Class Of Regression Models Research Articles

Applications of machine learning in quantitative trading

Abstract. This paper also addresses quantitative trading where machine learning techniques are applied. Quantitative trading uses historical data and makes future predictions that inform and optimise trading strategies. There are two machine learning techniques used in quantitative trading: supervised and reinforcement learning. In a supervised learning setting, machine learning techniques such as regression analysis, classification models and time series prediction are used to forecast future possible outcomes in the market. Data such as historical prices, volume and financial indicators are used to determine trends that inform and optimise trading decisions. Reinforcement learning is another machine learning technique in quantitative trading where adaptive strategies using a reward system are designed. These strategies then analyse market dynamics and optimise trades through interaction and feedback. The third technique, deep learning, takes neural networks and uses them to process large-scale, complex data. It's particularly useful in quantitative trading as it generates more accurate predictions; for instance, neural networks are capable of learning trading signals from price and volume data and even from news headlines. By enabling computers to analyse large quantities of training data, deep learning improves the predictive accuracy of future price trends. It also helps to generate trading signals that are more robust to market conditions.

A StarGAN and transformer-based hybrid classification-regression model for multi-institution VMAT patient-specific quality assurance.

The field of artificial intelligence (AI)-based patient-specific quality assurance (PSQA) for volumetric modulated arc therapy (VMAT) faces challenges in terms of developing general models across institutions due to the prevalence of multi-institution data collection and multivariate heterogeneity. Building a general model that is capable of handling diverse multi-institution data is critical for enabling large-scale integration and analysis. This study aims to develop a star generative adversarial network (StarGAN) and transformer-based hybrid classification-regression PSQA framework to address unification of heterogeneous data from different institutions. A StarGAN and transformer-based hybrid classification-regression model was developed as a general PSQA framework to predict gamma passing rates (GPRs) and classify quality assurance (QA) results as "Pass" or "Fail" at multiple institutions. A total of 1815 VMAT plans were collected from eight institutions to develop the general PSQA framework and perform clinical commissioning and implementation. Among them, 20 independent clinical plans from each of eight institutions, for a total of 160 plans, were used for the clinical commissioning, and 205 new clinical plans from eight institutions were used for clinical implementation. For the 3%/3, 3%/2, and 2%/2mm gamma criteria, the sensitivity of the proposed PSQA framework with pretraining was 90.13%, 92.03%, and 95.84%, respectively, while the specificity was 76.01%, 76.12%, and 85.34%, respectively. The mean absolute errors (MAEs) of the proposed PSQA framework with pretraining were 1.36%, 2.37%, and 3.96%, respectively, while the root-mean-square errors (RMSEs) were 2.31%, 3.89%, and 5.17%, respectively. The results demonstrated visible improvement at multiple institutions. For clinical commissioning, the deviations between the predicted and measured results were all within 3% for 3%/3 and 3%/2mm at eight institutions. For clinical implementation, all failure plans were correctly identified by the proposed PSQA framework. The general PSQA framework enables diverse clinical data sources to be handled to achieve enhanced model performance and generalizability, and provides a solution to the unification of heterogeneous data from different institutions to construct robust QA models. This approach can be clinically deployed for VMAT QA.

Asymmetric Effects of Real Exchange Rate on Turkey's Imports: Threshold Value Regression Model

Determinants of imports are one of the most discussed topics in the foreign trade literature. This study explores the asymmetric relationship between the real exchange rate and imports for Turkey example. The paper employs monthly data spanning from 2013:01 to 2023:06, estimating classical least square estimation and Threshold Regression Models, subsequently comparing their outcomes. The findings highlighted that the relations between imports and real exchange rate are asymmetric. According to the least square estimation model, it was noted that the real exchange rate had a negative impact on imports, aligning with anticipated outcomes. In the threshold regression model, it was seen that movements in the real exchange rate positively affected imports in the model below the threshold, while exerting a negative effect in the model above the threshold. This finding was interpreted as increasing imports by buying the expectation that the exchange rate increases, which are below the internalizable level, may increase further. It is considered essential for policymakers to take into account asymmetric relationships when analyzing the relationships between imports and the exchange rate.

