Perform Error Analysis Research Articles

Deep Learning-Driven Workload Prediction and Optimization for Load Balancing in Cloud Computing Environment

Abstract Cloud computing revolutionizes as a technology that succeeds in serving large-scale user demands. Workload prediction and scheduling tend to be factors dictating cloud performance. Forecasting the future workload in due to avoid unfair resource allocation, emerges to be a crucial inspecting feature for enhanced performance. The aforementioned issues of interest are addressed in our work by soliciting a Deep Learning driven Max-out prediction model, which efficiently forecasts the future workload by providing a balanced approach for enhanced scheduling with the Tasmanian Devil-Bald Eagle Search (TDBES) optimization algorithm. The results obtained proved that the TDBES scored efficacy in makespan with 16.75%, migration cost with 14.78%, and a migration efficiency rate of 9.36% over other existing techniques like DBOA, WACO, and MPSO, with additional error analysis of prediction performance using RMSE, MAP, and MAE, among which our contributed approach overrides traditional methods with least error.

Detecting Algorithmic Errors and Patient Harms for AI-Enabled Medical Devices in Randomized Controlled Trials: Protocol for a Systematic Review.

Artificial intelligence (AI) medical devices have the potential to transform existing clinical workflows and ultimately improve patient outcomes. AI medical devices have shown potential for a range of clinical tasks such as diagnostics, prognostics, and therapeutic decision-making such as drug dosing. There is, however, an urgent need to ensure that these technologies remain safe for all populations. Recent literature demonstrates the need for rigorous performance error analysis to identify issues such as algorithmic encoding of spurious correlations (eg, protected characteristics) or specific failure modes that may lead to patient harm. Guidelines for reporting on studies that evaluate AI medical devices require the mention of performance error analysis; however, there is still a lack of understanding around how performance errors should be analyzed in clinical studies, and what harms authors should aim to detect and report. This systematic review will assess the frequency and severity of AI errors and adverse events (AEs) in randomized controlled trials (RCTs) investigating AI medical devices as interventions in clinical settings. The review will also explore how performance errors are analyzed including whether the analysis includes the investigation of subgroup-level outcomes. This systematic review will identify and select RCTs assessing AI medical devices. Search strategies will be deployed in MEDLINE (Ovid), Embase (Ovid), Cochrane CENTRAL, and clinical trial registries to identify relevant papers. RCTs identified in bibliographic databases will be cross-referenced with clinical trial registries. The primary outcomes of interest are the frequency and severity of AI errors, patient harms, and reported AEs. Quality assessment of RCTs will be based on version 2 of the Cochrane risk-of-bias tool (RoB2). Data analysis will include a comparison of error rates and patient harms between study arms, and a meta-analysis of the rates of patient harm in control versus intervention arms will be conducted if appropriate. The project was registered on PROSPERO in February 2023. Preliminary searches have been completed and the search strategy has been designed in consultation with an information specialist and methodologist. Title and abstract screening started in September 2023. Full-text screening is ongoing and data collection and analysis began in April 2024. Evaluations of AI medical devices have shown promising results; however, reporting of studies has been variable. Detection, analysis, and reporting of performance errors and patient harms is vital to robustly assess the safety of AI medical devices in RCTs. Scoping searches have illustrated that the reporting of harms is variable, often with no mention of AEs. The findings of this systematic review will identify the frequency and severity of AI performance errors and patient harms and generate insights into how errors should be analyzed to account for both overall and subgroup performance. PROSPERO CRD42023387747; https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=387747. PRR1-10.2196/51614.

Polarization of Compton-Scattered Prompt Gamma-Ray Technique for HEU Detection at 186 keV

