Traditional Statistical Models Research Articles

Injury severity prediction and exploration of behavior-cause relationships in automotive crashes using natural language processing and extreme gradient boosting

Addressing the global challenge of traffic crashes necessitates transcending traditional statistical models, which often fail to fully capture the interactions between factors causing crashes. This oversight restricts the predictive accuracy and adaptability of current methodologies. Additionally, there is a notable gap in research that examines the links between behavior-cause relationships and crash injury severity. Our study deploys Natural Language Processing (NLP) and Frequent Pattern (FP) growth algorithm to mine crash narratives for behavior-cause connections, combines with the predictive strength of eXtreme Gradient Boosting (XGBoost) and the interpretative clarity offered by SHapley Additive exPlanations (SHAP), our approach not only predicts crash injury severity with satisfactory precision but also explains the influence of specific behavior-cause and environment conditions on crash outcomes. The integration of NLP and XGBoost, complemented by SHAP insights, has shown promising results with an accuracy of 0.79, outperforming traditional discrete choice models and competes closely with other machine learning approaches, including Support Vector Machines, Random Forest, Categorical Boosting (CatBoost), and Light Gradient Boosting Machine (LightGBM). Through detailed textual analysis and the establishment of a behavior-cause matrix, identifying five broad crash causes linked to 141 specific crash cause with behaviors, we uncover critical patterns such as the prominence of distracted driving in severe crashes. This comprehensive approach not only fills a critical research gap by linking behavior-cause relationships with injury severity but also sets the stage for developing targeted interventions to enhance road safety.

A Deformation Analysis Method for Sluice Structure Based on Panel Data

Deformation, as the most intuitive index, can reflect the operation status of hydraulic structures comprehensively, and reasonable analysis of deformation behavior has important guiding significance for structural long-term service. Currently, the health evaluation of dam deformation behavior has attracted widespread attention and extensive research from scholars due to its great importance. However, given that the sluice is a low-head hydraulic structure, the consequences of its failure are easily overlooked without sufficient attention. While the influencing factors of the sluice’s deformation are almost identical to those of a concrete dam, nonuniform deformation is the key issue in the sluice’s case because of the uneven property of the external load and soil foundation, and referencing the traditional deformation statistical model of a concrete dam cannot directly represent the nonuniform deformation behavior of a sluice. In this paper, we assume that the deformation at various positions of the sluice consist of both overall and individual effects, where overall effect values describe the deformation response trend of the sluice structure under external loads, and individual effect values represent the degree to which the deformation of a single point deviates from the overall deformation. Then, the random coefficient model of panel data is introduced into the analysis of sluice deformation to handle the unobservable overall and individual effects. Furthermore, the maximum entropy principle is applied, both to approximate the probability distribution function of individual effect extreme values and to determine the early warning indicators, completing the assessment and analysis of the nonuniform deformation state. Finally, taking a project as an example, we show that the method proposed can effectively identify the overall deformation trend of the sluice and the deviation degree of each measuring point from the overall deformation, which provides a novel approach for sluice deformation behavior research.

An Efficient Flood Forecasting Model Using an Optimal Deep Belief Network

<p>Floods inflict significant damage globally each year, underscoring the importance of accurate and timely flood prediction to mitigate property loss and life. Precise flood prediction provides governments with crucial preemptive alerts regarding potential flood disasters, enabling timely evacuations and life-saving measures. Although various ML (machine learning) models have demonstrated improved performance compared to traditional statistical models in flood prediction, they often overlook the spatial features crucial to understanding the generation of floods. DL (deep learning) is used in flood prediction to enhance the promptness and efficiency of flood levels predictions. This work presents an optimized DL model to forecast floods using time series data. Initially, the data set was cleaned and normalized by linear interpolation. Then, the DL model ODBN (optimal deep belief network) is utilized for flood forecasting time series prediction. ODBN is the integration of DBN and SCA (Sinh-Cosh algorithm). The experimental analysis is carried out on the real-time dataset and achieved better MSE and RMSE values of 0.75 and 0.94 respectively. The findings suggest that the use of an ODBN is an effective method of accurately forecasting floods.</p>

