Binary Model Research Articles

Taba Binary, Multinomial, and Ordinal Regression Models: New Machine Learning Methods for Classification.

The classification methods of machine learning have been widely used in almost every discipline. A new classification method, called Taba regression, was introduced for analyzing binary, multinomial, and ordinal outcomes. To evaluate the performance of Taba regression, liver cirrhosis data obtained from a Mayo Clinic study were analyzed. The results were then compared with an artificial neural network (ANN), random forest (RF), logistic regression (LR), and probit analysis (PA). The results using cirrhosis data revealed that the Taba regression model could be a competitor to other classification models based on the true positive rate, F-score, accuracy, and area under the receiver operating characteristic curve (AUC). Taba regression can be used by researchers and practitioners as an alternative method of classification in machine learning. In conclusion, the Taba regression provided a reliable result with respect to accuracy, recall, F-score, and AUC when applied to the cirrhosis data.

Binary Transformer Based on the Alignment and Correction of Distribution.

Transformer is a powerful model widely used in artificial intelligence applications. It contains complex structures and has extremely high computational requirements that are not suitable for embedded intelligent sensors with limited computational resources. The binary quantization technology takes up less memory space and has a faster calculation speed; however, it is seldom studied for the lightweight transformer. Compared with full-precision networks, the key bottleneck lies in the distribution shift problem caused by the existing binary quantization methods. To tackle this problem, the feature distribution alignment operation in binarization is investigated. The median shift and mean restore is designed to ensure consistency between the binary feature distribution and the full-precision transformer. Then, a knowledge distillation architecture for distribution correction is developed, which has a teacher-student structure comprising a full-precision and binary transformer, to further rectify the feature distribution of the binary student network to ensure the completeness and accuracy of the data. Experimental results on the CIFAR10, CIFAR100, ImageNet-1k, and TinyImageNet datasets show the effectiveness of the proposed binary optimization model, which outperforms the previous state-of-the-art binarization mechanisms while maintaining the same computational complexity.

Development of a Predictive Model for N-Dealkylation of Amine Contaminants Based on Machine Learning Methods

Amines are widespread environmental pollutants that may pose health risks. Specifically, the N-dealkylation of amines mediated by cytochrome P450 enzymes (P450) could influence their metabolic transformation safety. However, conventional experimental and computational chemistry methods make it difficult to conduct high-throughput screening of N-dealkylation of emerging amine contaminants. Machine learning has been widely used to identify sources of environmental pollutants and predict their toxicity. However, its application in screening critical biotransformation pathways for organic pollutants has been rarely reported. In this study, we first constructed a large dataset comprising 286 emerging amine pollutants through a thorough search of databases and literature. Then, we applied four machine learning methods—random forest, gradient boosting decision tree, extreme gradient boosting, and multi-layer perceptron—to develop binary classification models for N-dealkylation. These models were based on seven carefully selected molecular descriptors that represent reactivity-fit and structural-fit. Among the predictive models, the extreme gradient boosting shows the highest prediction accuracy of 81.0%. The SlogP_VSA2 descriptor is the primary factor influencing predictions of N-dealkylation metabolism. Then an ensemble model was generated that uses a consensus strategy to integrate three different algorithms, whose performance is generally better than any single algorithm, with an accuracy rate of 86.2%. Therefore, the classification model developed in this work can provide methodological support for the high-throughput screening of N-dealkylation of amine pollutants.

An investigation into the relationships between technical collision behaviours and physical characteristics in world-class, international female rugby players

ABSTRACT This study first investigated how the probability of winning collision events is affected by technical characteristics among world-class, international female rugby union players, and second, whether enhanced performance of these technical characteristics was related to physical attributes. Carry and tackle events from 16 international matches played by a top-two world ranking team were coded according to technical characteristics and performance outcomes. Binary classification tree models revealed that carry performance was successfully predicted (p < 0.01) by combinations of the variables: carrier velocity at the line, change of direction and straightening angle, leg drive, body mass and system mass (carrier combined with assistance from team-mate(s)). Tackle performance was predicted by combinations of the variables: initial line-speed, tackle direction, tackle type, collision zone entry, body mass, system mass, arm use and leg drive. Cumulative link mixed effects models subsequently revealed that performance increases of ~2% in single-leg isometric squat, counter-movement jump, bench press, single-leg drop jump, 10 m acceleration momentum and velocity, and skinfolds and body mass; were associated with increasing and decreasing likelihoods of superior technical performance, depending on the investigated variable. These findings may increase the precision of practices, physical training and assessment methods, among elite-standard female rugby union players.

