Adaptive Prediction Model Research Articles

Optimizing Service Stipulation Uncertainty with Deep Reinforcement Learning for Internet Vehicle Systems

Fog computing brings computational services near the network edge to meet the latency constraints of cyber-physical System (CPS) applications. Edge devices enable limited computational capacity and energy availability that hamper end user performance. We designed a novel performance measurement index to gauge a device’s resource capacity. This examination addresses the offloading mechanism issues, where the end user (EU) offloads a part of its workload to a nearby edge server (ES). Sometimes, the ES further offloads the workload to another ES or cloud server to achieve reliable performance because of limited resources (such as storage and computation). The manuscript aims to reduce the service offloading rate by selecting a potential device or server to accomplish a low average latency and service completion time to meet the deadline constraints of sub-divided services. In this regard, an adaptive online status predictive model design is significant for prognosticating the asset requirement of arrived services to make float decisions. Consequently, the development of a reinforcement learning-based flexible x-scheduling (RFXS) approach resolves the service offloading issues, where x = service/resource for producing the low latency and high performance of the network. Our approach to the theoretical bound and computational complexity is derived by formulating the system efficiency. A quadratic restraint mechanism is employed to formulate the service optimization issue according to a set of measurements, as well as the behavioural association rate and adulation factor. Our system managed an average 0.89% of the service offloading rate, with 39 of delay over complex scenarios (using three servers with a 50% service arrival rate). The simulation outcomes confirm that the proposed scheme attained a low offloading uncertainty, and is suitable for simulating heterogeneous CPS frameworks.

Semi-Supervised Domain-Adaptive Seizure Prediction via Feature Alignment and Consistency Regularization

The inter-patient variability still poses a great challenge for the real-world application of EEG-based seizure prediction, where most previous methods could only work under the patient-specific fashion and fail to generalize across patients. To address this issue, some latest studies applied supervised domain adaptation (SDA), utilizing limited labeled data from the target patient to calibrate the model. However, as the epileptic EEG representation even varies within one single patient, limited target data could hardly cover the data distribution, thus the model still shows poor generalization on the target patient. To this end, we introduce a novel semi-supervised domain adaptive seizure prediction model (SSDA-SPM), using limited labeled target data and extra unlabeled target data for adaptation. SSDA-SPM mainly consists of two unsupervised modules, namely feature alignment (FA) module and consistency regularization (CR) module. The FA module aims to transfer knowledge from existing patients to the target patient coarsely by globally aligning the data distribution between them. Then the CR module further enhances the discriminability on the target patient by pushing the decision boundary into the low-density area. Our proposed method achieves 88.8% sensitivity, 0.182/h false prediction rate (FPR) and 0.849 AUC on the CHB-MIT database and 75.7% sensitivity, 0.165/h FPR and 0.763 AUC on the Kaggle database. Experimental results demonstrate that our method has provided a promising solution to improve the cross-patient generalization for seizure prediction.

Time series regression based on Bayesian model averaging and principal component analysis

This paper proposed an adaptive prediction model for high-dimensional time series data based on model averaging method and principal component analysis. Specifically, this paper considers the case where the response variable is a scalar and the predictor variable is a time series. Firstly, the high-dimensional time series data is extracted information by principal component analysis. Secondly, the Bayesian model averaging method is used to perform the forecast task based on the principal component projection matrix. The proposed method can effectively deal with the unsupervised nature of PCA and avoid the problem of selecting the number of PCA. It is demonstrated that the proposed method is competitive compared with the lasso regression and the ridge regression by real data analyses.

Adaptive prediction of turbine profile loss and multi-objective optimization in a wide incidence range

The loss prediction model plays a crucial role in turbine design for fast performance prediction and a shorter design cycle. Owing to the design requirements of high-efficiency turbines under a wide range of operating conditions, the loss prediction of the off-design incidence is increasingly important. However, limited by the modeling database and traditional modeling methods, the accuracy and adaptability of the existing off-design incidence loss predictions are insufficient. This paper proposes an adaptive prediction method based on machine learning and develops a multi-objective optimization process based on adaptive prediction. Machine learning (neural network) is applied for more flexible and accurate loss predictions over a wide incidence range. Compared with two classic loss models (Ainley and Mathieson model and Benner model), the adaptive prediction model significantly improves the ability to predict turbine profile loss with off-design incidence, particularly under large incidence conditions. The prediction root mean square error can be reduced by up to 73.8% (absolute value: 0.063). Furthermore, the multi-objective optimization method based on adaptive prediction is applied to the aerodynamic optimization of the original cascades with a wide incidence range. The weighted objective of the optimized cascade (Cri = 0.211) is reduced by 8.7% compared with that of the original cascade (Cri = 0.231). Within the range of full incidence angle (−40° to +20°), the variation of profile loss is reduced by 24.0%. This study is a preliminary exploration aimed at establishing an accurate turbine loss prediction system based on machine learning, the feasibility, and superiority of this approach are confirmed.

