Off-line Data Research Articles

Research on Construction of Offline Data Warehouse for Ship Shore Power based on DolphinScheduler and Hive

In today's networked, intelligent, and data-driven era, the shore power industry is facing the challenge of rapidly growing data. This paper presents the construction of a professional offline data warehouse system for shore power based on DolphinScheduler and Hive. Firstly, MySQL is adopted as the backend database, combined with Sqoop to synchronize business data to HDFS, ensuring data reliability and integrity. Secondly, the Flume-Kafka-Flume architecture is utilized to achieve real-time collection and caching of user behavior data, providing data support for subsequent analysis. Thirdly, HQL statements are written in Hive to clean, merge, and analyze shore power data, calculating key indicators such as electricity consumption and usage trends. Fourthly, data visualization is achieved through the integration of Superset, displaying data analysis results via a web interface. Fifthly, DolphinScheduler is employed for timed scheduling, ensuring dependency control among various tasks and the smooth operation of the project. This system fully leverages the replication mechanism of HDFS to enhance reliability, dynamically adds nodes to achieve system scalability, and fully utilizes the fault tolerance of the Yarn scheduler. It saves time and computational costs for the shore power industry, realizing higher value and benefits.

A Data-Driven Approach for Leveraging Inline and Offline Data to Determine the Causes of Monoclonal Antibody Productivity Reduction in the Commercial-Scale Cell Culture Process.

The monoclonal antibody (mAb) manufacturing process comes with high profits and high costs, and thus mAb productivity is of vital importance. However, many factors can impact the cell culture process, and lead to mAb productivity reduction. Nowadays, the biopharma industry is actively employing manufacturing information systems, which enable the integration of both online data and offline data. Although the volume of data is large, related data mining studies for mAb productivity improvement are rare. Therefore, a data-driven approach is proposed in this study to leverage both the inline and offline data of the cell culture process to discover the causes of mAb productivity reduction. The approach consists of four steps, namely data preprocessing, phase division, feature extraction and fusion, and cluster comparing. First, data quality issues are solved during the data preprocessing step. Next, the inline data are divided into several phases based on the moving window k-nearest neighbor method. Then, the inline data features are extracted via functional data analysis and combined with the offline data features. Finally, the causes of mAb productivity reduction are identified using the contrasting clusters via the principal component analysis method. A commercial-scale cell culture process case study is provided in this research to verify the effectiveness of the approach. Data from 35 batches were collected, and each batch contained nine inline variables and seven offline variables. The causes of mAb productivity reduction were identified to be the lack of nutrients, and recommended actions were taken according to the result, which was subsequently proven by six validation batches.

An online battery–state of charge estimation method using the varying forgetting factor recursive least square - unscented Kalman filter algorithm on electric vehicles

Accurate and fast estimation of the state of charge is important for the battery management system of electric vehicles. This paper proposes a method to estimate the state of charge of Lithium-ion batteries by the <span lang="EN-US">variable</span> forgetting factor recursive least square (VFFRLS) – unscented Kalman filter (UKF) algorithm in real-time without the off-line battery testing data. Since the state observation requires an accurate model, an equivalent circuit model was constructed. Then, the VFFRLS algorithm is used to identify online the battery model parameters based on voltage and current measurements. An advantage of this algorithm is that it requires less initial information and shorter identification time than offline parameter identification. After the model parameters are well identified, the unscented Kalman filter estimates the state of charge and minimizes noise characteristics and uncertainty in the parameter identification process. The VFFRLS algorithm applied in this paper has shown a good result with the model output error of less than 1%, and the identification achieves real-time response. The state of charge obtained by the UKF algorithm has shown satisfactory estimation results with fast convergence speed and small errors. The UKF filter provides the results with a 1.5% error from the reference and converges after 10 cycles.

Robustly Improving Bandit Algorithms with Confounded and Selection Biased Offline Data: A Causal Approach

This paper studies bandit problems where an agent has access to offline data that might be utilized to potentially improve the estimation of each arm’s reward distribution. A major obstacle in this setting is the existence of compound biases from the observational data. Ignoring these biases and blindly fitting a model with the biased data could even negatively affect the online learning phase. In this work, we formulate this problem from a causal perspective. First, we categorize the biases into confounding bias and selection bias based on the causal structure they imply. Next, we extract the causal bound for each arm that is robust towards compound biases from biased observational data. The derived bounds contain the ground truth mean reward and can effectively guide the bandit agent to learn a nearly-optimal decision policy. We also conduct regret analysis in both contextual and non-contextual bandit settings and show that prior causal bounds could help consistently reduce the asymptotic regret.