An Extension of the Unified Skew-Normal Family of Distributions and its Application to Bayesian Binary Regression

We consider the Bayesian binary regression model and we introduce a new class of distributions, the Perturbed Unified Skew-Normal ( pSUN , henceforth), which generalizes the Unified Skew-Normal ( SUN ) class. We show that the new class is conjugate to any binary regression model, provided that the link function may be expressed as a scale mixture of Gaussian CDFs. We discuss in detail the popular logit case, and we show that, when a logistic regression model is combined with a Gaussian prior, posterior summaries such as cumulants and normalizing constants can easily be obtained through the use of an importance sampling approach, opening the way to straightforward variable selection procedures. For more general prior distributions, the proposed methodology is based on a simple Gibbs sampler algorithm. We also claim that, in the p > n case, our proposal presents better performances - both in terms of mixing and accuracy - compared to the existing methods. We illustrate the performance through several simulation studies and two data analyses. Supplementary Materials for this article, including the R package pSUN , are available online.

PTSP-BERT: Predict the thermal stability of proteins using sequence-based bidirectional representations from transformer-embedded features.

Thermophilic proteins, mesophiles proteins and psychrophilic proteins have wide industrial applications, as enzymes with different optimal temperatures are often needed for different purposes. Convenient methods are needed to determine the optimal temperatures for proteins; however, laboratory methods for this purpose are time-consuming and laborious, and existing machine learning methods can only perform binary classification of thermophilic and non-thermophilic proteins, or psychrophilic and non-psychrophilic proteins. Here, we developed a deep learning model, PSTP-BERT, based on protein sequences that can directly perform Three classes identification of thermophilic, mesophilic, and psychrophilic proteins. By comparing BERT-bfd with other deep learning models using five-fold cross-validation, we found that BERT-bfd-extracted features achieved the highest accuracy under six classifiers. Furthermore, to improve the model's accuracy, we used SMOTE (synthetic minority oversampling technique) to balance the dataset and light gradient-boosting machine to rank BERT-bfd-extracted features according to their weights. We obtained the best-performing model with five-fold cross-validation accuracy of 89.59% and independent test accuracy of 85.42%. The performance of the PSTP-BERT is significantly better than that of existing models in Three classes identification task. In order to compare with previous binary classification models, we used PSTP-BERT to perform binary classification tasks of thermophilic and non-thermophilic protein, and psychrophilic and non-psychrophilic protein on an independent test set. PSTP-BERT achieved the highest accuracy on both binary classification tasks, with an accuracy of 93.33% for thermophilic protein binary classification and 88.33% for psychrophilic protein binary classification. The accuracy of the independent test of the model can reach between 89.8% and 92.9% after training and optimization of the training set with different sequence similarities, and the prediction accuracy of the new data can exceed 97%. For the convenience of future researchers to use and reference, we have uploaded source code of PSTP-BERT to GitHub.

Climate risk perception and adaptation strategies of smallholder farmers in The Gambia

Climate risk poses significant challenges to agriculture in The Gambia, especially for smallholder farmers reliant on rain-fed farming. Adaptation efforts will be ineffective unless farmers' perspectives of climate change are understood. The objectives are to examine smallholder farmers’ perceptions of climate risk, identify their adaptation strategies, and examine the factors that influence the adoption of these adaptation strategies at farm and off-level. Data were collected from 420 smallholder farmers across three regions of rural Gambia using a survey and multistage sampling technique. Binary logistic and multivariate probit models were used to analyse the determinants of farmers' decision to adopt adaptation strategies. The empirical result of the binary logistic model indicates that land tenure, access to government support, access to markets, witness to extreme weather events, and climate change perception influenced farmers' choice to adopt adaptation measures. The findings further indicate that female farmers are less likely to adopt adaptation measures. However, the multivariate probit model revealed that age, education, access to extension services, membership of farm groups, and perceived climate change are the main determinants of on-farm adaptation strategies. Off-farm strategies such as petty business are mainly influenced by being female, middle age, and primary education. The results suggest that gender-disaggregated support and targeted policies are necessary to encourage the adoption of adaptation strategies among female farmers. It is recommended that the government enhance access to credit, provide tailored support for women farmers, and improve extension services to foster effective adaptation in rural Gambia.