Marked shifts of harmful algal blooms in the Bohai Sea linked with combined impacts of environmental changes

The Bohai Sea, a semi-enclosed inland sea in China and an important mariculture region, has experienced extensive harmful algal blooms (HABs) and their negative impacts for several decades. To investigate the changes of HABs and their potential drivers over time and space, a dataset of 230 HAB events (1952–2017), along with corresponding environmental data (1990–2017) was compiled. The frequency of HAB events in the Bohai Sea has increased over time but plateaued in the last decade, and our analysis showed that history of HABs in the Bohai Sea could be categorized into three periods based on their frequency, scale, and HAB-forming species. The seasonal window of HAB events has started earlier and lasted longer, and the main hotspot has moved from Bohai Bay to coastal waters of Qinhuangdao over time. There were marked shifts in the representative HAB-forming microalgae, from dinoflagellates in the first period (before 2000) to haptophytes in the second period (2000–2009), and pelagophytes in the third period (2009 onwards). These community changes are accompanied by a trend toward diversification of HAB-forming microalgae, decrease in cell-size, and increase in negative impacts. Statistical analyses indicate that long-term changes in HABs in the Bohai Sea are linked with the combined effects of climate change, eutrophication and mariculture development. The results of the present study require to refine future monitoring programs, develop adaptive management strategies and predictive models for HABs in the Bohai Sea.

Adaptive Linear Predictive Model of an Improved Predictive Control of Permanent Magnet Synchronous Motor Over Different Speed Regions

Model predictive control (MPC) strategy can provide significant benefits for controlling nonlinear systems over classical cascade field-oriented control (FOC). However, the MPC is still in the development stage for a high-performance predictive model. Therefore, the proposed MPC in this article updates the internal linear predictive model at each time step to accurately represent the nonlinear plant of a permanent-magnet synchronous motor (PMSM) plant over different speed regions. In other words, the adaptive discrete linear plant model (ADLPM) is designed to update the current operating conditions of the machine parameters and the equilibrium points of the measured stator currents, speed, and load torque. For the operation in the flux-weakening region, the proposed MPC depends on a performance control algorithm (PCA) to obtain high dynamic performance. In this PCA algorithm, the proposed MPC depends on the modified reference velocity rather than the original reference velocity, which can calculate the required <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">d</i> -axis cur <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">r</u> ent directly. Moreover, the proposed cost function is designed directly in terms of the error values of the velocity and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">d</i> -axis current, which fits the motor performance based on the further constraint of the maximum magnitude of the drawn stator current provided to control the acceleration of the rotor. Finally, comprehensive simulations and experiments have fully demonstrated that the proposed MPC can reduce the speed drop, and torque ripple in response to those of the FOC and traditional MPC strategies.

A novel structure adaptive new information priority discrete grey prediction model and its application in renewable energy generation forecasting

• A novel accumulation generation operator is developed by weighting coefficient. • A novel structure adaptive new information priority grey model is proposed. • The robustness analysis of the new model is established. • Comparative study shows that new model is superior to other various grey models. Renewable energy has made a significant contribution to global power generation. Therefore, accurate mid-to-long term renewable energy generation forecasting is becoming more and more important for integrating renewable energy systems with smart grid and energy strategic planning. For this purpose, a novel structure adaptive new information priority discrete grey prediction model is established, and the disturbance analysis shows that it is suitable for small sample modeling. Firstly, the fractional dynamic weighted coefficient is introduced to define the accumulative generating operator satisfying the new information priority principle, which can realize the effective utilization of system information with insufficient information and accurately extract the development mode of system sequence from sparse samples. In terms of model structure, the nonlinear term and periodic fluctuation term are introduced to simulate the nonlinear and periodic fluctuation trend of renewable energy generation data, which improves the adaptability of the grey prediction model to the nonlinear and fluctuating time series. By designing the comparative experiment of optimization algorithm, Grey Wolf Optimizer (GWO) is selected to optimize the structural parameters of the model, giving the proposed model higher flexibility and stronger adaptability. In order to illustrate the performance of the model, the new model is compared with statistical econometrics technology, artificial neural network (ANN) and various existing grey models in three cases: World biofuel power generation, China's renewable energy power generation and small-scale solar photovoltaic power generation in the United States. The experimental results show that the proposed model is superior to other competitive models in terms of fitting accuracy and prediction accuracy. In addition, Monte-Carlo Simulation and probability density analysis further show that the proposed model can provide the best prediction effect and has high robustness.