ІННОВАЦІЙНИЙ ПІДХІД У ПРОГНОЗУВАННІ ЧАСОВИХ РЯДІВ: ВІД ТРАДИЦІЙНИХ МЕТОДІВ ДО НОВАТОРСЬКОЇ МОДЕЛІ TIMESFM

Forecasting time series is crucial for companies as it enables them to make informed decisions using historical data and future predictions. Various approaches exist for time series forecasting, each tailored to different data characteristics and forecasting tasks. Currently, widely known models include ARIMA, LSTM networks, Prophet, and XGBoost. However, these models often require significant preprocessing time. To address this issue and leverage recent advancements in generative AI, Google introduced TimesFM, a decoder-only model specifically designed for time series forecasting. TimesFM utilizes transformer layers and a multi-layer perceptron block to transform time series fragments into tokens, enabling efficient forecasting with minimal generation steps. Synthetic and real-world data are combined for pretraining to capture fundamental temporal patterns and enhance model generalization. Evaluation demonstrates TimesFM's competitive performance across various benchmark time series datasets compared to traditional statistical methods and DL models.

Minimizing of Forecasting Error in Fuzzy Time Series Model Using Graph-Based Clustering Method

In recent years, numerous fuzzy time series (FTS) forecasting models have been developed to address complex and incomplete problems. However, the accuracy of these models is specific to the problem at hand and varies across datasets. Despite claims of superiority over traditional statistical and single machine learning-based models, achieving improved forecasting accuracy remains a formidable challenge. In FTS models, the lengths of intervals and fuzzy relationship groups are considered crucial factors influencing forecasting accuracy. Hence, this study introduces an FTS forecasting model based on the graph-based clustering technique. The clustering algorithm, utilized during the fuzzification stage, enables the derivation of unequal interval lengths. The proposed model is applied to forecast two numerical datasets: enrollment data from the University of Alabama and the datasets of Gas prices RON95 in Vietnam. Comparisons of forecasting results between the proposed model and others are conducted for enrollment forecasts at the University of Alabama. The findings reveal that the proposed model achieves higher forecasting accuracy across all orders of fuzzy relationships when compared to its counterparts

GIS and Machine Learning Models Target Dynamic Settlement Patterns and Their Driving Mechanisms from the Neolithic to Bronze Age in the Northeastern Tibetan Plateau

Traditional GIS-based statistical models are intended to extrapolate patterns of settlements and their interactions with the environment. They contribute significantly to our knowledge of past human–land relationships. Yet, these models are often criticized for their empiricism, lopsided specific factors, and for overlooking the synergy between variables. Though largely untested, machine learning and artificial intelligence methods have the potential to overcome these shortcomings comprehensively and objectively. The northeastern Tibetan Plateau (NETP) is characterized by diverse environments and significant changes to the social system from the Neolithic to Bronze Age. In this study, this area serves as a representative case for assessing the complex relationships between settlement locations and geographic environments, taking full advantages of these new models. We have explored a novel modeling case by employing GIS and random forests to consider multiple factors, including terrain, vegetation, soil, climate, hydrology, and land suitability, to construct classification models identifying environmental variation across different cultural periods. The model exhibited strong performance and a high archaeological prediction value. Potential living maps were generated for each cultural stage, revealing distinct environmental selection strategies from the Neolithic to Bronze Age. The key environmental parameters of elevation, climate, soil erosion, and cultivated land suitability were calculated with high weights, influencing human environmental decisions synergistically. Furthermore, we conducted a quantitative analysis of temporal dynamics in climate and subsistence to understand driving mechanisms behind environmental strategies. These findings suggest that past human environmental strategies were based on the comprehensive consideration of various factors, coupled with their social economic scenario. Such subsistence-oriented activities supported human beings in overcoming elevation limitation, and thus allowed them to inhabit wider pastoral areas. This study showcases the potential of machine learning in predicting archaeological probabilities and in interpreting the environmental influence on settlement patterns.