DCNN: a novel binary and multi-class network intrusion detection model via deep convolutional neural network

Network security become imperative in the context of our interconnected networks and everyday communications. Recently, many deep learning models have been proposed to tackle the problem of predicting intrusions and malicious activities in interconnected systems. However, they solely focus on binary classification and lack reporting on individual class performance in case of multi-class classification. Moreover, many of them are trained and tested using outdated datasets which eventually impact the overall performance. Therefore, there is a need for an efficient and accurate network intrusion detection system. In this paper, we propose a novel intelligent detection system based on convolutional neural network, namely DCNN. The proposed model can be utilized to efficiently analyze and detect attacks and intrusions in intelligent network systems (e.g., suspicious network traffic activities and policy violations). The DCNN model is applied against three benchmark datasets and compared with state-of-the-art models. Experimental results show that the proposed model improved resilience to intrusions and malicious activities for binary as well as multi-class classification, expanding its applicability across different intrusion detection scenarios. Furthermore, our DCNN model outperforms similar intrusion detection systems in terms of positive predicted value, true positive rate, F1 measure, and accuracy. The scores obtained for binary and multi-class classifications on the CICIoT2023 dataset are 99.50% and 99.25%, respectively. Additionally, for the CICIDS-2017 dataset, DCNN attains a score of 99.96% for both binary and multi-class classifications, while the CICIoMT2024 dataset attains a score of 99.98% and 99.86% for binary and multi-class classifications, respectively.

Establishing a Pharmacoinformatics Repository of Approved Medicines: A Database to Support Drug Product Development.

Advanced predictive modeling approaches have harnessed data to fuel important innovations at all stages of drug development. However, the need for a machine-readable drug product library which consolidates many aspects of formulation design and performance remains largely unmet. This study presents a scripted, reproducible approach to database curation and explores its potential to streamline oral medicine development. The Product Information files for all centrally authorized drug products containing a small molecule active ingredient were retrieved programmatically from the European Medicines Agency Web site. Text processing isolated relevant information, including the maximum clinical dose, dosage form, route of administration, excipients, and pharmacokinetic performance. Chemical and bioactivity data were integrated through automated linking to external curated databases. The capability of this database to inform oral medicine development was assessed in the context of drug-likeness evaluation, excipient selection, and prediction of oral fraction absorbed. Existing filters of drug-likeness, such as the Rule of Five, were found to poorly capture the chemical space of marketed oral drug products. Association rule learning identified frequent patterns in tablet formulation compositions that can be used to establish excipient combinations that have seen clinical success. Binary prediction models of oral fraction absorbed constructed exclusively from regulatory data achieved acceptable performance (balanced accuracytest = 0.725), demonstrating its modelability and potential for use during early stage molecule prioritization tasks. This study illustrates the impact of highly linked drug product data in accelerating clinical translation and underlines the ongoing need for accuracy and completeness of data reported in the regulatory datasphere.

Efficient Separation of Adjacent Rare Earths in One Step by Using Ion‐microporous Metal–Organic Frameworks

AbstractHigh‐purity rare earth (REEs) materials are key raw materials for the development of cutting‐edge technologies. However, due to similar physical and chemical properties, separating adjacent REEs from actual samples faces a formidable challenge. To overcome this challenge, an ion‐microporous metal–organic framework (ATZ‐BTC‐Zn; MOFs) is designed and synthesized, featuring densely packed nano‐trap pockets constructed from non‐coordinating carboxyl and amino groups. The synergistic effect of open nano‐trap pockets and suitable ions channel in ATZ‐BTC‐Zn is highly responsive to the size variation of lighter REEs. Notably, the configuration of Zn(H2O)62+ counterion in ATZ‐BTC‐Zn is similar to lanthanide hydrated ions (Ln(H2O)63+), facilitating the efficient ion exchange between them in the high‐speed ion transport channel. This accelerates the diffusion of REEs within the MOF pores, enhancing the utilization of active adsorption sites and promoting the efficient capture of adjacent REEs by nano‐traps. The binary model experiments show high separation factors for adjacent REEs (SFLa/Nd = 908, SFCe/La = 543), demonstrating efficient separation by ATZ‐BTC‐Zn in one step. This capability achieves the selective separation of adjacent REEs in tailings wastewater, providing a strategy to infer the compatibility between MOFs and REEs.

