Traditional Statistical Approaches Research Articles

Integrating a machine learning-driven fraud detection system based on a risk management framework

This article explores the application of machine learning techniques, specifically focusing on ensemble methods like Random Forests, for detecting fraudulent activities in digital financial transactions. Highlighting the evolution from traditional statistical approaches to modern machine learning models, it underscores the effectiveness of Random Forests in handling the inherent challenges of imbalanced datasets typical in fraud detection scenarios. Using a Kaggle dataset of credit card transactions, the study optimizes Random Forest parameters through rigorous parameter tuning, achieving significant improvements in model performance metrics such as Area Under the Curve (AUC). The findings underscore the critical role of machine learning in enhancing fraud detection capabilities, emphasizing the ongoing evolution and future potential of these methodologies in financial risk management.

Evaluating Salinity Tolerance of Pomegranate Cultivars Using Subordinate Function Analysis and Machine Learning

ABSTRACT The application of machine learning in salinity tolerance evaluation studies is limited. To address this gap, a pot culture experiment was conducted with six pomegranate cultivars at Punjab Agricultural University Ludhiana, India, in 2020 and 2021. The data were subjected to subordinated function analysis (SFA), random forest regression (RFR), support vector regression (SVR) and general linear modeling (GLM) after 60 and 120 days of first saline water treatment. The SFA identified ‘Bhagwa’ as the most salt tolerant cultivar, followed by ‘Wonderful.’ The RFR highlighted leaf K, catalase activity, and relative growth rate (RGR) as key parameters influencing salinity tolerance of the tested cultivars after 60 days. The SVR demonstrated root S content and net assimilation rate as critical factors. After 120 days, both RFR and SVR identified stomatal conductance and relative growth rate as crucial markers displaying salinity tolerance of cultivars. Both RFR and SVR showed high predictive accuracy, especially after 120 days, compared to GLM. However, GLM provided a broader set of variables for a better understanding of the underlying mechanism across different time intervals. Therefore, this research emphasizes the importance of using both machine learning techniques and traditional statistical approaches to gain a deeper understanding of the subject.

Health Analysis and Recommendation Based on Food Using Data Mining

Suitable nutritional diets have been widely recognized as important measures to prevent and control non-communicable diseases (NCDs). However, there is little research on nutritional ingredients in food now, which are beneficial to the rehabilitation of NCDs. In this project, we profoundly analyzed the relationship between nutritional ingredients and diseases by using data mining methods. First, more than n number of diseases was obtained, and we collected the recommended food and taboo food for each disease. The experiments on reallife data show that our method based on data mining improves the performance compared with the traditional statistical approach. We can assist doctors and disease researchers to find out positive nutritional ingredients that are conducive to the rehabilitation of the diseases as accurately as possible. At present, some data is not available, because they are still in the medical verification. The dataset uploaded will be pre-processed, Feature Extraction, noisy data will be removed, and classification of dataset will take places using random forest algorithm based on this analysis the diseases prediction takes places for the food intake by the individual

Understanding key mineral supply chain dynamics using economics‐informed material flow analysis and Bayesian optimization

AbstractThe low‐carbon energy transition requires significant increases in production for many mineral commodities. Understanding demand, technological requirements, and prices associated with this production increase requires understanding the supply chain dynamics of many minerals simultaneously, and via a consistent framework. A generalized economics‐informed material flow method, global materials modeling using Bayesian optimization, captures the market dynamics of key mineral commodities. The method relies only on a limited set of widely available historical data as input, enabling quantification of economic relationships (elasticities) for supply chain components where data are sparse, and relationships cannot be obtained via traditional statistical approaches. Building upon established material flow analysis (MFA) and economic modeling techniques, Bayesian optimization was applied to fit an economics‐informed MFA model to global historical demand, supply, and price for aluminum, copper, gold, lead, nickel, silver, iron, tin, and zinc. This approach enables estimates for the evolution of ore grades, mine costs, refining charges, sector‐specific demand, and scrap collection for each commodity. Economic relationships were quantified and compared with a database compiled from the literature, including 1333 values from 213 analyses across 65 publications. Discrepancies in methods and limited coverage make use of these parameters in modeling efforts difficult. This work provides a single, homogeneous, probabilistic approach to identifying economic relationships across mineral supply chains, with uncertainty quantification, a literature database for comparison, and a modeling framework in which to use them. This article met the requirements for a Gold‐Gold JIE data openness badge described at http://jie.click/badges.