Radiotherapy-Immunotherapy Related Pneumonitis Prediction Based on an Adaptive Fusion Model of Deep CT and Clinical Parameters

<h3>Purpose/Objective(s)</h3> It is critical yet challenging to select clinical signatures from a wide variety of non-imaging clinical parameters and image features. This study aims to develop a unified and learnable CT and clinical parameter fusion model for symptomatic pneumonia prediction (grade ≥ 2) in non-small cell lung cancer (NSCLC) patients who received radiotherapy combined with immunotherapy. <h3>Materials/Methods</h3> This retrospective study contains 243 NSCLC patients (62 symptomatic pneumonitis) who underwent radiotherapy combined with immunotherapy between January 2015 and June 2021. Patient demographics, tumor histologic type, TNM stage, therapeutic regimen, inflammatory biomarkers and pre-treatment CT were collected. We perform five-fold cross-validation, where each fold has 195 training cases (50 symptomatic pneumonitis) and 48 validation cases (12 symptomatic pneumonitis). Pre-processings include performing one-hot encoding over categorical parameters and normalizing numerical features. The adaptive fusion and prediction model consists of a) 3D CNN encoders for deep feature extraction from tumor and lung volumes, b) multilayer perceptron (MLP) for deep signature extraction from categorical and numerical clinical parameters, and c) graph attention layers (GAT) to learn adaptive weights of different types of features for embedding, and d) softmax for final classification. <h3>Results</h3> Our prediction model achieved area under ROC (auc), sensitivity (sen), and specificity (spe) of 0.841, 0.784, and 0.832, which improved the performance of other comparing methods including our model using CT alone (auc 0.823, sen 0.767, and spe 0.810), our model without GAT (auc 0.812, sen 0.758, and spe 0.786), and conventional radiomics model based on logistic regression (auc 0.782, sen 0.763, spec 0.784). The improvement was statistically significant (p<0.001). <h3>Conclusion</h3> Our unified model improved pneumonia prediction results using adaptively weighted clinical parameters and CT, which can support effective safety management and individualized treatment for NSCLC patients.

An Adaptive Long-Term Bus Arrival Time Prediction Model with Cyclic Variations

Real-time bus arrival information systems at transit stops can be useful to passengers for efficient trip planning and reducing waiting times. The accuracy of such systems depends upon the ability of the model to account for variations in the data series and to adjust according to changing traffic conditions. Many of the existing studies on passenger information systems have modeled the system based on stationary relations, not taking into account the cyclic variations in data, which is often suitable for demonstration purposes but not for long-term implementation. The present study models the changing relationships using Double-Seasonal Holt-Winter’s Exponential Smoothing approach, which allows for self-updating parameters at four levels. It accounts for both long-term and short-term seasonal fluctuations in data while maintaining the dynamic treatment of real-time bus information. The model also takes into account delays using the real-time running information of the bus and incorporates it into subsequent forecasts for better accuracy. Real-time data from GPS transmitters in buses were used for validation of the proposed model. The results show that the proposed model performs better than the currently-used elementary field methods and is able to forecast bus travel times with a reasonable accuracy.