Prompt gamma-ray polarization is a practical method for detecting highly enriched uranium (HEU) in concealed sources. It also provides information on their geometry, magnetic fields, and radiation mechanisms. However, prompt gamma-ray polarization measurements have rarely been applied in nuclear nonproliferation areas to detect HEU. In this study, the feasibility of detecting the characteristic energy peak of 186 keV, which is associated with the asymmetry of the activation mechanism and the detection of energy-dependent polarization from concealed HEU sources, was evaluated using the Compton scattering approach. A Monte Carlo N-particle transport code simulation was used to realize the activation mechanism of HEU via two 1.4-mm strips of converter material [i.e., cesium lead tribromide (CsPbBr3)], transported by secondary scattered gamma rays during the three-stage process of Compton scattering, polarization, and detection. This paper presents the mathematical model, the physics of Compton scattering, and the polarization mechanism for the detection technique. In this case, the physics is relevant to both processes in which the emitted secondary scattered gamma rays undergo initial orthogonal polarization. Specifically, to meet the objective of testing the technical protocol for the enhanced detection of energy peaks associated with HEU, particularly 186 keV, simulations were conducted to quantify the HEU volume and neutron source strength in the MCNP data card to perform error analysis. The detector system had the potential to acquire good resolved photopeak with a 4.5% relative error or less, with a 1 Ci source activity, and a peak-to-background ratio of 1.15. This resolution took 163 s for high-purity germanium detection, which is comparable to current methods used for material detection placed within 100 to 900 s to completion. The small error difference was due to the attributes of the phenomenal enhancement properties of cesium tribromide and polarimetry. The identified photo peaks included K-shell X-rays from 235U, 61 keV from fission, 511-keV annihilation, and the peak of interest at 186 keV. The result from spectral analyses showed clear signatures related to pure and adulterated HEU. HEU detection with the low neutron yield and the easiness of shielding the yield of the HEU sample showed that the HEU characterization was feasible when shielded, with the highest success rate under both enhancement approaches. The optimization and scale-up of this technique are expected to enhance its applications in a large-scale HEU detection design.

Investigation of optimized machine learning models with PSO for forecasting the shear capacity of steel fiber-reinforced SCC beams with/out stirrups

This article presents a comprehensive investigation of the applicability of optimized machine learning (ML) models with particle swarm optimization (PSO) for forecasting the shear strength of steel fiber-reinforced self-compacting concrete (SFR-SCC) beams with/without stirrups in engineering applications. Firstly, a database containing the results of 101 specimens with nine input features is adopted to train the models. As ML models such as random forest (RF), adaptive boosting regression (AdaBoost), extreme gradient boosting (XGBoost), support vector regression (SVR), and K-nearest neighbors regression (KNN) are considered, whereas the hyper-parameters of these models are set as default by the sklearn module. On the other hand, PSO-ML models (PSO-RF, PSO-AdaBoost, PSO-XGBoost, PSO-SVR, and PSO–KNN) are constructed using particle swarm optimization to find the optimal combination of the hyper-parameters of these default ML models. Afterwards, the forecasting ability of each model is extensively assessed using various performance metrics, error analysis, and score analysis, and the model with the best forecasting ability is determined and compared with existing empirical models. Moreover, Shapley additive explanation (SHAP) analysis is also utilized to ensure the interpretability of the forecasting models and to overcome the “black box” problem of ML methods. Lastly, based on the best forecasting model developed in this study, a graphical user interface (GUI) has been developed to easily forecast the shear strength of SFR-SCC beams in practical applications. The results of the study clearly illustrate that PSO-ML models exhibit better forecasting capabilities than default models. It can be emphasized from here that the PSO algorithm can be an effective tool to improve the performance of ML models. It should also be pointed out that the use of PSO in simpler algorithms instead of tree-based models can further improve forecasting efficiency. On the other hand, the PSO-RF model has the best performance, with a lower error value and a high final score. And this makes it a more reliable option for predicting the shear strength of the SFR-SCC beams compared to empirical equations. In addition, according to the results of SHAP feature importance analysis, the most important input parameters affecting the shear strength of SFR-SCC beams are stirrup rebar ratio (ρv), stirrup yield strength (fyv) and longitudinal rebar ratio (ρt). This information can assist engineers in paying special attention to these features in their design and assessment processes.

Some finite difference methods for solving linear fractional KdV equation

The time-fractional Korteweg de Vries equation can be viewed as a generalization of the classical KdV equation. The KdV equations can be applied in modeling tsunami propagation, coastal wave dynamics, and oceanic wave interactions. In this study, we construct two standard finite difference methods using finite difference methods with conformable and Caputo approximations to solve a time-fractional Korteweg-de Vries (KdV) equation. These two methods are named as FDMCA and FDMCO. FDMCA utilizes Caputo's derivative and a finite-forward difference approach for discretization, while FDMCO employs conformable discretization. To study the stability, we use the Von Neumann Stability Analysis for some fractional parameter values. We perform error analysis usingL1&L∞norms and relative errors, and we present results through graphical representations and tables. Our obtained results demonstrate strong agreement between numerical and exact solutions when the fractional operator is close to 1.0 for both methods. Generally, this study enhances our comprehension of the capabilities and constraints of FDMCO and FDMCA when used to solve such types of partial differential equations laying some ground for further research.