Verification of a system utilizing heartbeat-induced acoustic pulse waves for estimating the time at which bladder urine increases to a level requiring drainage among individuals with spinal cord injury

BackgroundSpinal cord injury (SCI) often leads to the loss of urinary sensation, making urination difficult. In a previous experiment involving six healthy participants, we measured heartbeat-induced acoustic pulse waves (HAPWs) at the mid-back, calculated time-series power spectra of heart rate gradients at three ultralow/very low frequencies, distinguished and formulated waveform characteristics (one characteristic for each power spectrum, nearly uniform across participants) at times of increased urine in the bladder and heightened urges to urinate, and developed an algorithm with five of these power spectra to identify when urination is needed by extracting the waveform portion (continuous timepoints) where all of the characteristics were consistent with the formulated characteristics. The objective of this study was to verify the validity of the algorithm fed with data from measured HAPW of participants with SCI and to adapt the algorithm for these individuals.MethodsIn ten participants with SCI, we measured HAPWs continuously and urine volume intermittently, and obtained scores related to urinary sensation. A Boolean output at each data point was obtained by the algorithm fed with the calculated power spectra from each participant's HAPW. Notable times included when the output was positive or when the need to urinate (= ( +)) was judged from the urine volume and urinary sensation scores. The outputs at these notable times were examined with the need to urinate and determined to be true/false. The accuracy of the algorithm was evaluated by the number of true/false-positive/negative points via the F-score with a binary classification model. We attempted to adapt the algorithm for participants with SCI.ResultsThe outputs at 13 notable times were examined, yielding seven true-positive, one false-positive, and five false-negative times, with an F-score of 0.70. The algorithm was modified by replacing three thresholds that determine the extraction condition for the slope in the power spectral waveform with new values that included all 12 true-positive points.ConclusionsWithout changing the use of ultralow/very low frequencies or significantly modifying the extraction conditions, the modified algorithm did not miss any true urination times or identify false urination times in ten participants with SCI.

129I and 236U distribution in the subpolar North Atlantic unravels water mass provenance in AR7W and A25 lines

The subpolar North Atlantic (SPNA) is crucial in the global ocean circulation system and one of the few regions where deep convection occurs. The intermediate and deep waters formed in the SPNA have long been investigated, yet their sources and pathways are not fully understood. In this study, we employ a combination of two radionuclide tracers, namely, 129I and 236U, to understand water mass provenance and mixing in the SPNA. The concentrations measured between Portugal and Greenland and across the Labrador Sea in 2020/2021 agreed with previously observed tracer distributions. The highest tracer concentrations were measured in the East Greenland Current (EGC), Denmark Strait Overflow Water (DSOW), and, to a lesser extent, in the eastward-flowing Labrador Sea Water (LSW). In contrast, waters of southern origin such as the North East Antarctic Bottom Water and North East Atlantic Central Water (ENACW) carried comparably smaller amounts of 129I. By using a binary mixing model, we estimated that the EGC contains about 29%–32% of the Polar Surface Water outflowing the Fram Strait. DSOW was mainly derived from 20% to 35% Return Atlantic Water and mixed with LSW. The Iceland Scotland Overflow Water (ISOW) evolved into North East Atlantic Deep Water in the Irminger and Labrador seas primarily by mixing with LSW and, to a lesser extent, with DSOW. The 129I and 236U binary mixing approach was less conclusive for LSW, reaching the current limitation of the model. This study suggests potential benefits and limitations of using 129I and 236U to investigate the mixing and provenance of water masses in the SPNA.

Effects of long-acting versus short-acting granulocyte colony stimulating factor after radiotherapy in gynecologic malignancies: a prospective observational cohort study

Little is known about the role of the protective effects of granulocyte colony-stimulating factor (G-CSF) in patients after radiotherapy. The aim of the present study was to explore the prophylactic effects of long-acting granulocyte colony-stimulating factor (G-CSF) on febrile neutropenia (FN) and myelosuppression in chemotherapy patients with gynecologic malignancies after pelvic radiotherapy. Patients voluntarily participated in a study group (long-acting G-CSF for all chemotherapy cycles) and a control group (short-acting G-CSF) after they were educated about G-CSF utilization. The incidences of FN and myelosuppression, as well as adverse events, were compared between the two groups. A regression model was used to determine the risk factors for FN and myelosuppression. From January 6, 2019, to August 22, 2019, 61 patients were included in the final analysis, with 286 chemotherapy cycles. There were 14 (23.0%) and 57 (77.0%) patients in the study and control groups, respectively. The study group had significantly fewer complete blood count tests, fewer outpatient clinic visits, fewer short-acting G-CSF doses, and lower incidences of FN and myelosuppression per chemotherapy cycle. According to the binary regression model, the use of long-acting G-CSF was the only factor associated with a decreased incidence of myelosuppression but not FN. The major adverse event related to G-CSF was mild bone pain. In conclusion, long-acting G-CSF may effectively reduce the incidence of FN and myelosuppression with mild adverse effects during chemotherapy after radiotherapy.Trial registrationRegistered at https://www.clinicaltrials.gov/ on January 4, 2019 (NCT03793205).