Defining the optimal setting for transcriptomic analyses on blood samples for response prediction in immunotherapy-treated NSCLC patients

Transcriptomic profiling of blood immune cells offers a promising alternative to invasive, sampling bias-prone tissue-based biomarker assays for predicting immune checkpoint inhibitor (ICI) therapy response in non-small cell lung cancer (NSCLC) patients. However, the optimal analytical approach to identify systemic correlates of response still needs to be explored. We collected peripheral blood mononuclear cells and whole blood (WB) samples from 33 ICI-treated NSCLC patients before ICI treatment and at the first response evaluation. After bulk polyadenylated RNA-sequencing, we assessed differences in gene expression profiles between non-responders and responders using differential expression analysis, single sample gene set enrichment analysis (ssGSEA), and cell type deconvolution. We evaluated gene expression values, ssGSEA scores, and deconvolved cell type proportions to distinguish non-responders from responders via ROC curve (AUC) analysis, training a logistic regression classification model. Gene expression values and deconvolved proportions yielded the best results with WB samples after treatment (AUC = 0.87 and 0.85, respectively). Overall, ssGSEA scores showed superior classification performance across all sample types and timepoints (AUC > 0.7). In conclusion, transcriptomic analysis through ssGSEA demonstrated the best performance as a non-invasive biomarker for predicting clinical benefit in ICI-treated NSCLC patients, with gene expression and deconvolution on post-treatment WB samples also showing promising results.

Machine learning-based personalized composite score dissects risk and protective factors for cognitive and motor function in older participants.

With age, sensory, cognitive, and motor abilities decline, and the risk for neurodegenerative disorders increases. These impairments influence the quality of life and increase the need for care, thus putting a high burden on society, the economy, and the healthcare system. Therefore, it is important to identify factors that influence healthy aging, particularly ones that are potentially modifiable through lifestyle choices. However, large-scale studies investigating the influence of multi-modal factors on a global description of healthy aging measured by multiple clinical assessments are sparse. We propose a machine learning model that simultaneously predicts multiple cognitive and motor outcome measurements on a personalized level recorded from one learned composite score. This personalized composite score is derived from a large set of multi-modal components from the TREND cohort, including genetic, biofluid, clinical, demographic, and lifestyle factors. We found that a model based on a single composite score was able to predict cognitive and motor abilities almost as well as a classical flexible regression model specifically trained for each single clinical score. In contrast to the flexible regression model, our composite score model is able to identify factors that globally influence cognitive and motoric abilities as measured by multiple clinical scores. The model identified several risk and protective factors for healthy aging and recovered physical exercise as a major, modifiable, protective factor. We conclude that our low parametric modeling approach successfully recovered known risk and protective factors of healthy aging on a personalized level while providing an interpretable composite score. We suggest validating this modeling approach in other cohorts.

Bayesian network modeling applied to food risks: Data from General Administration of Customs of China as an example

We introduce a multidimensional Bayesian Network (BN) modeling approach as an alternative to the classical multivariate regression approach commonly used for risk factor analysis. BN modeling is a rich and flexible analytical technique capable of elucidating complex food data. It is a data-driven graphical modeling technique and exploratory tool that visually presents dependencies and causal associations while retaining statistical rigor in holistic level inference. We used data on monthly unauthorized food imports to China for demonstrating a full Additive BN model in the open-source R language with all code necessary to reproduce our analyses. Compared to the classical regression model, the Additive BN model of this study described more relationships between variables, and overcome the problems of classical discretization of general BN. The analysis showed that food category and import weight were two actionable drivers directly related to risk and can be controlled to manage risk. The risk was also related to where the food is produced, but not where it is likely to be sold. The BN model also is able to systematically elaborate uncertainty and obtain information through probabilistic inference, which might be useful for giving additional new insights potentially not captured by classical methods. Finally, we discuss the potential and evolution of BN, such as dynamic BN.

Development and Internal Validation of Machine Learning to Predict Postoperative Worse Functional Status after Surgical Treatment for Thoracic Spinal Stenosis.