Net Power Optimization Based on Extremum Search and Model-Free Adaptive Control of PEMFC Power Generation System for High Altitude

The operating characteristics of proton exchange membrane fuel cell (PEMFC) generation systems will change in high altitude. In order to improve the system output performance, this paper proposes a net power optimization (NPO) based on extremum search and model-free adaptive control (MFAC) of PEMFC power generation system. A PEMFC power generation system model for high-altitude environment is established, the off-line analysis shows that with the increase of altitude, the performance of the air compressor decreases, the optimal oxygen excess ratio (OER) of the system decreases, and surge phenomenon is more likely to occur. Considering the non-minimum phase characteristic of the net power to OER, the OER is optimized online by extremum search strategy and the optimization step and cycle ensure the convergence of the strategy. Combined with a model-free adaptive control strategy based on compact form dynamic linearization (CFDL), the optimal OER is tracked real time, the data model identified online ensures the adaptability of the system. Finally, a hardware-in-the-loop platform is built and comparative experiment is carried out. The results show that: compared with the off-line optimization method (OOM), (1) NPO can improve the net power of the system at different altitudes through the adaptive adjustment of OER and no off-line data is required inappropriate static operating points are avoided. (2) in high altitude areas, NPO has higher operational stability and can effectively avoid surge phenomenon.

Design and Implementation of Power Big Data Platform

In recent years, with the advancement of the national construction of smart grid and the development of power grid enterprise integration system, the traditional power data platform has some defect, such as insufficient scalability, repeated implementation of two sets of logic of off-line data warehouse and real-time data warehouse, highly correlated storage of data warehouse during computation, and difficulty in platform migration, which restrict the deeper mining and analysis of power big data. Therefore, this paper adopts a technical architecture incorporating distributed calculating, micro-service, flow-batch integration and lake-warehouse integration, combining the current popular Web front-end and back-end technologies such as Vue, SpringCloud and Flask, and utilizing big data components such as Hadoop, Flink, Hudi and Kafka as well as Docker containerization technology. We explore and build a power grid system operation oriented big data platform that assembles real-time computing, multi-source heterogeneous data storage, and develop APIs for authority management, information management, data analysis and visualization, machine learning and other data supporting services that assist power enterprises in managing electrical equipment and users’ power consumption.With the help of this platform, power enterprises can intelligently track the equipment status, view the offline and real-time visual charts to analyze the power consumption by users, so as to have better basis in the power dispatching, maintenance and scheme adjustment of electrical market.

An unsupervised machine learning approach to reduce nonlinear FE[formula omitted] multiscale calculations using macro clustering

Solving nonlinear multiscale methods with history-dependent behaviors and fine macroscopic meshes is a well-known challenge. In this work, an unsupervised machine learning-based clustering approach is developed to reduce nonlinear Multilevel Finite Element-FE2 calculations. In contrast with most available techniques which aim at developing Reduced Order Models (ROM) or AI-based surrogate models for the microscale nonlinear problems, the present technique reduces the problem from the macro scale by creating clusters of macro Gauss points which are assumed to be in close mechanical states. Then, a single micro nonlinear Representative Volume Element (RVE) calculation is performed for each cluster. A linear approximation of the macro stress is used in each cluster. Handling internal variables is carried out by using anelastic macro strains in the clustering vectors in addition to the macro strains components. Finally, some convergence issues related to the use of clusters at the macro scale are addressed through a cluster freezing algorithm. The technique is applied to nonlinear hyperelastic, viscoelastic and elastoplastic composites. In contrast to available ROM or machine-learning -based acceleration techniques, the present method does not require neither preliminary off-line calculations, nor training, nor data base, nor reduced basis at the macro scale, while maintaining typical speed-up factors about 20 as compared to classical FE2.

Feature Selection in the Data Stream Based on Incremental Markov Boundary Learning.