Evaluation of Data Fitting Approaches for Speed/Flow Density Relationships

This paper presents guidance on data-fitting approaches in the context of pedestrian and evacuation dynamics research. In particular, it examines parametric and non-parametric regression techniques for analysing speed/flow density relationships. Parametric models assume predefined functional forms, while non-parametric models provide flexibility to capture complex relationships. This paper evaluates a range of traditional statistical approaches and machine-learning techniques. It emphasises the importance of weighting unbalanced datasets to enhance model accuracy. Practical applications are illustrated using traffic and pedestrian evacuation data. This paper is intended to stimulate discussion on best practices for developing, calibrating, and testing macroscopic and microscopic evacuation models. It does not prescribe a one-size-fits-all solution for evacuation data fitting approaches, but it provides an overview of existing methods and analyses their advantages and limitations.

Predicting Future Birth Rates with the Use of an Adaptive Machine Learning Algorithm: A Forecasting Experiment for Scotland.

The total fertility rate is influenced over an extended period of time by shifts in population socioeconomic characteristics and attitudes and values. However, it may be impacted by macroeconomic trends in the short term, although these effects are likely to be minimal when fertility is low. With the objective of forecasting monthly deliveries, this study concentrates on the analysis of registered births in Scotland. Through this approach, we examine the significance of precisely forecasting fertility trends, which can subsequently aid in the anticipation of demand in diverse sectors by allowing policymakers to anticipate changes in population dynamics and customize policies to tackle emerging demographic challenges. Consequently, this has implications for fiscal stability, national economic accounts and the environment. In conducting our analysis, we incorporated non-linear machine learning methods alongside traditional statistical approaches to forecast monthly births in an out-of-sample exercise that occurs one step in advance. The outcomes underscore the efficacy of machine learning in generating precise predictions within this particular domain. In sum, this research will comprehensively demonstrate a cutting-edge model of machine learning that utilizes several attributes to assist in clinical decision-making, predict potential complications during pregnancy and choose the appropriate delivery method, as well as help in medical diagnosis and treatment.

A longevity level-oriented wellness target area identification method: a case study of Yunnan Province, China.

Aging, as a global demographic issue, is characterized by its rapid growth, which drives an increase in people's healthcare awareness. The emergence of wellness bases caters to this market demand. Therefore, the identification of potential areas suitable for wellness activities and the construction of wellness bases, referred to as Wellness Target Areas (WTAs), becomes a crucial first step. Currently, commonly used identification methods are mostly based on traditional statistical approaches, which are often complex, cumbersome, and subject to potential risks of subjective assumptions, affecting the reliability of WTAs identification results. Longevity level serves as a comprehensive indicator reflecting the natural and socio-economic environment of a region, making it the most indicative of the regional wellness environment status. This study proposes using longevity level as the benchmark for WTAs identification to simplify the identification process and reduce the impact of subjective bias on the results. The study focuses on 129 county-level units in Yunnan Province. Firstly, the Geodetector (GD) is utilized to explore the complex interaction between the longevity level and the geographical environment to determine regional wellness factors. Secondly, using ArcGIS and geographical weighted regression (GWR), the study investigates the role of different wellness factors, ultimately classifying and grading the WTAs. The longevity level in Yunnan Province exhibits a pattern of multi-point clustering, forming three major longevity regions. Factors that significantly influence longevity level include annual average precipitation, sunshine duration, PM2.5 content, per capita disposable income, density of tourist attractions, and distance from residential areas to hospitals. Based on the degree of longevity and the contribution rate of influencing factors, Yunnan Province's WTAs are classified into three levels and two types (natural and comprehensive). Our study aims to establish a connection between longevity level and the selection of wellness bases, exploring regional wellness factors through the relationship between longevity phenomena and geographical environment, identifying potential construction areas for wellness bases (i.e., WTAs), providing new insights for the precise selection of wellness bases, effectively enhancing the scientificity of site selection, promoting population health, and contributing to the global aging process with better health.