BACKGROUND The objective of this study was to develop and validate machine learning (ML) algorithms to predict the 30-day and 6-month risk of deteriorating functional status following surgical treatment for thoracic spinal stenosis (TSS). We aimed to provide surgeons with tools to identify patients with TSS who have a higher risk of postoperative functional decline. MATERIAL AND METHODS The records of 327 patients with TSS who completed both follow-up visits were analyzed. Our primary endpoint was the dichotomized change in the perioperative Japanese Orthopedic Association (JOA) score, categorized based on whether it deteriorated or not. The models were developed using Naïve Bays, LightGBM, XGBoost, logistic regression, and random forest classification models. The model performance was assessed by accuracy and the c-statistic. ML algorithms were trained, optimized, and tested. RESULTS The best-performing algorithms for predicting functional decline at 30 days and 6 months after TSS surgery were XGBoost (accuracy=88.17%, c-statistic=0.83) and Naïve Bays (accuracy=86.03%, c-statistic=0.80). Both algorithms presented good calibration and discrimination in our testing data. We identified several significant predictors, including poor quality of intraoperative SSEP/MEP baseline, poor quality of preoperative SSEP, duration of symptoms, operated level, and motor dysfunction of the lower extremity. CONCLUSIONS The best-performing algorithms for predicting functional decline at 30 days and 6 months after TSS surgery were XGBoost (accuracy=88.17%, c-statistic=0.83) and Naïve Bays (accuracy=86.03%, c-statistic=0.80). Both algorithms presented good calibration and discrimination in our testing data. We identified several significant predictors, including poor quality of intraoperative SSEP/MEP baseline, poor quality of preoperative SSEP, duration of symptoms, operated level, and motor dysfunction of the lower extremity.

The Impact of Prospect Theory on Asset Forecasting: A Comparative Study Based on Random Forest and Linear Regression

This study focuses on the application of predictive models and prospect theory in the field of machine learning to predict financial asset prices. Here, the linear regression and random forest algorithms were employed to forecast the price values and fluctuations of two different financial assets: the S&P 500 index and Bitcoin prices. Meanwhile, this study compared the performance of different financial asset datasets before and after the introduction of prospect theory models. Through calculation and comparative analysis, this article provides an in-depth discussion on the predictive performance of regression models and classification models. The research results indicate that in most cases, simple linear regression models have high prediction accuracy. Meanwhile, this paper also found that introducing prospect theory can effectively improve the accuracy of prediction models for specific financial assets. This result has a positive impact on the financial industry, helping to optimize risk management and investment decision-making, improving efficiency at the market level and promoting information transparency, while also promoting the development of emerging financial formats in financial innovation

Candidate Biomarkers For Response To Treatment In Psoriatic Disease.

To determine whether biologic therapy alters serum CXCL10, MMP3, S100A8, ACP5, and CCL2 levels in psoriatic arthritis (PsA) and psoriasis (PsC) patients, and whether baseline levels of these proteins predict response to treatment for PsA. We included PsA patients on TNF inhibitors (TNFi), Interleukin-17 inhibitors (IL-17i), methotrexate (MTX), and untreated with bDMARDs or csDMARDs (untreated) and PsC patients on bDMARDs and matched untreated. Serum samples at baseline and 3-6 months follow-up were retrieved from the biobank. Protein levels were quantified using a Luminex multiplex assay. We compared follow-up vs baseline protein levels within groups and change in levels between groups. For the predictive potential of the biomarkers, we developed logistic regression classification models. Response to treatment was defined as: 1. Achieving low disease activity or remission (according to DAPSA), 2. 75% reduction in PASI, and 3. 50% reduction in actively inflamed joint count. In PsA, TNFi reduced serum levels of all five proteins, IL-17i increased ACP5 and CCL2, and MTX reduced MMP3. Changes in MMP3 and S100A8 levels were significantly different between untreated PsA and matched biologic-treated PsA (p<0.05). There were no significant differences between treated or untreated PsC patients. Baseline levels of CXCL10, MMP3, S100A8, and ACP5 had good predictive value (AUC>0.8) for response to biologics in PsA patients. Biologics treatment and MTX affect serum CXCL10, MMP3, S100A8, ACP5, and CCL2 levels in PsA patients. MMP3, S100A8, ACP5, and CXCL10 have potential use as serum biomarkers to predict response to treatment for PsA.