Recent years have witnessed the proliferation of techniques for streaming data mining to meet the demands of many real-time systems, where high-dimensional streaming data are generated at high speed, increasing the burden on both hardware and software. Some feature selection algorithms for streaming data are proposed to tackle this issue. However, these algorithms do not consider the distribution shift due to nonstationary scenarios, leading to performance degradation when the underlying distribution changes in the data stream. To solve this problem, this article investigates feature selection in streaming data through incremental Markov boundary (MB) learning and proposes a novel algorithm. Different from existing algorithms focusing on prediction performance on off-line data, the MB is learned by analyzing conditional dependence/independence in data, which uncovers the underlying mechanism and is naturally more robust against the distribution shift. To learn MB in the data stream, the proposal transforms the learned information in previous data blocks to prior knowledge and employs them to assist MB discovery in current data blocks, where the likelihood of distribution shift and reliability of conditional independence test are monitored to avoid the negative impact from invalid prior information. Extensive experiments on synthetic and real-world datasets demonstrate the superiority of the proposed algorithm.

Prediction of systemic free and total valproic acid by off-line analysis of exhaled breath in epileptic children and adolescents.

Therapeutic drug monitoring (TDM) of medications with a narrow therapeutic window is a common clinical practice to minimize toxic effects and maximize clinical outcomes. Routine analyses rely on the quantification of systemic blood concentrations of drugs. Alternative matrices such as exhaled breath are appealing because of their inherent non-invasive nature. This is especially the case for pediatric patients. We have recently showcased the possibility of predicting systemic concentrations of valproic acid (VPA), an anti-seizure medication by real-time breath analysis in two real clinical settings. This approach, however, comes with the limitation of the patients having to physically exhale into the mass spectrometer. This restricts the possibility of sampling from patients not capable or available to exhale into the mass spectrometer located on the hospital premises. In this work, we developed an alternative method to overcome this limitation by collecting the breath samples in customized bags and subsequently analyzing them by secondary electrospray ionization coupled to high-resolution mass spectrometry (SESI-HRMS). A total ofn= 40 patients (mean ± SD, 11.5 ± 3.5 y.o.) diagnosed with epilepsy and taking VPA were included in this study. The patients underwent three measurements: (i) serum concentrations of total and free VPA, (ii) real-time breath analysis and (iii) off-line analysis of exhaled breath collected in bags. The agreement between the real-time and the off-line breath analysis methods was evaluated using Lin's concordance correlation coefficient (CCC). CCC was computed for ten mass spectral predictors of VPA concentrations. Lin's CCC was >0.6 for all VPA-associated features, except for two low-signal intensity isotopic peaks. Finally, free and total serum VPA concentrations were predicted by cross validating the off-line data set. Support vector machine algorithms provided the most accurate predictions with a root mean square error of cross validation of 29.0 ± 7.4 mg l-1and 3.9 ± 1.4 mg l-1for total and free VPA (mean ± SD), respectively. As a secondary analysis, we explored whether exhaled metabolites previously associated with side-effects and response to medication could be rendered by the off-line analysis method. We found that five features associated with side effects showed a CCC > 0.6, whereas none of the drug response-associated peaks reached this cut-off. We conclude that the clinically relevant free fraction of VPA can be predicted by this combination of off-line breath collection with rapid SESI-HRMS analysis. This opens new possibilities for breath based TDM in clinical settings.

An Efficient Online Offline Data Collection Software Solution for Creating Reference Population to Established Genomic Selection Under Small Holders’ Dairy System in India

This article as part of an Enhanced Genetics Project assumes special significance in the light of emergence of electronic mode of data collection system. Pen and paper method of data collection although very familiar to field people, but is generates the errors and requires digitization and data cleaning. The software solution surveying is easier mode of ensuring uniformity in data collection, reducing errors and better storage and analysis of data. In this regard, this article gives a clearer picture of the digital tablet based data collection system employed part of the Enhanced Genetics Project across 49 selected Data Recorders (Data Enumerators) in Maharashtra, Uttar Pradesh, Jharkhand, Punjab, Odisha, Gujarat, Andhra Pradesh, Rajasthan, and Bihar. The manual outlines the processes to be fulfilled by the digital mobile based solution, the user interface involved and the software and hardware infrastructure to support such a platform independent system. In this project we have enrolled the 33,293 dairy animals and collecting there production, reproduction, body weight data, as well as, collecting the baseline & herd management information of that farm. This system increase Data Accuracy GPS (Global Positioning System) enabled data collection system for genomic evolution. This is very unique software solution developed for to collect the structured phenotype data of the dairy animals in Indian conditions on realtime basis. As well, show case the robust data collection system for the Genomic Selection for dairy animal’s initiative. This sub-system also focused on the various activities and operations related to molecular biological laboratory. In this software solution system the two major feedback systems for actively participation of the farmers were participation certificate and Graphical Animal Ranking system for the farmer and animal. The locational data collection with seamless interfacing with the user and the data repository for Genetic Improvement Data collection system. Data capture system was quite successful in gathering field phenotype data for Genetic Improvement Program in the complex dairy production system like India.