Innovation and Practice of Educational Management System in Colleges and Universities Based on the Concept of Data Analysis Education

Abstract This paper proposes the evaluation method of college education management mode based on deep belief network (DBN), establishes the evaluation index of college education management mode, takes it as the input of the deep belief network, adopts the divergence algorithm and backpropagation algorithm to learn and optimize the parameters of the deep belief network, and establishes the evaluation model of the education management mode of universities. The evaluation method of the college education management model evaluation of the deep belief network is verified by evaluating the results and performing error analysis on 438 questionnaire test samples. The results show that the number of training steps of the adopted DBN model in the supervised tuning stage is 214, and the error fluctuation range is ±0.6, which achieves the expected effect, indicating that the adoption of the model has accuracy in the evaluation of college education management model. Under the concept of data analysis education, the evaluation method proposed in this paper has a positive significance for the innovation and practice of educational and teaching management systems.

IncRna: The R Package for Optimizing lncRNA Identification Processes.

In silico identification of long noncoding RNAs (lncRNAs) is a multistage process including filtering of transcripts according to their physical characteristics (e.g., length, exon-intron structure) and determination of the coding potential of the sequence. A common issue within this process is the choice of the most suitable method of coding potential analysis for the conducted research. Selection of tools on the sole basis of their single performance may not provide the most effective choice for a specific problem. To overcome these limitations, we developed the R library lncRna, which provides functions to easily carry out the entire lncRNA identification process. For example, the package prepares the data files for coding potential analysis to perform error analysis. Moreover, the package gives the opportunity to analyze the effectiveness of various combinations of the lncRNA prediction methods to select the optimal configuration of the entire process.

T5 for Hate Speech, Augmented Data, and Ensemble

We conduct relatively extensive investigations of automatic hate speech (HS) detection using different State-of-The-Art (SoTA) baselines across 11 subtasks spanning six different datasets. Our motivation is to determine which of the recent SoTA models is best for automatic hate speech detection and what advantage methods, such as data augmentation and ensemble, may have on the best model, if any. We carry out six cross-task investigations. We achieve new SoTA results on two subtasks—macro F1 scores of 91.73% and 53.21% for subtasks A and B of the HASOC 2020 dataset, surpassing previous SoTA scores of 51.52% and 26.52%, respectively. We achieve near-SoTA results on two others—macro F1 scores of 81.66% for subtask A of the OLID 2019 and 82.54% for subtask A of the HASOC 2021, in comparison to SoTA results of 82.9% and 83.05%, respectively. We perform error analysis and use two eXplainable Artificial Intelligence (XAI) algorithms (Integrated Gradient (IG) and SHapley Additive exPlanations (SHAP)) to reveal how two of the models (Bi-Directional Long Short-Term Memory Network (Bi-LSTM) and Text-to-Text-Transfer Transformer (T5)) make the predictions they do by using examples. Other contributions of this work are: (1) the introduction of a simple, novel mechanism for correcting Out-of-Class (OoC) predictions in T5, (2) a detailed description of the data augmentation methods, and (3) the revelation of the poor data annotations in the HASOC 2021 dataset by using several examples and XAI (buttressing the need for better quality control). We publicly release our model checkpoints and codes to foster transparency.

Dimension measurement method of ring forgings based on laser scanning and multiple regression filtering algorithm

In order to detect the radial is characterized and the isolated points that deviate from the target point cloud are removed using statistical section dimensions of ring parts in the forging process, we present a new method that processes laser scanning data of the outer radial section profile of ring forgings. Based on regression analysis and point cloud filtering, a multiple regression filtering algorithm is proposed to process the point cloud data for the measurement of the outer diameter of ring forgings. First, the laser scanned point cloud data filtering methods. The outlier clusters are then removed using regression analysis and processed by Gaussian regression filtering algorithms to obtain point cloud data with good circularity characteristics. Next, the Least Squares principle is used to fit a circle to the processed data to obtain information on the external diameter of the ring forgings to be measured. Finally, the feasibility of the method is verified by measuring the outer dimensions of the elliptical ring forgings in the experiment and performing error analysis.