Tracking of Uncertain Robotic Manipulators Using Event-Triggered Model Predictive Control With Learning Terminal Cost

This paper presents an event-triggered model predictive control (MPC) strategy with learning terminal cost for robotic manipulators containing model uncertainty and input constraints. In the proposed MPC structure, an adaptive predictive model for the robotic system is established by radial basis function neural networks (RBFNNs) firstly. Then, a terminal cost adjusted by the global learning mechanism is constructed. Both global steady-state optimization and transient fast convergence are achieved by adding the learning terminal cost to the MPC scheme. After that, a triggering condition of the MPC solving is developed based on the predictive model’s weights and the predictive tracking error. Besides, the condition to avoid Zeno behavior is obtained. The recursive feasibility of the proposed MPC strategy is verified, and the ultimately uniformly boundedness (UUB) of all variables is proved according to the Lyapunov theorem. Finally, experiments based on an xMate7 Pro robot are conducted to demonstrate the effectiveness of the presented method. Note to Practitioners—The tracking control of robotic manipulators is a common and important problem in industrial applications, such as grasping, loading, unloading, et al. There exist some limitations in existing control methods. For example, general control strategies such as adaptive control, sliding mode control neglect the balance between control costs and expected performance; existing optimal control approaches rarely consider global steady-state optimization and transient fast convergence under the influence of model uncertainty. This paper is motivated by these limitations of robotic control design and inspired by model predictive control and optimal control theory. It develops a novel strategy for tracking control of robotic manipulators, which includes three following items: (1) adopt an approximation model constructed by neural networks as the predictive model for estimating the robotic system; (2) introduce a critic network into the terminal cost of the objective function of MPC for learning global optimal solution; (3) establish an event-triggered mechanism for solving of the optimization problem. The proposed method is verified by experimental results. In our future work, the proposed control strategy will be applied to more industrial processes.

Energy Management Strategy and Optimal Sizing for Hybrid Energy Storage Systems Using an Evolutionary Algorithm

Energy management strategy (EMS) of hybrid energy storage systems has an essential mission of ensuring safety, enhancing reliability and improving system efficiency. This paper focuses on optimizing sizing of HESS and parameters of EMS simultaneously. Firstly, an improved model is employed in adaptive predictive model control (AMPC). Secondly, in order to minimize the cost of supercapacitors and the capacity degradation of batteries at the same time, the multiple objective optimization problems are solved by Multi-objective Evolutionary Algorithm based on Decomposition (MOEA/D). The nine variables are decided including seven parameters of AMPC and series-parallel number of supercapacitors. Subsequently, the support vector machine combined with the gray wolf optimization algorithm (GWO-SVM) is employed to enhance the applicability of EMS. Finally, in order to prove the superiority of the proposed EMS, a combined cycle and a real cycle are performed for validation. The results illustrate that the proposed method can improve the efficiency of the system and enhance the life span of the battery. Comparing with AMPC-only, the total energy loss can be reduced by at least 17.9%. A reduction of not less than 8.8% in RMS of battery current is obtained. The Ah-throughput can be reduced by at least 2.5%. In further discussions, the superiority of MOEA/D and GWO-SVM is proved by comparing with other methods.

Identifying Exposure of Urban Area to Certain Seismic Hazard Using Machine Learning and GIS: A Case Study of Greater Cairo

The 1992 Cairo earthquake, with a moment magnitude of 5.8, is the most catastrophic earthquake to shock the Greater Cairo (GC) in recent decades. According to the very limited number of seismological stations at that time, the peak ground acceleration (PGA) caused by this event was not recorded. PGA calculation requires identification of nature of the earthquake source, the geologic setting of the path between the source and site under investigation and soil dynamic properties of the site. Soil dynamic properties are acquired by geotechnical investigations and/or geophysical survey. These two methods are costly and are not applicable in regional studies. This study presents an adaptive and reliable PGA prediction model using machine learning (ML) along with six standard geographic information system (GIS) interpolation methods (IDW, Kriging, Natural, Spline, TopoToR, and Trend) to predict the spatial distribution of PGA caused by this event over the GC. The model is employed to estimate the exposure of the urban area and population in the GC based on the available geotechnical and geophysical investigations. The exposure (population) data is from free and easy-access sources, e.g., Landsat images and the Global Human Settlement Population Grid (GHS-POP). The results show that Natural, Spline, and Trend are not suitable GIS interpolation techniques for generating seismic hazard maps (SHMs), while the Kriging Method shows sufficient prediction. Interestingly, with an accuracy of 96%, the ML model outperforms the classical GIS methodologies.