STUCK PIPE MITIGATION: FINDINGS, ANALYSIS, AND MACHINE LEARNING SOLUTIONS

The occurrence of stuck pipes during drilling presents significant challenges, leading to nonproductive time (NPT) and potential damage to equipment. This paper delves into the reasons behind, effects of, and methods for addressing incidents of stuck pipes in drilling actions. Extensive literature review and analysis reveal the prevalence of stuck pipe events and their diverse underlying causes, ranging from differential sticking to mechanical jamming. Additionally, the study identifies key drilling parameters and signals indicative of potential stuck pipe incidents, emphasizing the importance of early detection and intervention to prevent adverse outcomes. Proposing the utilization of machine learning algorithms and statistical models for predictive analytics as a proactive strategy to mitigate the risk of stuck pipe incidents. The paper discusses the limitations of traditional statistical models and highlights the efficacy of machine learning methods, such as artificial neural networks (ANNs) and support vector machines (SVMs), in accurately modeling drilling processes. Furthermore, practical machine learning models using real-time drilling data are presented as effective tools for early detection and timely intervention in stuck pipe incidents. The study also describes various software tools used to assess and mitigate risks in drilling operations. Overall, this paper presents prospects for improving operational efficiency and safety of drilling operations using advanced analytical methods and software solutions. Keywords: stuck pipe, type of stuck, drillstring immobilization, nonproductive time, machine learning in drilling, risk analysis software.

Prognostic models and factors identifying end-of-life in non-cancer chronic diseases: a systematic review

BackgroundPrecise prognostic information, if available, is very helpful for guiding treatment decisions and resource allocation in patients with non-cancer non-communicable chronic diseases (NCDs). This study aimed to systematically review the...

Application of the improved dung beetle optimizer, muti-head attention and hybrid deep learning algorithms to groundwater depth prediction in the Ningxia area, China

Due to the lack of accurate data and complex parameterization, the prediction of groundwater depth is a challenge for numerical models. Machine learning can effectively solve this issue and has been proven useful in the prediction of groundwater depth in many areas. In this study, two new models are applied to the prediction of groundwater depth in the Ningxia area, China. The two models combine the improved dung beetle optimizer (DBO) algorithm with two deep learning models: The Multi-head Attention–Convolution Neural Network–Long Short Term Memory networks (MH-CNN-LSTM) and the Multi-head Attention–Convolution Neural Network–Gated Recurrent Unit (MH-CNN-GRU). The models with DBO show better prediction performance, with larger R (correlation coefficient), RPD (residual prediction deviation), and lower RMSE (root-mean-square error). Compared with the models with the original DBO, the R and RPD of models with the improved DBO increase by over 1.5%, and the RMSE decreases by over 1.8%, indicating better prediction results. In addition, compared with the multiple linear regression model, a traditional statistical model, deep learning models have better prediction performance.摘要本研究将两个新模型应用于位于中国西北干旱半干旱区的宁夏地区地下水深度预测. 这两个模型将改进的蜣螂优化 (DBO) 算法与两个深度学习模型相结合, 即多头注意力-卷积神经网络-长短期记忆网络和多头注意力-回旋神经网络-门控递归单元. 带有DBO的模型预测结果表现出更大的相关系数 (R) , 残差预测偏差 (RPD) 和较低的均方根误差 (RMSE) , 预测结果更好. 此外, 与DBO模型相比, 改进后的DBO模型的R和RPD增加了1.5%以上, RMSE降低了1.8%以上, 表明预测结果更好. 与传统的统计模型多元线性回归模型相比, 深度学习模型具有更好的预测性能.

A Bayesian convolutional neural network-based generalized linear model.