"한국어-스페인어 기계 번역 성능 분석 연구 - 구글 번역기와 네이버 파파고를 중심으로"

This article performs performance evaluation and error analysis of the Korean-Spanish translation service engine. Unlike previous studies, the research is conducted by diversifying the types of texts to be studied and dividing them into texts with and without official translations. The translation services for study are Google Translate and Naver Papago. The translation engine performance evaluation uses BLEU score, which is one of the types of automatic evaluation, and as a result, both translation services showed relatively low evaluation scores. Also, in the case of error analysis, it is divided into four types: words, grammar, sentences, and typos. As a result of the error evaluation, word-related errors were seen the most, and the word substitution phenomenon was particularly prominent. In the case of word errors, it is characteristic that grammatical errors have a great influence on the meaning transfer of the translation, whereas there are often cases where it does not have a great influence on the meaning transfer of the translation. This study is significant in that it suggested the future direction of the development of Korean-Spanish machine translation through the evaluation of the Korean-Spanish translation service.

The Prediction of Enterprise Stock Change Trend by Deep Neural Network Model.

This study aims to accurately predict the changing trend of stocks in stock trading so that company investors can obtain higher returns. In building a financial forecasting model, historical data and learned parameters are used to predict future stock prices. Firstly, the relevant theories of stock forecasting are discussed, and problems in stock forecasting are raised. Secondly, the inadequacies of deep neural network (DNN) models are discussed. A prediction trend model of enterprise stock is established based on long short-term memory (LSTM). The uniqueness and innovation lie in using the stock returns of Bank of China securities in 2022 as the training data set. LSTM prediction models are used to perform error analysis on company data training. The 20-day change trend of the company's stock returns under different models is predicted and analyzed. The results show that as the number of iterations increases, the loss rate of the LSTM training curve keeps decreasing until 0. The average return price of the LSTM prediction model is 14.01. This figure is closest to the average real return price of 13.89. Through the forecast trend analysis under different models, LSTM predicts that the stock change trend of the enterprise model is closest to the changing trend of the actual earnings price. The prediction accuracy is better than other prediction models. In addition, this study explores the characteristics of high noise and complexity of corporate stock time series, designs a DNN prediction model, and verifies the feasibility of the LSTM model to predict corporate stock changes with high accuracy.

InteractEva: A Simulation-Based Evaluation Framework for Interactive AI Systems

Evaluating interactive AI (IAI) systems is a challenging task, as their output highly depends on the performed user actions. As a result, developers often depend on limited and mostly qualitative data derived from user testing to improve their systems. In this paper, we present InteractEva; a systematic evaluation framework for IAI systems. InteractEva employs (a) a user simulation backend to test the system against different use cases and user interactions at scale with (b) an interactive frontend allowing developers to perform important quantitative evaluation tasks, including acquiring a performance overview, performing error analysis, and conducting what-if studies. The framework has supported the evaluation and improvement of an industrial IAI text extraction system, results of which will be presented during our demonstration.

Data-driven modelling and optimization of hydrogen adsorption on carbon nanostructures

In the present study, using modelling based on experimental data, models for predicting the hydrogen adsorption isotherm were presented. The three Automatic Learning of Algebraic Models (ALAMO), feed-forward artificial neural networks (ANNs), and group method of data handling-type polynomial neural networks (GMDH-PNN) were constructed. The created models were evaluated to predict the equilibrium data of hydrogen storage on carbon nanostructures, including activated carbons doped with palladium (Pd) nanoparticles, fullerene pillared graphene nanocomposites, and nickel (Ni)-decorated carbon nanotubes. The inputs were nanostructure characteristics such as surface area, pore-volume, and thermodynamic conditions such as pressure. The generalization of the trained models was acceptable, and the models successfully predicted the hydrogen adsorption isotherm for new inputs. The relative error percentage for most data points is less than 4%, which demonstrates their applicability in determining adsorption isotherms for any operating conditions. By performing error analysis calculations, it was shown that the ALAMO model has the highest accuracy. Also, sensitivity analysis calculations show that pressure is the most influential parameter in the adsorption process. Besides, by performing Genetic Algorithm (GA) optimization using the ALAMO model, the amount of pressure and adsorbent properties were determined so that the amount of hydrogen adsorption is maximized. According to the optimization results based on the GA, the higher the pressure, the greater the amount of hydrogen adsorption. The nanotubes with a surface area of 194.15 m2/g, a total volume of 1.8 cm3/g, micropore volume of 0.097 cm3/g, and mesopore volume of 0.963 cm3/g, graphene with a surface area of 2977.13 m2/g, a total volume of 1.5134 cm3/g, density of 617.45 kg/m3, and activated carbon at pressures less than 30 bar with a surface of 2546.36 m2/g, a total volume of 1.237 cm3/g, micropore volume of 0.839 cm3/g, and activated carbon at pressures more than 30 bar with a surface of 3027 m2/g, a total volume of 1.343 cm3/g, a micropore volume of 0.9582 cm3/g, and a mesopore volume of 1.23 cm3/g, have the highest amount of stored hydrogen.