Adaptive Residual Life Prediction for Small Samples of Mechanical Products Based on Feature Matching Preprocessor-LSTM

In order to solve the problem of predicting the residual life of mechanical products accurately based on small-sample data, this paper proposes a small-sample adaptive residual life prediction model of mechanical products based on feature matching preprocessor-LSTM. First, aiming at the problem of low accuracy of remaining life prediction for small samples of mechanical products caused by multiple time scales and multiple fault states, the failure time data and performance degradation data are fused, and the failure rate and standard deviation are used as the remaining life prediction criteria to intuitively reflect The possibility of failure of a component or system at a certain point in time. Considering the demand of adaptive small-sample residual life prediction data, this paper establishes the adaptive matching pre-processor model of life characteristics. On this basis, the LSTM neural network is used to establish a small-sample adaptive residual life prediction model. Then, the XJTU-SY bearing life data set and the test data of the small-sample life characteristics measured by the RV reducer are used as the research objects, and a small amount of the data set is randomly selected. The remaining life expectancy is predicted from the sample data and compared with its standard remaining life, respectively. The comparison results show that the overall prediction error is small. This study shows that the remaining life prediction model established can better predict the remaining life of mechanical product sub-sample data and provides a feasible method for predicting the remaining life of mechanical product sub-samples.

Forecasting China's energy production and consumption based on a novel structural adaptive Caputo fractional grey prediction model

China's total energy consumption and production ranks first in the world. However, China's energy structure is not perfect. Therefore, accurate prediction of future energy trends is of great significance for the Chinese government to adjust the energy structure. Therefore, this paper proposes a novel structural adaptive grey model FCSAGM(p, 1) with Caputo fractional derivative and a new Caputo fractional order accumulation generation operator. Through the comparison between optimization algorithms, Particle Swarm Optimization (PSO) is selected to optimize the parameters of the model. In particular, aiming at the problem of model reliability caused by the optimization algorithm, the robustness of the model is analyzed, and the results show that the model is stable and reliable. Finally, four actual cases of China's total energy consumption, China's total primary energy production, China's thermal power generation and China's hydropower generation are predicted; the prediction result shows that the new model has higher prediction accuracy than the other six grey models.

Dual‐stage time series analysis on multifeature adaptive frequency domain modeling

Time series research in academic and industrial fields has attracted wide attention. However, the frequency information contained in time series still lacks effective modeling. The studies found that time series forecasting relies on different frequency patterns: short-term series forecasting relies more on high-frequency components, while long-term forecasting focuses more on low-frequency data. To better describe the multifrequency mode, a dual-stage multifeature adaptive frequency domain prediction model (DMAFD) is proposed in this paper. DMAFD contains two stages. First, it adopts the XGBoost algorithm to obtain a feature vector by analyzing the feature importance. Second, the frequency feature extraction of time series and the frequency aware modeling of the target sequence is integrated, for building an end-to-end prediction network based on the dependence of time series on frequency mode. The innovation is reflected in the fact that the prediction network can automatically focus on multifrequency components according to the dynamic evolution of the input sequence. Extensive experiments on four real data sets from different fields show that DMAFD obtains higher accuracy and smaller lags in time step analysis compared with state-of-the-art algorithms.

An adaptive Grey-Markov model based on parameters Self-optimization with application to passenger flow volume prediction

It has been demonstrated that local prediction approaches show better prediction performance compared with global ones. The paper proposes a novel local prediction model named as the adaptive grey-Markov prediction model for incomplete and complex dynamic systems. This model displays favorable adaptability, flexibility and universality because of four advantages. Firstly, because of the randomness and non-repeatability of time series, the paper proposes adaptive exponential accumulated generating operators to describe different contribution degrees of historical information to future change trends of system characteristics. Secondly, this paper proposes a new paradigm of adaptive grey background value, which can automatically adjust parameters based on historical information to optimize this model and reduce prediction errors. Thirdly, the optimization model based on average absolute percentage error is constructed, and the corresponding adaptive parameters are searched by Ant Lion Optimizer algorithm. Fourthly, this model consists of a quadratic polynomial function describing nonlinearity and a periodic function generated by a Fourier series characterizing noise and period. Thereupon, the adaptive odd-period grey model is proposed. Its residual modified model, named as adaptive grey-Markov modified model, is given based on Markov chain (MC) and virtual linguistic trust degree (VLTD) to further improve prediction accuracy. The modified values of absolute errors are determined by the score functions produced from trust degrees of absolute errors to states on MC calculated by VLTD method. This modified way retains the influence degrees of absolute errors on the modified values and improves the prediction accuracy. Finally, the paper further demonstrates performance and practicability of the proposed prediction model through predicting passenger flow of Chengdu Metro Line1 and comparative analysis.