Convolutional neural networks (CNNs) provide flexible function approximations for a wide variety of applications when the input variables are in the form of images or spatial data. Although CNNs often outperform traditional statistical models in prediction accuracy, statistical inference, such as estimating the effects of covariates and quantifying the prediction uncertainty, is not trivial due to the highly complicated model structure and overparameterization. To address this challenge, we propose a new Bayesian approach by embedding CNNs within the generalized linear models (GLMs) framework. We use extracted nodes from the last hidden layer of CNN with Monte Carlo (MC) dropout as informative covariates in GLM. This improves accuracy in prediction and regression coefficient inference, allowing for the interpretation of coefficients and uncertainty quantification. By fitting ensemble GLMs across multiple realizations from MC dropout, we can account for uncertainties in extracting the features. We apply our methods to biological and epidemiological problems, which have both high-dimensional correlated inputs and vector covariates. Specifically, we consider malaria incidence data, brain tumor image data, and fMRI data. By extracting information from correlated inputs, the proposed method can provide an interpretable Bayesian analysis. The algorithm can be broadly applicable to image regressions or correlated data analysis by enabling accurate Bayesian inference quickly.

Cryptocurrency assets valuation based on statistics model and machine learning

As a matter of fact, cryptocurrency is an exchange media, which uses cryptographic functions to trade, which achieves decentralisation, transparency, and invariance using blockchain. With the rapid development since the first white book proposed in 2008. The market value of cryptocurrency has been exponentially expanded rapidly in recent years. On this basis, programmatic trading strategy are paid more attention by people in trading cryptocurrency contemporarily. With this in mind, this study provides three models, that is, Exponential Smoothing Model (ESM), Autoregressive Integrated Moving Average (ARIMA) model and Long short-term memory (LSTM) model to forecast the future price of cryptocurrency and use sum of square error to analyse the results of these prediction. According to the analysis, LSTM is the best model to forecast the price of cryptocurrency in this article. Overall, machine learning might be more useful to forecast in this fields than traditional statistical models. Investors and researchers in the future could continue to forecast the price of cryptocurrency combined with LSTM model.

Gait kinematics and kinetics in patients with chronic ankle instability and healthy controls: A statistical parametric mapping analysis

BackgroundChronic ankle instability (CAI) is associated with changes in gait biomechanics which may be related to chronic dysfunction. Traditional statistical models may be limited in their ability to assess the complex 3D movement of the lower extremity during gait. Multivariate analysis of the lower extremity kinematics may reveal unique biomechanical differences associated with CAI. Research Question: Do patients with CAI differ from healthy controls in their lower extremity biomechanics and GRF when comparing 3D biomechanics? MethodsThirty-nine young, active adults participated in this study. Data was collected using a 3D motion analysis system while patients walked and jogged. Statistical parametric mapping (SPM) was used to explore 3D GRF, kinematics and kinetics of the of the lower extremity of CAI and healthy patients. ResultsDuring walking, patients with CAI had greater inversion from 68–100% of gait cycle (p < 0.001, mean difference=3.2°). During jogging, patients with CAI had greater inversion from 20–92% (p < 0.001, mean difference=4.6°). Greater plantar flexion moments were found from 65–71% (p = 0.05, mean difference=347.4Nm/kg) and greater eversion moments were found from 95–100% (p = 0.03, mean difference=74.6Nm/kg) in the CAI group. No differences in GRF were found.Significance:Greater inversion may present a potentially injurious position. A faulty position of the rearfoot may require greater muscle function in order to correct the position of the joint resulting in greater eversion moments at the ankle. However, this kinetic change does not appear to correct the ankle position.

Machine Learning Approaches to Predict Symptoms in People With Cancer: Systematic Review.

