguration time, it is important to understand whether the preemptive or non-preemptive strategy provides better performance under given traffic conditions and system configurations. To address the question, this paper presents a performance analysis that allows to quantitatively compare the two scheduling strategies in terms of packet delay and potential energy savings obtained from sleep mode implementation. Analytical results reveal that the non-preemptive strategy outperforms the preemptive one for typical configurations used in data centre networks. In addition, the tuning and wake-up times of optical transceivers are shown to be key performance-determining factors. The proposed framework and obtained findings are not limited to the considered intra-rack communication scenario, but can be applied to any centralized optical switching systems featuring multipoint-to-multipoint transmissions with non-negligible reconfiguration time.

Each mobile network architecture able to meet the traffic requirements of future 5G services comes with its own set of benefits vs. requirements for the radio and the transport segments. This paper focuses on the energy performance of four mobile network architectures, each one with different splitting options for the baseband processing functions. The radio segment under exam is based on a new radio access technology referred to as 5G-NX, while the transport segment is based on wavelength division multiplexing (WDM). The energy consumption of each mobile network architecture is weighted against (i) the benefits for the radio segment as a function of the level of centralization of the baseband processing resources and (ii) the required transport capacity and consequently the power consumption levels needed to accommodate the backhaul, midhaul, and/or fronthaul traffic generated at each base station. Our results show that a fully centralized radio access network (C-RAN) with centralization of all the baseband functions is not a practical approach for 5G mobile networks, while a partially centralized C-RAN architecture in which the physical layer baseband processing is performed at the base station site represents a promising solution.

Mobile networks are the largest contributor to the carbon footprint of the telecom sector and their contribution is expected to rapidly increase in the future due to the foreseen traffic growth. Therefore, there is an increasing urgency in the definition of green mobile network deployment strategies. This paper proposes a four-step design and power assessment methodology for mobile networks, taking into consideration both radio and transport segments. A number of mobile network deployment architectures for urban residential areas based on different radio (i.e., macro base station, distributed indoor radio, femto cell) and transport (i.e., microwave, copper, optical fiber) technologies are proposed and evaluated to identify the most energy efficient solution. The results show that with low traffic the conventional macro base station deployment with microwave based backhaul is the best option. However, with higher traffic values heterogeneous networks with macro base stations and indoor small cells are more energy efficient. The best small cell solution highly depends on the transport network architecture. In particular, our results show that a femto cell based deployment with optical fiber backhaul is the most energy efficient, even if a distributed indoor radio architecture (DRA) deployment with fiber fronthaul is also a competitive approach.

Mobile operators are facing an exponential traffic growth due to the proliferation of portable devices that require a high-capacity connectivity. This, in turn, leads to a tremendous increase of the energy consumption of wireless access networks. A promising solution to this problem is the concept of heterogeneous networks, which is based on the dense deployment of low-cost and low-power base stations, in addition to the traditional macro cells. However, in such a scenario the energy consumed by the backhaul, which aggregates the traffic from each base station towards the metro/core segment, becomes significant and may limit the advantages of heterogeneous network deployments. This paper aims at assessing the impact of backhaul on the energy consumption of wireless access networks, taking into consideration different data traffic requirements (i.e., from todays to 2020 traffic levels). Three backhaul architectures combining different technologies (i.e., copper, fiber, and microwave) are considered. Results show that backhaul can amount to up to 50% of the power consumption of a wireless access network. On the other hand, hybrid backhaul architectures that combines fiber and microwave performs relatively well in scenarios where the wireless network is characterized by a high small-base-stations penetration rate.

In this paper the performance of a hybrid optical network composed of multiple Ethernet Passive Optical Networks (EPONs) at the access connected to an Optical Burst Switching (OBS) core is evaluated. The EPONs make use of the IPACT algorithm to manage the upstream transmissions, whereas the OBS edge node employs a time-based assembly scheme for burst aggregation. This leads to interoperability issues between access and core which are investigated in this paper. The hybrid network scalability is proved by assessing the transport layer performance with different numbers of EPONs served by the same burst assembler and different numbers of Optical Network Units (ONUs) per EPON. The impact of the main design parameters affecting the interaction between EPONs and OBS is estimated. The transport layer performance is evaluated considering heterogeneous traffic flows by simulating both TCP and UDP sources.