Detection of stellar wakes in the Milky Way: A deep learning approach

Due to poor observational constraints on the low-mass end of the subhalo mass function, the detection of dark matter (DM) subhalos on sub-galactic scales would provide valuable information about the nature of DM. Stellar wakes, induced by passing DM subhalos, encode information about the mass (properties) of the inducing perturber and thus serve as an indirect probe for the DM substructure within the Milky Way. Our aim is to assess the viability and performance of deep learning searches for stellar wakes in the Galactic stellar halo caused by DM subhalos of varying mass. We simulated massive objects (subhalos) moving through a homogeneous medium of DM and star particles with phase-space parameters tailored to replicate the conditions of the Galaxy at a specific distance from the Galactic centre. The simulation data was used to train deep neural networks with the purpose of inferring both the presence and mass of the moving perturber. We then investigated the performance of our deep learning models and identified the limitations of our current approach. We present an approach that allows for quantitative assessment of subhalo detectability in varying conditions of the Galactic stellar and DM halos. We find that our binary classifier is able to infer the presence of subhalos in our generated mock datasets, showing non-trivial performance down to a mass of $5 \, M_ In a multiple-hypothesis case, we are also able to discern between samples containing subhalos of different mass. By simulating datasets describing subhalo orbits at different Galactocentric distances, we tested the robustness of our binary classification model and found that it performs well with data generated from different initial physical conditions. Based on the phase-space observables available to us, we conclude that overdensity and velocity divergence are the most important features for subhalo detection performance.

Comparison of Current Techniques for Securing the Cystic Duct in Laparoscopic Cholecystectomy: A Systematic Review and Meta-Analysis.

Laparoscopic cholecystectomy is a widely performed procedure, with securing the cystic duct being a critical step to prevent bile leakage. Traditional metal clips are commonly used, but alternative methods, such as non-absorbable polymer clips, absorbable clips, sutures, and ultrasonic shears, are also utilized. This systematic review and meta-analysis evaluates the safety and efficacy of various cystic duct securing techniques. Following PRISMA guidelines, a comprehensive search was conducted across the EMBASE, MEDLINE, and Cochrane databases without date restrictions. Eligible studies compared different techniques for cystic duct closure in laparoscopic cholecystectomy, focusing on bile leakage as a primary outcome. Data extraction and synthesis were performed using a binary random-effects model. Meta-analyses were conducted for absorbable clips, sutures, hem-o-lok clips, and ultrasonic devices compared to standard metal clips. A total of 26 studies met the inclusion criteria. Absorbable polymer clips demonstrated a statistically significant reduction in postoperative bile leakage compared to metal clips (OR 0.159; 95%CI 0.031-0.818; P=0.028). In contrast, no significant differences were observed when comparing metal clips with suture ties (OR 0.459, 95%CI 0.139-1.522, P=0.203), non-absorbable clips (OR 0.166, 95%CI 0.025-1.109, P=0.064), or the clipless technique with a harmonic device (OR 1.332, 95%CI 0.425-4.169, P=0.623).

Determinants of last-mile travel mode choice under different COVID-19 alert levels: A case study of Batasan Hills, Quezon City, Philippines

The COVID-19 outbreak has led to remarkable changes in the transport sector and people’s travel behavior. The suspension of public transport leads to an increase in the number of private car users and the number of walking activities. The last mile, being one of the weakest links in the transport network, has become more challenging to manage with the imposition of different travel restrictions. Using the data collected from the households of Barangay Batasan Hills, Quezon City, Philippines, this study aimed to understand people’s travel behavior during the pandemic. Specifically, a binary logit model was used to determine the significant factors that affect the last-mile travel mode choice under different alert levels. Results showed that age during the pandemic, monthly household income, the purpose of travel, travel expense, travel time, departure time, origin, compliance with COVID-19 measures, and trip duration have significant factors in last-mile travel mode choice. In addition, risk perception on public transport was also a determinant of last-mile travel mode under alert levels 1 and 2. Analyzing travel behavior during the COVID-19 pandemic is deemed beneficial in devising strategies and interventions that will help mitigate the spread of the virus while still allowing economic activity and the movement of people to happen.