People with cancer frequently experience severe and distressing symptoms associated with cancer and its treatments. Predicting symptoms in patients with cancer continues to be a significant challenge for both clinicians and researchers. The rapid evolution of machine learning (ML) highlights the need for a current systematic review to improve cancer symptom prediction. This systematic review aims to synthesize the literature that has used ML algorithms to predict the development of cancer symptoms and to identify the predictors of these symptoms. This is essential for integrating new developments and identifying gaps in existing literature. We conducted this systematic review in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) checklist. We conducted a systematic search of CINAHL, Embase, and PubMed for English records published from 1984 to August 11, 2023, using the following search terms: cancer, neoplasm, specific symptoms, neural networks, machine learning, specific algorithm names, and deep learning. All records that met the eligibility criteria were individually reviewed by 2 coauthors, and key findings were extracted and synthesized. We focused on studies using ML algorithms to predict cancer symptoms, excluding nonhuman research, technical reports, reviews, book chapters, conference proceedings, and inaccessible full texts. A total of 42 studies were included, the majority of which were published after 2017. Most studies were conducted in North America (18/42, 43%) and Asia (16/42, 38%). The sample sizes in most studies (27/42, 64%) typically ranged from 100 to 1000 participants. The most prevalent category of algorithms was supervised ML, accounting for 39 (93%) of the 42 studies. Each of the methods-deep learning, ensemble classifiers, and unsupervised ML-constituted 3 (3%) of the 42 studies. The ML algorithms with the best performance were logistic regression (9/42, 17%), random forest (7/42, 13%), artificial neural networks (5/42, 9%), and decision trees (5/42, 9%). The most commonly included primary cancer sites were the head and neck (9/42, 22%) and breast (8/42, 19%), with 17 (41%) of the 42 studies not specifying the site. The most frequently studied symptoms were xerostomia (9/42, 14%), depression (8/42, 13%), pain (8/42, 13%), and fatigue (6/42, 10%). The significant predictors were age, gender, treatment type, treatment number, cancer site, cancer stage, chemotherapy, radiotherapy, chronic diseases, comorbidities, physical factors, and psychological factors. This review outlines the algorithms used for predicting symptoms in individuals with cancer. Given the diversity of symptoms people with cancer experience, analytic approaches that can handle complex and nonlinear relationships are critical. This knowledge can pave the way for crafting algorithms tailored to a specific symptom. In addition, to improve prediction precision, future research should compare cutting-edge ML strategies such as deep learning and ensemble methods with traditional statistical models.

A Novel Ensemble Machine Learning Model for Oil Production Prediction with Two-Stage Data Preprocessing

Petroleum production forecasting involves the anticipation of fluid production from wells based on historical data. Compared to traditional empirical, statistical, or reservoir simulation-based models, machine learning techniques leverage inherent relationships among historical dynamic data to predict future production. These methods are characterized by readily available parameters, fast computational speeds, high precision, and time–cost advantages, making them widely applicable in oilfield production. In this study, time series forecast models utilizing robust and efficient machine learning techniques are formulated for the prediction of production. We have fused the two-stage data preprocessing methods and the attention mechanism into the temporal convolutional network-gated recurrent unit (TCN-GRU) model. Firstly, the random forest (RF) algorithm is employed to extract key dynamic production features that influence output, serving to reduce data dimensionality and mitigate overfitting. Next, the mode decomposition algorithm, complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), is introduced. It employs a decomposition–reconstruction approach to segment production data into high-frequency noise components, low-frequency regular components and trend components. These segments are then individually subjected to prediction tasks, facilitating the model’s ability to capture more accurate intrinsic relationships among the data. Finally, the TCN-GRU-MA model, which integrates a multi-head attention (MA) mechanism, is utilized for production forecasting. In this model, the TCN module is employed to capture temporal data features, while the attention mechanism assigns varying weights to highlight the most critical influencing factors. The experimental results indicate that the proposed model achieves outstanding predictive performance. Compared to the best-performing comparative model, it exhibits a reduction in RMSE by 3%, MAE by 1.6%, MAPE by 12.7%, and an increase in R2 by 2.6% in Case 1. Similarly, in Case 2, there is a 7.7% decrease in RMSE, 7.7% in MAE, 11.6% in MAPE, and a 4.7% improvement in R2.