An intelligent multi-step predictive control method of the Small Modular Reactor

As Small Modular Reactors (SMRs) feature great complexity and uncertainty in response and are varying over time during long-term operation, it is of difficulty to design a well-performed controller that can deal with different working conditions. In this paper, an intelligent multi-step predictive scheme is created to support the development of autonomous control for SMRs. In this scheme, an adaptive prediction model is constructed using T-S fuzzy identification with training data generated by cases of reactor system model, and a real-time adaptive algorithm is utilized to update model parameters to improve accuracy of prediction. Particle swarm optimization (PSO) has also been carried out to search for the optimal decision. The simulation results manifest the effectiveness and the good performance of the proposed approach.

Adaptive Multi-Factor Quantitative Analysis and Prediction Models: Vaccination, Virus Mutation and Social Isolation on COVID-19.

Different countries have adopted various control measures for the COVID-19 pandemic in different periods, and as the virus continues to mutate, the progression of the pandemic and preventive measures adopted have varied dynamically over time. Thus, quantitative analysis of the dynamic impact of different factors such as vaccination, mutant virus, social isolation, etc., on transmission and predicting pandemic progress has become a difficult task. To overcome the challenges above and enable governments to formulate reasonable countermeasures against the ongoing COVID-19 pandemic, we integrate several mathematical methods and propose a new adaptive multifactorial and geographically diverse epidemiological model based on a modified version of the classical susceptible-exposed-infectious-recovered (SEIR) model. Based on public datasets, a multi-center study was carried out considering 21 regions. First, a retrospective study was conducted to predict the number of infections over the next 30 days in 13 representative pandemic areas worldwide with an accuracy of 87.53%, confirming the robustness of the proposed model. Second, the impact of three scenarios on COVID-19 was quantified based on the scalability of the model: two different vaccination regimens were analyzed, and it was found that the number of infections would progressively decrease over time after vaccination; variant virus caused a 301.55% increase in infections in the United Kingdom; and 3-tier social lockdown in the United Kingdom reduced the infections by 47.01%. Third, we made short-term prospective predictions for the next 15 and 30 days for six countries with severe COVID-19 transmission and the predicted trend is accurate. This study is expected to inform public health responses. Code and data are publicly available at https://github.com/yuanyuanpei7/covid-19.

A novel adaptive discrete grey prediction model for forecasting development in energy consumption structure—from the perspective of compositional data

PurposeThe purpose of this study is to make a prediction of China's energy consumption structure from the perspective of compositional data and construct a novel grey model for forecasting compositional data.Design/methodology/approachDue to the existing grey prediction model based on compositional data cannot effectively excavate the evolution law of correlation dimension sequence of compositional data. Thus, the adaptive discrete grey prediction model with innovation term based on compositional data is proposed to forecast the integral structure of China's energy consumption. The prediction results from the new model are then compared with three existing approaches and the comparison results indicate that the proposed model generally outperforms existing methods. A further prediction of China's energy consumption structure is conducted into a future horizon from 2021 to 2035 by using the model.FindingsChina's energy structure will change significantly in the medium and long term and China's energy consumption structure can reach the long-term goal. Besides, the proposed model can better mine and predict the development trend of single time series after the transformation of compositional data.Originality/valueThe paper considers the dynamic change of grey action quantity, the characteristics of compositional data and the impact of new information about the system itself on the current system development trend and proposes a novel adaptive discrete grey prediction model with innovation term based on compositional data, which fills the gap in previous studies.

An adaptive prediction model for sparse data forecasting

Sparse data generated by the limitations of data acquisition are ubiquitous for prediction. However, the general prediction model is challenging to deal with those sparse data. Therefore, this paper aims to propose an adaptive sparse data prediction model. Firstly, we introduced the aXreme Gradient Boosting (XGBoost) algorithm to build an adaptive prediction model to correct sparse data constantly. Secondly, the sparsity perception of the XGBoost algorithm is used for parallel tree learning. Finally, we applied the model to the PM2.5 concentration forecasting of Nanjing, China. We trained the model and adjusted the parameters to get better prediction results, and compared the prediction results with actual data to prove the feasibility of the model.