Data-Driven Reduced Order Models Using Invariant Foliations, Manifolds and Autoencoders

This paper explores how to identify a reduced order model (ROM) from a physical system. A ROM captures an invariant subset of the observed dynamics. We find that there are four ways a physical system can be related to a mathematical model: invariant foliations, invariant manifolds, autoencoders and equation-free models. Identification of invariant manifolds and equation-free models require closed-loop manipulation of the system. Invariant foliations and autoencoders can also use off-line data. Only invariant foliations and invariant manifolds can identify ROMs, and the rest identify complete models. Therefore, the common case of identifying a ROM from existing data can only be achieved using invariant foliations. Finding an invariant foliation requires approximating high-dimensional functions. For function approximation, we use polynomials with compressed tensor coefficients, whose complexity increases linearly with increasing dimensions. An invariant manifold can also be found as the fixed leaf of a foliation. This only requires us to resolve the foliation in a small neighbourhood of the invariant manifold, which greatly simplifies the process. Combining an invariant foliation with the corresponding invariant manifold provides an accurate ROM. We analyse the ROM in case of a focus type equilibrium, typical in mechanical systems. The nonlinear coordinate system defined by the invariant foliation or the invariant manifold distorts instantaneous frequencies and damping ratios, which we correct. Through examples we illustrate the calculation of invariant foliations and manifolds and at the same time show that Koopman eigenfunctions and autoencoders fail to capture accurate ROMs under the same conditions.

An Integrated Off-Line Echo Signal Acquisition System Implemented in SoC-FPGA for High Repetition Rate Lidar

High repetition rate lidar is typically equipped with a low-energy, high repetition rate laser, and small aperture telescopes. Therefore, it is small, compact, low-cost, and can be networked for observation. However, its data acquisition and control functions are generally not specially designed, and the data acquisition, storage, and control programs need to be implemented on an IPC (Industrial Personal Computer), which increases the complexity and instability of the lidar system. Therefore, this paper designs an integrated off-line echo signal acquisition system (IOESAS) for lidar developed based on SoC FPGA (System-On-Chip Field Programmable Gate Array). Using a hardware–software co-design approach, the system is implemented in a heterogeneous multi-core chip ZYNQ-7020 (integrated FPGA and ARM). The FPGA implements dual-channel echo data acquisition (gated counting and hardware accumulation). At the same time, the ARM performs laser control and monitoring, laser pointing control, pulse energy monitoring, data storage, and wireless transmission. Offline data acquisition and control software was developed based on LabVIEW, which can remotely control the status of the lidar and download the echo data stored in IOESAS. To verify the performance of the data acquisition system, IOESAS was compared with the photon counting card P7882 and MCS-PCI, respectively. The test results show that they are in good agreement; the linear correlation coefficients were 0.99967 and 0.99884, respectively. IOESAS was installed on lidar outdoors for continuous detection, and the system was able to work independently and stably in different weather conditions, and control functions were tested normally. The gating delay and gating width time jitter error are ±5 ns and ±2 ns, respectively. The IOESAS is now used in several small lidars for networked observations.

Off-line measuring sampling data identification parameters for digital twins mirroring load modelling and stability analysis

Currently, the methods used to represent loads do not differ between the characteristics that compose them or the nature of these. Therefore, the purpose of this research is to develop digital twins mirroring load models that can be used for more precise studies on power-flows and stability within the National Transmission Grid (NTG). Off-line sampling data of different electric measurements have been used in six substations of the Guayaquil (Ecuador) area. These values were organized by statistical methods and by time periods, to determine the parameters that make up the static load model. Dynamic models are also constructed for the same six substations using the analysis of current and voltage signals obtained from the substations. All data is organized to show a digital twin mirroring visual representation of the disturbances that may occur in the substation buses. A more accurate description of the static and dynamic responses can be obtained by replacing the general model that is currently used by engineers and planners with off-line sampling data. Digital twins help the electric utility businesses gather, visualise, and contextualise data from different sources, and enable to act on data, and to understand what-if modelling stability scenarios.