Optimizing Stock Price Prediction: Exploring LSTM Architectural Parameters in Financial Forecasting

Accurate stock price prediction is a crucial concern in the field of finance, impacting investors, traders, and the broader economic landscape. Traditional time series analysis and statistical models often fall short in capturing the dynamic and nonlinear nature of financial markets. Financial market forecasting has advanced significantly with the advent of deep learning technology, especially Long Short-Term Memory (LSTM) networks. This study conducts an in-depth exploration of LSTM networks' critical architectural parameters and their impact on stock price prediction. The performance of the LSTM-based model is assessed using historical stock price data from Amazon, with particular attention paid to hyperparameters like the number of neurons, dropout rates, and LSTM layers. The research reveals that a well-balanced model architecture with moderate units per layer (e.g., 64), a single layer, and a dropout rate of 0.7 is key to achieving accurate predictions in financial markets. The outcomes of this research hold practical significance for both investors and financial analysts. Furthermore, they emphasize the promising potential of LSTM networks in the realm of financial forecasting, opening avenues for future studies in finance. This investigation deepens the comprehension of LSTM network architecture and its applicability in predicting stock prices, providing valuable insights that extend beyond the finance sector.

Classifying Beta-Secretase 1 Inhibitor Activity for Alzheimer’s Drug Discovery with LightGBM

This study explores the utilization of LightGBM, a gradient-boosting framework, to classify the inhibitory activity of beta-secretase 1 inhibitors, addressing the challenges of Alzheimer's disease drug discovery. The study aims to enhance classification performance by focusing on overcoming the limitations of traditional statistical models and conventional machine-learning techniques in handling complex molecular datasets. By sourcing a dataset of 7298 compounds from the ChEMBL database and calculating molecular descriptors for each compound as features, we employed LightGBM in conjunction with a set of carefully selected molecular descriptors to achieve a nuanced analysis of compound activities. The model's efficiency was benchmarked against traditional machine-learning algorithms, revealing LightGBM's superior accuracy (84.93%), precision (87.14%), sensitivity (89.93%), specificity (77.63%), and F1-score (88.17%) in classifying beta-secretase 1 inhibitor activity. The study underscores the critical role of molecular descriptors in understanding drug efficacy, highlighting LightGBM's potential in streamlining the virtual screening process. Conclusively, the findings advocate for LightGBM's adoption in computational drug discovery, offering a promising avenue for advancing Alzheimer's disease therapeutic development by facilitating the identification of potential drug candidates with enhanced precision and reliability.

Predicting dental anxiety in young adults: classical statistical modelling approach versus machine learning approach

ObjectivesTo predict and identify the key demographic and clinical exposure factors associated with dental anxiety among young adults, and to compare if the traditional statistical modelling approach provides similar results to the machine learning (ML) approach in predicting factors for dental anxiety.MethodsA cross-sectional study of Western Illinois University students. Three survey instruments (sociodemographic questionnaire, modified dental anxiety scale (MDAS), and dental concerns assessment tool (DCA)) were distributed via email to the students using survey monkey. The dependent variable was the mean MDAS scores, while the independent variables were the sociodemographic and dental concern assessment variables. Multivariable analysis was done by comparing the classical statistical model and the machine learning model. The classical statistical modelling technique was conducted using the multiple linear regression analysis and the final model was selected based on Akaike information Criteria (AIC) using the backward stepwise technique while the machine learining modelling was performed by comparing two ML models: LASSO regression and extreme gradient boosting machine (XGBOOST) under 5-fold cross-validation using the resampling technique. All statistical analyses were performed using R version 4.1.3.ResultsThe mean MDAS was 13.73 ± 5.51. After careful consideration of all possible fitted models and their interaction terms the classical statistical approach yielded a parsimonious model with 13 predictor variables with Akaike Information Criteria (AIC) of 2376.4. For the ML approach, the Lasso regression model was the best-performing model with a mean RMSE of 0.617, R2 of 0.615, and MAE of 0.483. Comparing the variable selection of ML versus the classical statistical model, both model types identified 12 similar variables (out of 13) as the most important predictors of dental anxiety in this study population.ConclusionThere is a high burden of dental anxiety within this study population. This study contributes to reducing the knowledge gap about the impact of clinical exposure variables on dental anxiety and the role of machine learningin the prediction of dental anxiety. The predictor variables identified can be used to inform public health interventions that are geared towards eliminating the individual clinical exposure triggers of dental anxiety are recommended.