Close Contact Tracing and Risky Area Identification Using Alpha Shape Algorithm and Binary Contact Detection Model Based on Bluetooth 5.1

Close Contact Tracing and Risky Area Identification Using Alpha Shape Algorithm and Binary Contact Detection Model Based on Bluetooth 5.1

How does distraction affect cyclists' severe crashes? A hybrid CatBoost-SHAP and random parameters binary logit approach.

Cyclists are among the most vulnerable road users, increasingly subject to various sources of distraction, including the use of mobile phones and engagement in other tasks while navigating urban environments. Understanding and mitigating the impact of these distractions on cyclist safety is crucial. Despite the importance of this issue, the effect of distraction on injury severity in cycling crashes has not been extensively studied. This research analyzes four years of U.S. crash data (2019-2022) from the Crash Report Sampling System (CRSS) database, employing a hybrid framework that integrates CatBoost-based SHAP algorithm and the random parameters binary logit model with heterogeneity in means and variances (RPBL-HMV). The proposed approach confirms the significant role of cyclist distraction in crash injury severity. Subsequently, the analysis identifies several factors influencing the likelihood of severe injuries in distracted cyclist crashes. Crashes involving the front of motor vehicles, occurring in rural areas, on two-way roads, at higher speed limits, and during weekends were associated with a higher probability of severe injuries. Conversely, crashes at T-intersections, involving the side or rear of motor vehicles, where cyclists wore helmets, or during rush hour were linked to a reduced likelihood of severe injuries. Notably, interaction effects reveal nuanced patterns. For instance, while crossing roadway actions and rush hour periods individually decrease the likelihood of severe crashes, their combination increases the probability of such outcomes. The findings suggest targeted safety measures and policy interventions aimed at enhancing cyclist safety and promoting safer cycling environments by mitigating distraction-related risks.

Determination of the relationship between the numbers of Linepithema humile (Hymenoptera: Formicidae) on irrigation pipes and tree trunks to facilitate automated monitoring in Southern California citrus orchards.

Argentine ant, Linepithema humile (Mayr) (Hymenoptera: Formicidae), is a pest in southern California citrus orchards because it protects honeydew-producing hemipteran pests from natural enemies. A major impediment to controlling L. humile is estimating ant densities in orchards. Ants use irrigation lines to travel across orchard floors to reach trees infested with hemipterans. However, for making ant control decisions, it is the number of ants in trees, not on pipes that is critical. Work completed here demonstrates that the number of ants counted on pipes is highly correlated with the number of ants counted on trunks. Densities of ants counted on trunks are correlated with trunk diameter, citrus variety, and time of year and time of day counts. Six regression models, linear regression, zero-inflated Poisson regression, and zero-inflated negative binomial regression models, and each of their mixed model extensions, indicated a strong positive relationship between ant counts on irrigation pipes and ant counts on tree trunks. Mean squared prediction error and 5-fold cross-validation analyses indicated that the best performing of these 6 models was the zero-inflated Poisson mixed regression model. A binary classification model developed using support vector machine learning for ant infestation severity levels, categorized as low (<100 ants counted in 1min) or high (≥100 ants counted in minutes), predicted ant densities on trunks with 85% accuracy. These models can be used to estimate the number of ants on the trunks of citrus trees by using counts of ants made on irrigation pipes.

Machine Learning Models for Predicting Polymer Solubility in Solvents across Concentrations and Temperatures.

Artificial intelligence and machine learning have become essential tools in predicting material properties to aid in the accelerated design of new materials. Polymer solubility, critical for new formulations and solution processing, is one such property. However, current models are limited by inadequate experimental data sets that cannot capture the complexity and detail for many features contributing to polymer solubility. Here, we provide a data set for polymer solution behavior based on Crystal16 turbidity measurements that includes high quality percent transmission data for polymer solutions for a variety of polymers, solvents, concentrations and temperatures. We use this data set to train a model that predicts the experimental transmission data at many temperatures and multiple concentrations. From this, we are able to classify the polymer/solvent pairs into three solubility categories providing a level of granularity to predictions beyond prior binary classification models considering only solvent/nonsolvent classes. The inclusion of multiple concentrations, temperatures and partially soluble data expands solubility prediction capability beyond prior work into predictions more attractive for use by formulators and process designers working with industrial polymer solutions.