Measuring Policy Performance in Online Pricing with Offline Data: Worst-case Perspective and Bayesian Perspective

Measuring Policy Performance in Online Pricing with Offline Data: Worst-case Perspective and Bayesian Perspective

Deep active learning for constitutive modelling of granular materials: From representative volume elements to implicit finite element modelling

Constitutive relation remains one of the most important, yet fundamental challenges in the study of granular materials. Instead of using closed-form phenomenological models or numerical multiscale modelling, machine learning has emerged as an alternative paradigm to revolutionise the constitutive modelling of granular materials. However, deep neural networks (DNNs) require massive training data and often fail to make credible extrapolations. This study aims to develop a deep active learning strategy to (i) identify unreliable forecasts without knowing the ground truth; and (ii) continuously improve and verify a data-driven constitutive model until the desired generalisation is satisfied. The role of active learning in constitutive modelling is instantiated through three scenarios: (i) off-line strain-stress data pool of granular materials; (ii) interactive constitutive training and strain-stress data labelling; and (iii) finite element modelling (FEM) driven by deep learning-based constitutive models. The results confirm the capability of active learning in advancing data-driven constitutive modelling of granular materials toward developing a faithful surrogate constitutive model with less data. The same active learning strategy can also be applied to other data-centric applications across various science and engineering fields.

Learning configurations of wires for real-time shape estimation and manipulation planning

Robotic manipulation of a wire by its ends requires rapid reasoning of its shape in real-time. A recent development of an analytical model has shown that sensing of the force and torque on one end can be used to determine its shape. However, the model relies on assumptions that may not be met in real world wires and do not take into account gravity and non-linearity of the Force/Torque (F/T) sensor. Hence, the model cannot be applied to any wire with accurate shape estimation. In this paper, we explore the learning of a model to estimate the shape of a wire based solely on measurements of F/T states and without any visual perception. Visual perception is only used for off-line data collection. We propose to train a Supervised Autoencoder with convolutional layers that reconstructs the spatial shape of the wire while enforcing the latent space to resemble the space of F/T. Then, the encoder operates as a descriptor of the wire where F/T states can be mapped to its shape. On the other hand, the decoder of the model is the inverse problem where a desired goal shape can be mapped to the required F/T state. With the same collected data, we also learn the mapping from F/T states to grippers poses. Then, a motion planner can plan a path within the F/T space to a goal while avoiding obstacles. We validate the proposed data-based approach on Nitinol and standard electrical wires, and demonstrate the ability to accurately estimate their shapes.

Feasibility evaluation and pre-conceptual design of the Iter Tokamak systems monitor

Several thousands of sensors will be fitted in the main Tokamak Systems to oversee the behavior of the ITER machine. This engineering instrumentation will measure key parameters related to availability, prediction of failures, as well as schedule and needs related to their maintenance. However, despite the large amount of instrumentation envisaged, the Tokamak operator needs more data than those solely based on direct measurements, as it is not feasible to install sensors in every single location expected to be critical during the different operational modes.The “Tokamak Systems Monitor (TSM)” is a software suite specifically designed aiming at overcoming these limitations. By making use of the available measurements, it will provide an integrated view of the overall hydraulic, thermal, structural and electromagnetic response of the main Tokamak Systems (Vacuum Vessel, Magnets, Blankets, Divertor, Cryostat, Thermal Shields and the Diagnostics Port Plugs). Near real time processing shall allow an online evaluation of the Tokamak Systems status with only a few seconds latency. Offline data assessment shall expand the online calculation capabilities for in-depth detailed analysis after the plasma discharges. Sensor measurements can be used directly (e.g., the calculation of electromagnetic loads from measured currents and magnetic fields) or indirectly, through the resolution of ill-posed, inverse problems. The operator will receive all this information within the control room for managing scarce resources, operational limits, and decision-making. Eventually, the TSM will contribute to optimize the trade-off between maximal performance and equipment lifetime.This paper attempts to summarize the initial feasibility and concept design efforts towards the development, validation and practical implementation of reconstruction and Tokamak simulation algorithms.