Application of Artificial Intelligence and Machine Learning Was Not Able to Reliably Predict Poor Outcomes in People With Hemophilia.

Background Artificial intelligence (AI) and machine learning (ML) are currently used in the clinical field to improve the outcome predictions on disease diagnosis and prognosis. However, to date, few AI/ML applications have been reported in rare diseases, such as hemophilia. In this study, taking advantage of the ATHNdataset, an extensive repository of hemostasis and thrombosis data, we aimed to demonstrate the application of AI/ML approaches to build predictive models to identify persons with hemophilia (PwH) who are at risk of poor outcome and to inform providers in clinical decision-making towards helping patients prevent long-term complications. Materials and methods This project was carried out in two steps. First, the data were mined from ATHN 7, a subset study of the ATHNdataset, to determine markers that defined "poor outcome." Second, we applied multiple AI/ML approaches on the ATHNdataset to validate our findings and to develop predictive models to identify PwH at risk of poor outcomes. The classical regression-based predictive model was used as a reference to evaluate the performance of various AI/ML models. Results Our models included features similarly distributed to response variables of interest, resulting in a limited ability to distinguish poor outcomes. Low recall (<53%) resulted in no single model reliably predicting poor outcomes out of all actual positive cases. Our results suggest that, to build a more useful AI/ML model, we may need a larger dataset size along with additional features. Furthermore, our results showed that most of the AI/ML models outperformed the classical logistic regression model in both model accuracy and precision. Conclusions Our AI and ML model showed limited ability to predict poor outcomes in people with hemophilia.

Prediction of cyber-attacks in air transport using neural networks

This article addresses the topic of cyber-attacks in air transport, which is crucial for ensuring the safety and reliability of airports and air transport operations. The aim of the article was to present a new method for predicting cyber-attacks in air transport based on neural networks. The task of the neural network was to determine the multiple regression function based on which the probability of a cyberattack occurring at a specified hour and on a specific day of the week is predicted. The probability, depending on the time of the cyberattack occurrence, was determined using theoretical distributions. The method was verified with real data. Verification of the method confirmed its high effectiveness, determined at the level of 92%. The study examined the effectiveness of using the classical multiple regression method in predicting cyber-attacks in air transport. The classical multiple regression model covered only 0.14 of the input data, while the regression model generated by the neural network covered 0.99, indicating the high efficiency of the developed neural network.

On local power of likelihood-based tests in von Mises regressions

The von Mises distribution has played a central role as a distribution on the circle. Its associated circular regression model has been applied in a number of areas. In this paper, we consider the von Mises regression model and, under a sequence of Pitman alternatives, derive the nonnull asymptotic expansions of the cumulative distribution functions of the likelihood ratio, Wald, Rao score, and gradient test statistics for testing a subset of the von Mises regression parameters, as well as for testing the concentration parameter. We then compare analytically the local power of these likelihood-based tests on the basis of the asymptotic expansions and provide conditions where one test can be more locally powerful than the other one in this class of regression models. Consequently, on the basis of the general conditions established, the user can choose the most powerful test to make inferences on the model parameters. We also provide a numerical example to illustrate the usefulness and applicability of the general result.

CD34+ cell yield among healthy donors: Large-scale model development and validation.