Machine Learning Approach to Facilitate Knowledge Synthesis at the Intersection of Liver Cancer, Epidemiology, and Health Disparities Research.

Liver cancer is a global challenge, and disparities exist across multiple domains and throughout the disease continuum. However, liver cancer's global epidemiology and etiology are shifting, and the literature is rapidly evolving, presenting a challenge to the synthesis of knowledge needed to identify areas of research needs and to develop research agendas focusing on disparities. Machine learning (ML) techniques can be used to semiautomate the literature review process and improve efficiency. In this study, we detail our approach and provide practical benchmarks for the development of a ML approach to classify literature and extract data at the intersection of three fields: liver cancer, health disparities, and epidemiology. We performed a six-phase process including: training (I), validating (II), confirming (III), and performing error analysis (IV) for a ML classifier. We then developed an extraction model (V) and applied it (VI) to the liver cancer literature identified through PubMed. We present precision, recall, F1, and accuracy metrics for the classifier and extraction models as appropriate for each phase of the process. We also provide the results for the application of our extraction model. With limited training data, we achieved a high degree of accuracy for both our classifier and for the extraction model for liver cancer disparities research literature performed using epidemiologic methods. The disparities concept was the most challenging to accurately classify, and concepts that appeared infrequently in our data set were the most difficult to extract. We provide a roadmap for using ML to classify and extract comprehensive information on multidisciplinary literature. Our technique can be adapted and modified for other cancers or diseases where disparities persist.

Legal Information Retrieval and Entailment Based on BM25, Transformer and Semantic Thesaurus Methods

We describe the techniques applied by the University of Alberta (UA) team in the most recent Competition on Legal Information Extraction and Entailment (COLIEE 2021). We participated in retrieval and entailment tasks for both case law and statute law; we applied a transformer-based approach for the case law entailment task, an information retrieval technique based on BM25 for legal information retrieval, and a natural language inference mechanism using semantic knowledge applied to statute law texts. This competition included 25 teams from 14 countries; our case law entailment approach was ranked no. 4 in Task 2, the BM25 technique for legal information retrieval was ranked no. 3 in Task 3, and the natural language inference technique incorporating semantic information was ranked no. 4 in Task 4. The combination of the latter two techniques on Task 5 was ranked no. 2. We also performed error analysis of our system in Task 4, which provides some insight into current state-of-the-art and research priorities for future directions.

A Method for Accurate 3-D Gaze Estimation With a Single Camera and Two Collinear Light Sources

The simplest configuration for estimating 3D line-of-sight based on corneal reflection and pupil refraction uses a single camera and two non-collinear light sources. The corneal radius and 3D corneal center can be determined by solving a system of six equations with six unknowns. However, when two light sources are collinear with the camera, the reflection planes of the two light sources are coplanar. In this case, the estimation of 3D corneal center is under-determined in the absence of any pre-known eyeball parameters (e.g., corneal radius). To solve this issue, this paper proposes to calibrate the corneal radius with corneal reflections from two light sources combined with an iris imaging model, so that the 3D corneal center is estimated with a known corneal radius. Then, the pupil features or iris features can be selected to estimate the 3D pupil center or 3D iris center, which uses an optimization solution for only the center parameters to avoid unnecessary errors. Thus, the optical axis of the eyeball is constructed with two spatial centers, and the kappa angle calculated in real time is used to convert the optical axis of the eyeball to the 3D line-of-sight. We simulated this gaze estimation process using pupil and iris features, and performed error analysis on the proposed method. In practical experiments, the proposed method was verified using iris features, and the gaze accuracy was 1.22° and 1.13° in the x- and y-directions, respectively, which is comparable to that of more sophisticated systems. A single-camera-two-collinear-light-source system is the most typical and common system configuration, and the realization of 3D gaze estimation under such a system configuration can effectively promote the wide application of eye-tracking interaction.