The Effect of Offline Medical Resource Distribution on Online Physician-Patient Interaction: Empirical Study With Online and Offline Data

The relationship between online health communities (OHCs) and offline medical care is unclear because both provide physician-patient interaction services and channels. Taking advantage of information and communication technology, patients have been using OHCs widely. However, some physical medical resources (such as hospital beds and medical devices) cannot be replicated by information and communication technologies. Therefore, it is worth studying how offline medical resources affect physician-patient interactions in OHCs and how OHCs help to solve resource scarcity and the uneven distribution of traditional medical treatment. This study aimed to support the notion that physician-patient consultations in OHCs are influenced by the objective distribution of offline health care capital (accessibility and availability) and to provide suggestions for the allocation of medical resources in practice through the judicious use of offline and online channels. The empirical data in this study were collected from both online and offline channels. The offline data include 9 years (2006-2014) of medical resource statistics of 31 provincial administrative regions in mainland China. Moreover, data regarding the geolocation-based physician-patient interaction network in the OHC were also collected. The online data come from one of China's largest OHCs. We obtained 92,492 telephone consultation records of 6006 physicians using an automatic web crawler program. Social network analysis was used to visualize the descriptive statistics of the offline geolocation-based physician-patient interaction network in the OHC. A regression model with a squared variable was applied to analyze online and offline empirical data to further test our hypothesis. Two types of robustness tests were used to increase the reliability of the test results of the initial model. The results of our social network analysis show that there is a uniform geographic distribution of patients who use OHCs, whereas the physician relies more on geographic advantage (eg, a higher medical resource capability). Moreover, the empirical results of the regression model support the notion that physician-patient telephone consultations are positively influenced by physicians' online contributions (βcontribution=.210; P<.001) and capital availability (βbed=.935; P=.07), and, interestingly, spatial accessibility has an inverted U-shaped effect (βdistance=.199; P<.001 and βdistance2=-.00449; P=.008). The results indicate that the use of OHCs, although constrained by offline medical resources, provides a channel for offline resources to flow from areas with high availability to those with low availability. This study explores the relationship between online and offline channels by investigating online physician-patient interactions and offline medical resources. In particular, this study analyzes the impact of offline channels on online channels and verifies the possibility of OHC capital use shifting from a high-availability area to a low-availability area. In addition, it provides a theoretical and practical basis for understanding the interaction of online and offline channels of medical care.

Controllability Maximization of Network Systems: Gradient Computation Based on Offline Data

In the field of network systems, controllability maximization has become more important in terms of efficient control. When an exact model of a network system is not available, data-driven approaches are useful. In this paper, we establish a framework for maximizing the controllability of networked systems by using off-line data. In particular, the maximization with respect to the network topology of the network system is addressed. First, we develop a data-driven method for solving the Lyapunov equation which describes the properties of a system different from the system associated with the data. Second, based on this result, we derive a data-driven method for computing the gradient of a controllability measure (the trace of the controllability Gramian) with respect to the network topology of the network system. Finally, we show that our gradient computation can be used for controllability maximization based on off-line data. The effectiveness of the data-driven methods is numerically demonstrated.

Long Short-Term Memory-Based Neural Networks for Missile Maneuvers Trajectories Prediction⋆

Due to its extensive applications in different contexts, moving target tracking has become a hot topic in the last years, above all in the military field. Specifically, missile tracking research received a great effort, mainly for its importance in terms of security and safety. Herein, traditional solutions, e.g. Interacting Multiple Model (IMM) based on the Kalman estimation theory, achieve good performance under the main restrictive assumption of the a priori knowledge of the target model, so neglecting the unavoidable presence of model uncertainties and limiting the achievable tracking accuracy only by the presence of the measurement noise. With the specific aim of overcoming this narrowness, this work investigates the capability of deep neural networks in predicting the missile maneuvering trajectories in a model-free fashion. The idea is to leverage the Long-Short Term Memory (LSTM) net due to its excellent capability in learning long-term dependencies of temporal information. Two different LSTM-based architectures have been hence designed to predict both position and velocity of a missile using raw and noisy measurements provided by a realistic radar system, exploiting a large database abundant of realistic off-line data. Training results and theoretical derivations are verified through non-trivial scenarios in order to assess the capability of predicting unknown and realistic 3D missile maneuvers. Finally, the proposed approach has been also compared with a performing model-based IMM algorithm, suitably tuned to deal with realistic missile maneuvers, confirming the excellent generalization abilities of the developed data-driven architectures for different datasets.