Successful engraftment in hematopoietic stem cell transplantation necessitates the collection of an adequate dose of CD34+ cells. Thus, the precise estimation of CD34+ cells harvested via apheresis is critical. Current CD34+ cell yield prediction models have limited reproducibility. This study aims to develop a more reliable and universally applicable model by utilizing a large dataset, enhancing yield predictions, optimizing the collection process, and improving clinical outcomes. A secondary analysis was conducted using the Center for International Blood and Marrow Transplant Research database, involving data from over 17 000 healthy donors who underwent filgrastim-mobilized hematopoietic progenitor cell apheresis. Linear regression, gradient boosting regressor, and logistic regression classification models were employed to predict CD34+ cell yield. Key predictors identified include pre-apheresis CD34+ cell count, weight, age, sex, and blood volume processed. The linear regression model achieved a coefficient of determination (R2) value of 0.66 and a correlation coefficient (r) of 0.81. The gradient boosting regressor model demonstrated marginally improved results with an R2 value of 0.67 and an r value of 0.82. The logistic regression classification model achieved a predictive accuracy of 96% at the 200 × 106 CD34+ cell count threshold. At thresholds of 400, 600, 800, and 1000 × 106 CD34+ cell count, the accuracies were 88%, 83%, 83%, and 88%, respectively. The model demonstrated a high area under the receiver operator curve scores ranging from 0.90 to 0.93. This study introduces advanced predictive models for estimating CD34+ cell yield, with the logistic regression classification model demonstrating remarkable accuracy and practical utility.

Contribution to Characterizing the Meat Quality of Protected Designation of Origin Serrana and Preta de Montesinho Kids Using the Near-Infrared Reflectance Methodology.

The aims of this study were to describe and compare the meat quality characteristics of male and female kids from the "Serrana" and "Preta de Montesinho" breeds certified as "Cabrito Transmontano" and reinforce the performance of near-infrared reflectance (NIR) spectra in predicting these quality characteristics and discriminating among breeds. Samples of Longissimus thoracis (n = 32; sixteen per breed; eight males and eight females) were used. Breed significantly affected meat quality characteristics, with only color and fatty acid (FA) (C12:0) being influenced by sex. The meat of the "Serrana" breed proved to be more tender than that of the "Preta de Montesinho". However, the meat from the "Preta de Montesinho" breed showed higher intramuscular fat content and was lighter than that from the "Serrana" breed, which favors its quality of color and juiciness. The use of NIR with the linear support vector machine regression (SVMR) classification model demonstrated its capability to quantify meat quality characteristics such as pH, CIELab color, protein, moisture, ash, fat, texture, water-holding capacity, and lipid profile. Discriminant analysis was performed by dividing the sample spectra into calibration sets (75 percent) and prediction sets (25 percent) and applying the Kennard-Stone algorithm to the spectra. This resulted in 100% correct classifications with the training data and 96.7% accuracy with the test data. The test data showed acceptable estimation models with R2 > 0.99.

Housing prices and points of interest in three Polish cities

This paper shows the usefulness of open source points-of-interest (POI) data for understanding the causes of volatility in residential property prices. Points of interest are unique locations or features cartographically mapped in space and explicitly connected to various aspects of human life. This point number, and density, primarily reflect a city’s spatial structure. They could therefore indicate a higher quality of life in a given urban zone, resulting in higher demand for housing in the area and, consequently, higher housing prices. This study was conducted in three Polish cities: Warsaw, Poznań and Olsztyn. Our research also attempted to establish which POI categories constitute stimulants and which are destimulants in the housing market. To determine the significance of POI in the price formation process, we used the quantile regression model and its spatial version, apart from classical regression models. The results of our research show the importance of POIs for the formation of housing prices. Individual POI categories are of minor importance, while the simultaneous presence of many different POIs definitely has a positive impact on housing prices.

ML Model for Stock Classification

Abstract: This study examines how effectively volatility indices can forecast stock market movements by applying different machine learning techniques. It uses a dataset from Kaggle for the TCS Company, including variables such as 'Date,' 'Symbol,' 'Series,' 'Prev Close,' 'Open,' 'High,' 'Low,' 'Last,' 'Close,' 'VWAP,' 'Volume,' 'Turnover,' 'Deliverable Volume,' and '%Deliverable.' Various models were tested and evaluated based on metrics like the AUC-ROC Curve, Accuracy, Precision, Recall, F1-Score, Cross-Validation Accuracy, and the Confusion Matrix. The findings reveal that the Linear Regression (Classification model) surpasses other models in forecasting stock market directions with the highest levels of accuracy across all evaluation metrics. This highlights the potential of machine learning methods in leveraging volatility indices to accurately predict stock market trends