A class of lattice Boltzmann models for the Burgers equation with variable coefficient in space and time

<abstract><p>In this paper, we study the numerical results of the Burgers equation with the variable coefficient in space and time and then put forward a lattice Boltzmann model of backward difference solution of nonlinear system. The macroscopic equation is recovered by using the Chapman-Enskog method and the direct Taylor-series expansion method. These two methods can recover the same hydrodynamic equations and analyze various nonlinear systems. In particular, it is much easier to perform error analysis by using the direct Taylor method. In this study, the two methods are used to analyze the Burgers equation with variable coefficient in space and time, the numerical results are discussed and are compared with the analytical solution. The numerical results verify the effectiveness of the model. The stability of the model ensures that we can use larger time step lengths. The improvement of lattice speed can improve the computational performance of the model, and the D1Q7 lattice performance is much better than the D1Q5 lattice performance.</p></abstract>

Detecting ethnicity-targeted hate speech in Russian social media texts

Ethnicity-targeted hate speech has been widely shown to influence on-the-ground inter-ethnic conflict and violence, especially in such multi-ethnic societies as Russia. Therefore, ethnicity-targeted hate speech detection in user texts is becoming an important task. However, it faces a number of unresolved problems: difficulties of reliable mark-up, informal and indirect ways of expressing negativity in user texts (such as irony, false generalization and attribution of unfavored actions to targeted groups), users’ inclination to express opposite attitudes to different ethnic groups in the same text and, finally, lack of research on languages other than English. In this work we address several of these problems in the task of ethnicity-targeted hate speech detection in Russian-language social media texts. This approach allows us to differentiate between attitudes towards different ethnic groups mentioned in the same text – a task that has never been addressed before. We use a dataset of over 2,6M user messages mentioning ethnic groups to construct a representative sample of 12K instances (ethnic group, text) that are further thoroughly annotated via a special procedure. In contrast to many previous collections that usually comprise extreme cases of toxic speech, representativity of our sample secures a realistic and, therefore, much higher proportion of subtle negativity which additionally complicates its automatic detection. We then experiment with four types of machine learning models, from traditional classifiers such as SVM to deep learning approaches, notably the recently introduced BERT architecture, and interpret their predictions in terms of various linguistic phenomena. In addition to hate speech detection with a text-level two-class approach (hate, no hate), we also justify and implement a unique instance-based three-class approach (positive, neutral, negative attitude, the latter implying hate speech). Our best results are achieved by using fine-tuned and pre-trained RuBERT combined with linguistic features, with F1-hate=0.760, F1-macro=0.833 on the text-level two-class problem comparable to previous studies, and F1-hate=0.813, F1-macro=0.824 on our unique instance-based three-class hate speech detection task. Finally, we perform error analysis, and it reveals that further improvement could be achieved by accounting for complex and creative language issues more accurately, i.e., by detecting irony and unconventional forms of obscene lexicon.

An Error Analysis Framework for Shallow Surface Realization

Abstract The metrics standardly used to evaluate Natural Language Generation (NLG) models, such as BLEU or METEOR, fail to provide information on which linguistic factors impact performance. Focusing on Surface Realization (SR), the task of converting an unordered dependency tree into a well-formed sentence, we propose a framework for error analysis which permits identifying which features of the input affect the models’ results. This framework consists of two main components: (i) correlation analyses between a wide range of syntactic metrics and standard performance metrics and (ii) a set of techniques to automatically identify syntactic constructs that often co-occur with low performance scores. We demonstrate the advantages of our framework by performing error analysis on the results of 174 system runs submitted to the Multilingual SR shared tasks; we show that dependency edge accuracy correlate with automatic metrics thereby providing a more interpretable basis for evaluation; and we suggest ways in which our framework could be used to improve models and data. The framework is available in the form of a toolkit which can be used both by campaign organizers to provide detailed, linguistically interpretable feedback on the state of the art in multilingual SR, and by individual researchers to improve models and datasets.1

Prediction of the Number of Pure Electric Vehicles Based on the Extended GM(1,1) Model

The gray system theory is used to study the changes in the number of pure electric vehicles nationwide. According to the data on the number of pure electric vehicles in the country from 2016 to 2020, the original time series prediction model is established based on the GM(1,1) model. Using MATLAB to establish the original GM(1,1), new information GM(1,1) and metabolic GM(1,1), and compare them, and select the metabolic GM(1,1) with the smallest sum of squared errors for model performance Inspection and error analysis showed that the degree of fit was up to standard, and the predicted results were obtained. The prediction results show that the number of pure electric vehicles in my country will increase year by year from 2021 to 2023, and the number of pure electric vehicles in my country will exceed 10 million by the end of 2023.