Physics-based vs. data-driven 24-hour probabilistic forecasts of precipitation for northern tropical Africa

Abstract Numerical weather prediction (NWP) models struggle to skillfully predict tropical precipitation occurrence and amount, calling for alternative approaches. For instance, it has been shown that fairly simple, purely data-driven logistic regression models for 24-hour precipitation occurrence outperform both climatological and NWP forecasts for the West African summer monsoon. More complex neural network based approaches, however, remain underdeveloped due to the non-Gaussian character of precipitation. In this study, we develop, apply and evaluate a novel two-stage approach, where we train a U-Net convolutional neural network (CNN) model on gridded rainfall data to obtain a deterministic forecast and then apply the recently developed, nonparametric Easy Uncertainty Quantification (EasyUQ) approach to convert it into a probabilistic forecast. We evaluate CNN+EasyUQ for one-day ahead 24-hour accumulated precipitation forecasts over northern tropical Africa for 2011–2019, with the Integrated Multi-satellitE Retrievals for GPM (IMERG) data serving as ground truth. In the most comprehensive assessment to date we compare CNN+EasyUQ to state-of-the-art physics-based and data-driven approaches such as a monthly probabilistic climatology, raw and postprocessed ensemble forecasts from the European Centre for Medium-Range Weather Forecasts (ECMWF), and traditional statistical approaches that use up to 25 predictor variables from IMERG and the ERA5 reanalysis. Generally, statistical approaches perform about on par with postprocessed ECMWF ensemble forecasts. The CNN+EasyUQ approach, however, clearly outperforms all competitors for both occurrence and amount. Hybrid methods that merge CNN+EasyUQ and physics-based forecasts show slight further improvement. Thus, the CNN+EasyUQ approach can likely improve operational probabilistic forecasts of rainfall in the tropics, and potentially even beyond.

Technical Perspective: Synthetic Data Needs a Reproducibility Benchmark

Synthetic data is a vital substitute for real sensitive personal data in supporting social science research and policy studies. Extensive prior research has delved into various models for generating synthetic data, from traditional statistical approaches to cutting-edge deep-learning methods. However, selecting the most suitable one for unforeseen applications poses a significant challenge due to the varying strengths and weaknesses, dependent on factors such as the application domain, data distribution, analytical requirements, and privacy considerations.

Improved Autuoformer Electricity load forecasting based on model fusion

The development of the power system is undergoing rapid changes. The future direction of power system development is inevitably smart grids. This paper proposes an optimization of the Autoformer model based on model fusion to improve the accuracy of electricity load forecasting. By incorporating the predictive results of traditional statistical multivariate linear regression (MLR) and machine learning method LightGBM into the Autoformer model, we aim to fully leverage the strengths of both traditional statistical and machine learning approaches to enhance the predictive accuracy of the Autoformer model. Experimental results, using mean MAE and MSE as the evaluation metrics, demonstrate that the algorithm proposed in this paper exhibits better performance in the prediction results.

ECG-surv: A deep learning-based model to predict time to 1-year mortality from 12-lead electrocardiogram

Electrocardiogram (ECG) abnormalities have demonstrated potential as prognostic indicators of patient survival. However, the traditional statistical approach is constrained by structured data input, limiting its ability to fully leverage the predictive value of ECG data in prognostic modeling. This study aims to introduce and evaluate a deep-learning model to simultaneously handle censored data and unstructured ECG data for survival analysis. We herein introduce a novel deep neural network called ECG-surv, which includes a feature extraction neural network and a time-to-event analysis neural network. The proposed model is specifically designed to predict the time to 1-year mortality by extracting and analyzing unique features from 12-lead ECG data. ECG-surv was evaluated using both an independent test set and an external set, which were collected using different ECG devices. The performance of ECG-surv surpassed that of the Cox proportional model, which included demographics and ECG waveform parameters, in predicting 1-year all-cause mortality, with a significantly higher concordance index (C-index) in ECG-surv than in the Cox model using both the independent test set (0.860 [95% CI: 0.859- 0.861] vs. 0.796 [95% CI: 0.791- 0.800]) and the external test set (0.813 [95% CI: 0.807- 0.814] vs. 0.764 [95% CI: 0.755- 0.770]). ECG-surv also demonstrated exceptional predictive ability for cardiovascular death (C-index of 0.891 [95% CI: 0.890- 0.893]), outperforming the Framingham risk Cox model (C-index of 0.734 [95% CI: 0.715-0.752]). ECG-surv effectively utilized unstructured ECG data in a survival analysis. It outperformed traditional statistical approaches in predicting 1-year all-cause mortality and cardiovascular death, which makes it a valuable tool for predicting patient survival.