Face Age Prediction Based on Machine Learning

Age recognition, an essential attribute of human identity, plays a pivotal role in various social and medical applications. Due to the influences of photography equipment, lighting, and angles, age recognition based on facial images has always been a challenging topic. Fortunately, with the development of machine learning, especially the breakthroughs achieved by neural network models in image-related fields, researchers have applied machine learning to age recognition. From the perspective of research methodologies, age prediction can be regarded as either a regression or classification task. Therefore, this article analyzes from both regression and classification perspectives. Firstly, it introduces the commonly used datasets in the field of facial recognition. Then, it discusses the commonly used regression models, classification models, and optimization methods in age recognition tasks. Finally, the article summarizes the entire text and proposes future research ideas, providing references for researchers in related fields.

A Theoretical Study of the Representational Power of Weighted Randomised Univariate Regression Tree Ensembles

Univariate regression trees have representation problems for non-orthogonal regression functions. Ensembles of univariate regression trees have better representational power. In some cases, weighted ensembles have shown better performance than unweighted ensembles. In this paper, we study the properties of ensembles of regression trees by using regression classification models. We propose a theoretical framework to study the representational power of infinite-sized weighted ensembles, consisting of randomised finite-sized regression trees. We show for some datasets that the weighted ensembles may have better representational power than unweighted ensembles, but the performance is highly dependent on the weighting scheme and the properties of datasets. Our model cannot be used for all the datasets. However, for some datasets, we can accurately predict the experimental results of ensembles of regression trees.

Linear mixed-effects models and least confounded residuals in the modelling of Holstein calves’ performance

Some experimental studies are carried out considering a longitudinal feature. In view of this, the classical regression models are not able to handle with it, since the independence assumption between the observations is violated. To handle with this kind of data it was proposed the called linear mixed-effects models, where it is possible to model the response variable taking into account the correlation between the observations, and even between the response variables, when there are two or more of them in study, setting a bivariate or multivariate scenario, respectively. For the diagnosis of the linear mixed-effects models the least confounded residuals are quite recommended due to their lower bias in relation to other types of residuals, but it is not so used in the literature. Using a data set of dairy calves’ performance according to three different diets over eight weeks, the linear mixed-effects models theory under univariate and bivariate approach was applied alongside the least confounded residuals in the diagnosis of the model for both approaches. Comparing the univariate and bivariate approaches, the last one was more informative presenting lower standard errors’ values for its estimates, while the least confounded residuals was more efficient than the classical residuals present in the literature.

Code-Free Machine Learning Approach for EVO-ICL Vault Prediction: A Retrospective Two-Center Study.

Establishing a development environment for machine learning is difficult for medical researchers because learning to code is a major barrier. This study aimed to improve the accuracy of a postoperative vault value prediction model for implantable collamer lens (ICL) sizing using machine learning without coding experience. We used Orange data mining, a recently developed open-source, code-free machine learning tool. This study included eye-pair data from 294 patients from the B&VIIT Eye Center and 26 patients from Kim's Eye Hospital. The model was developed using OCULUS Pentacam data from the B&VIIT Eye Center and was internally evaluated through 10-fold cross-validation. External validation was performed using data from Kim's Eye Hospital. The machine learning model was successfully trained using the data collected without coding. The random forest showed mean absolute errors of 124.8 µm and 152.4 µm for the internal 10-fold cross-validation and the external validation, respectively. For high vault prediction (>750 µm), the random forest showed areas under the curve of 0.725 and 0.760 for the internal and external validation datasets, respectively. The developed model performed better than the classic statistical regression models and the Google no-code platform. Applying a no-code machine learning tool to our ICL implantation datasets showed a more accurate prediction of the postoperative vault than the classic regression and Google no-code models. Because of significant bias in measurements and surgery between clinics, the no-code development of a customized machine learning nomogram will improve the accuracy of ICL implantation.