Credit risk modelling within the euro area in the COVID‐19 period: Evidence from an ICAS framework

AbstractThis paper develops a logistic regression model in an in‐house credit assessment system (ICAS) framework for predicting corporate defaults in the Greek economy. We consider the impact of the COVID‐19 pandemic and the associated government financial support schemes, aiming to protect against financial vulnerabilities, on the probability of default of non‐financial firms, as well as the relevant sectoral and firm‐size effects. In developing the ICAS framework, we address methodological issues such as the predictive performance of statistical versus machine learning approaches and the imbalanced dataset problem, indicating ways to evaluate such models with strong predictive power. Our findings suggest that the effect of the financial support measures dominates the pandemic shocks, thus substantially reducing the probability of firms' default, while the size‐ and industry‐based models show that firms in the micro and services sectors benefited the most. Furthermore, using a random forest model, our findings highlight the trade‐off between the transparency of traditional statistical models and the predictive value of machine learning models.

An Agent-Based Modeling Approach for the Diffusion Analysis of Electric Vehicles With Two-Stage Purchase Choice Modeling

Abstract Diffusion research of innovative technologies is crucial for new product positioning and strategic planning in product design. As a versatile system simulation method, agent-based modeling (ABM) has been used in many previous studies on the diffusion analysis of electric vehicles (EVs). In these simulations, modeling consumers' purchase decisions is a significant step. Previous studies often adopt simple rule-based decision criteria in this step, while an accurate purchase decision model can contribute to a more reasonable diffusion analysis of EVs. To fill this gap, this brief presents an agent-based modeling approach for the diffusion analysis of electric vehicles with two-stage choice modeling. The core idea is to separate consumers' decision-making process for purchasing cars into two stages. Consumers first form a small choice set from the whole auto market. Then, they make the final choice from the choice set built in the first stage. In addition, the word-of-mouth (WOM) effect and consumers' social networks are also considered in the ABM, which can further improve the accuracy of the diffusion analysis. A case study using data collected from Shanghai, China, is presented to demonstrate the proposed approach. Our approach outperforms other ablation models as well as traditional statistical models in the prediction accuracy of EV's market share. The influence of factors such as government policy and technological improvement on the diffusion of EVs is also discussed. These insights can assist automakers in improving their product design and enhancing their market competitiveness.

딥러닝 기반 시계열 분석 모델의 불확실성 정량화 비교 연구

With the advancement of artificial intelligence, machine learning, and deep learning, their applications in various industries, particularly finance, have increased. However, interpreting predictions from deep learning models poses challenges, especially in finance where result interpretation is important. This study aims to determine the uncertainty of stable deep learning models, despite changes in the model like dropout, to establish standards for reliable models and identify those detecting anomal data through model uncertainty. In the experiment, the traditional statistical model ARIMA and deep learning models mainly used for time series analysis, CNN, LSTM, MLP, and CNN-LSTM. Uncertainty was measured from a Bayesian perspective using MC Dropout. The experimental results confirmed that deep learning models performed better than statistical models for various patterns of time series data. It was observed that, even if the performance was not the best, LSTM based models exhibited low uncertainty, indicating stability. Consequently, this study highlights the importance of considering uncertainty along with accuracy in model selection. Moreover, it was confirmed that models with higher uncertainty are suitable for anomaly detection, making CNN based models particularly fitting for this purpose.