Time Series Analysis and Algorithm for Fluctuation Pattern Recognition and Forecasting of Economic Indicators

This study examines the use of artificial neural network (ANN) algorithms in machine learning for recognizing fluctuation patterns and forecasting economic indicators in the field of political economy. In today's changing global economic scene, economic forecasting is critical for informing policy decisions, directing strategic planning, and managing risks. Traditional statistical approaches and time series analysis have long been used for this purpose, but the advent of ANN algorithms provides a new way to capture the intricate dynamics of economic systems. Using ANNs' intrinsic flexibility and nonlinear modelling capabilities, this study analyzes their ability to effectively forecast future swings in economic indices. This work intends to understand the strengths, limits, and practical consequences of ANN-based forecasting models by conducting a complete assessment of literature, methodological methods, and empirical studies. This study adds to the continuing discussion of economic forecasting approaches by diving into theoretical frameworks, methodological concerns, and empirical data. Finally, incorporating ANN algorithms into machine learning offers promising opportunities for improving the understanding of economic processes and supporting informed decision-making in the political economy.

ML-Based Software Defect Prediction in Embedded Software for Telecommunication Systems (Focusing on the Case of SAMSUNG ELECTRONICS)

Software stands out as one of the most rapidly evolving technologies in the present era, characterized by its swift expansion in both scale and complexity, which leads to challenges in quality assurance. Software defect prediction (SDP) has emerged as a methodology crafted to anticipate undiscovered defects, leveraging known defect data from existing codes. This methodology serves to facilitate software quality management, thereby ensuring overall product quality. The methodologies of machine learning (ML) and one of its branches, deep learning (DL), exhibit superior accuracy and adaptability compared to traditional statistical approaches, catalyzing active research in this domain. However, it makes it hard to generalize, not only because of the disparity between open-source projects and commercial projects but also due to the differences in each industrial sector. Consequently, further research utilizing datasets sourced from diverse real-world sectors has become imperative to bolster the applicability of these findings. For this study, we utilized embedded software for use with the telecommunication systems of Samsung Electronics, supplemented by the introduction of nine novel features to train the model, and a subsequent analysis of the results ensued. The experimental outcomes revealed that the F-measurement metric has been enhanced from 0.58 to 0.63 upon integration of the new features, thereby signifying a performance augmentation of 8.62%. This case study is anticipated to contribute to bolstering the application of SDP methodologies within analogous industrial sectors.

Comparison of deep learning models for milk production forecasting at national scale

Forecasting of agricultural variables at national scale is essential for developing policies aimed at guaranteeing food security and increase the stability of the entire agri-food chain. However, this is often a challenging task due to the fact that data are usually sparse and long records are rarely available. This work aims at evaluating the potentiality of ten deep learning models for predicting monthly milk production in France, Germany, and Italy, using as input, climatic and economic variables from open datasets. The results indicate that deep learning models are a better alternative to traditional statistical approaches, providing robust results without requiring complex model architectures. Furthermore, the prominent autoregressive nature of milk production highlights the inability of environmental variables to capture external processes that influence the milk production (e.g., milk quota). However, the high accuracy achieved by the models lays the ground for possible application of deep learning models to accurately forecast agricultural variables at national scale with potential implications for the development of dairy insurance products, and risk management practices.

On the integration of inspection data with seismic resilience assessment of corroded reinforced concrete structures

This research aims to improve the traditional statistical approach to evaluate the seismic resilience of concrete structures exposed to chloride attack. In the mentioned approach, the uncertain parameters are modeled using certain probability distribution functions (PDFs) selected from some standards such as fib. However, the mentioned PDFs should contain the actual conditions of the desired structure. This paper introduces a Bayesian framework for updating PDFs used in deterioration analysis, which is an important part of the life-cycle seismic resilience assessment process of corroded concrete structures. The proposed methodology was applied to a two-story reinforced concrete school in Iran that suffered from chloride corrosion for twenty years. The seismic resilience of the structure was investigated for a 50-year life cycle. The results showed that the updated model can significantly reduce the uncertainty of the influencing parameters in the seismic resilience evaluation process. Using the updated model resulted in a conservative seismic resilience index compared to the non-updated model. This research shows that the role of inspection data in periodically updating the seismic-resistance calculations of corroded concrete structures is not only necessary but also possible and beneficial.

Integration and holistic analysis of multiple multidimensional soil data sets

Complex matrices such as soil have a range of measurable characteristics, and thus data to describe them can be considered multidimensional. These characteristics can be strongly influenced by factors that introduce confounding effects that hinder analyses. Traditional statistical approaches lack the flexibility and granularity required to adequately evaluate such matrices, particularly those with large dataset of varying data types (i.e. quantitative non-compositional, quantitative compositional). We present a statistical workflow designed to effectively analyse complex, multidimensional systems, even in the presence of confounding variables. The developed methodology involves exploratory analysis to identify the presence of confounding variables, followed by data decomposition (including strategies for both compositional and non-compositional quantitative data) to minimise the influence of these confounding factors such as sampling site/location. These data processing methods then allow for common patterns to be highlighted in the data, including the identification of biomarkers and determination of non-trivial associations between variables. We demonstrate the utility of this statistical workflow by jointly analysing the chemical composition and fungal biodiversity of New Zealand vineyard soils that have been managed with either organic low-input or conventional input approaches. By applying this pipeline, we were able to identify biomarkers that distinguish viticultural soil from both approaches and also unearth links and associations between the chemical and metagenomic profiles. While soil is an example of a system that can require this type of statistical methodology, there are a range of biological and ecological systems that are challenging to analyse due to the complex interplay of global and local effects. Utilising our developed pipeline will greatly enhance the way that these systems can be studied and the quality and impact of insight gained from their analysis.

Causal machine learning for predicting treatment outcomes.

Causal machine learning (ML) offers flexible, data-driven methods for predicting treatment outcomes including efficacy and toxicity, thereby supporting the assessment and safety of drugs. A key benefit of causal ML is that it allows for estimating individualized treatment effects, so that clinical decision-making can be personalized to individual patient profiles. Causal ML can be used in combination with both clinical trial data and real-world data, such as clinical registries and electronic health records, but caution is needed to avoid biased or incorrect predictions. In this Perspective, we discuss the benefits of causal ML (relative to traditional statistical or ML approaches) and outline the key components and steps. Finally, we provide recommendations for the reliable use of causal ML and effective translation into the clinic.

Faithful Trip Recommender Using Diffusion Guidance (Student Abstract)

Trip recommendation aims to plan user’s travel based on their specified preferences. Traditional heuristic and statistical approaches often fail to capture the intricate nuances of user intentions, leading to subpar performance. Recent deep-learning methods show attractive accuracy but struggle to generate faithful trajectories that match user intentions. In this work, we propose a DDPM-based incremental knowledge injection module to ensure the faithfulness of the generated trajectories. Experiments on two datasets verify the effectiveness of our approach.

A Machine Learning Study of Anxiety-related Symptoms and Error-related Brain Activity.

Changes in error processing are observable in a range of anxiety-related disorders. Numerous studies, however, have reported contradictory and nonreplicating findings, thus the exact mapping of brain response to errors (i.e., error-related negativity [ERN]; error-related positivity [Pe]) onto specific anxiety symptoms remains unclear. In this study, we collected 16 self-reported scores of anxiety dimensions and obtained spatial features of EEG recordings from 171 individuals. We then used machine learning to (1) identify symptoms that are central for elevated ERN/Pe and (2) estimate the generalizability of traditional statistical approaches. ERN was associated with rumination, threat overestimation, and inhibitory intolerance of uncertainty. Pe was associated with rumination, prospective intolerance of uncertainty, and behavioral inhibition. Our findings emphasize that not only the amplitude of ERN but also other sources of brain signal variance encode information relevant to individual differences in error processing. The results of the generalizability check reveal the need for a change in result-validation methods to move toward robust findings that reflect stable individual differences and clinically useful biomarkers. Our study benefits from the use of machine learning to improve the generalizability of results.

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.