Regression Tasks Research Articles

Learning subjective time-series data via Utopia Label Distribution Approximation

Subjective time-series regression (STR) tasks have gained increasing attention recently. However, most existing methods overlook the label distribution bias in STR data, which results in biased models. Emerging studies on imbalanced regression tasks, such as age and depth estimations, hypothesize the label distribution is uniform and known. But in reality, the label distribution of test set in STR tasks is usually non-uniform and unknown. Moreover, the time-series data exhibits continuity in both temporal context and label spaces, which has not been addressed by existing methods. To tackle these issues, we propose a Utopia Label Distribution Approximation (ULDA) method, which approximates the training label distribution to the real-world but unknown (utopia) label distribution for calibrating the training and test sets. The utopia label distribution is generated by convolving the original one using a Gaussian kernel. ULDA also has two new devised modules (Time-slice Normal Sampling (TNS) generating required new samples and Convolution Weighted Loss (CWL) lowering learning weights for redundant samples), which not only assist the model training, but also maintain the sample continuity in temporal context space. Extensive experiments demonstrate that ULDA lifts the state-of-the-art performance on STR tasks and shows a considerable generalization ability to other time-series tasks.

Multi-task metric learning for optical performance monitoring

In our experiments, applying few shot metric learning for optical performance monitoring (OPM), we set the dataset as 16-way-6-shot. Modulation format identification (MFI) was utilized as a classification task, and optical signal-to-noise ratio (OSNR) estimation was used as a regression task for joint analysis. Multi-task metric learning (MML) used the adaptive weights to balance the weights of the three metric functions, six modulation formats (QPSK, 8QAM, 16QAM, 32QAM, 64QAM, 128QAM) are classified correctly with 100 % accuracy after 200 epochs. Furthermore, the lowest mean square error (MSE) of OSNR is 0.431 dB. Then, Ablation experiments compute the corresponding similarity (SIM) for each metric function show that the MSE of MML, SIMLocal+Cosine, SIMCosine+Point, SIMLocal+Point, single-task metric learning (SML) and adaptive multi-task learning (AMTL) is 0.431 dB, 0.572 dB, 0.569 dB, 0.567 dB, 0.637 dB, 1.319 dB, respectively. The proposed model achieves the highest accuracy in MFI and the lowest MSE in OSNR estimation. Finally, when comparing the various metric functions while altering the transmission distance of the optical fiber, it was observed that MML stayed within an acceptable range between 200 km and 800 km. This shows that our algorithm requires only a small number of training samples to create a reasonably good model, offering a new approach to solving problems that arise in optical performance monitoring.

Estimation of Anthocyanins in Heterogeneous and Homogeneous Bean Landraces Using Probabilistic Colorimetric Representation with a Neuroevolutionary Approach

The concentration of anthocyanins in common beans indicates their nutritional value. Understanding this concentration makes it possible to identify the functional compounds present. Previous studies have presented color characterization as two-dimensional histograms, based on the probability mass function. In this work, we proposed a new type of color characterization represented by three two-dimensional histograms that consider chromaticity and luminosity channels in order to verify the robustness of the information. Using a neuroevolutionary approach, we also found a convolutional neural network (CNN) for the regression task. The results demonstrate that using three two-dimensional histograms increases the accuracy compared to the color characterization represented by one two-dimensional histogram. As a result, the precision was 93.00 ± 5.26 for the HSI color space and 94.30 ± 8.61 for CIE L*a*b*. Our procedure is suitable for estimating anthocyanins in homogeneous and heterogeneous colored bean landraces.

Optimizing Mean Fragment Size Prediction in Rock Blasting: A Synergistic Approach Combining Clustering, Hyperparameter Tuning, and Data Augmentation

Accurate estimation of the mean fragment size is crucial for optimizing open-pit mining operations. This study presents an approach that combines clustering, hyperparameter optimization, and data augmentation to enhance prediction accuracy using the Xtreme Gradient Boosting (XGBoost) regression model. A dataset of 110 blasts was divided into 97 blasts for training and testing, whereas a separate set of 13 new, unseen blasts was used to evaluate the robustness and generalization of the model. Hierarchical Agglomerative (HA) and K-means clustering algorithms were used, with HA clustering providing a higher cluster quality. To address class imbalance and improve model generalization, a synthetic minority oversampling technique for regression with Gaussian noise (SMOGN) was employed. Hyperparameter tuning was conducted using HyperOpt by comparing Random Search (RS) with the Advanced Tree-structured Parzen Estimator (ATPE). The combination of ATPE with HA clustering and SMOGN in an expanded search space produced the best results, achieving superior prediction accuracy and reliability. The proposed HAC1-SMOGN model, which integrates HA clustering, ATPE tuning, and SMOGN augmentation, achieved a mean squared error (MSE) of 0.0002 and an R2 of 0.98 on the test set. This study highlights the synergistic benefits of clustering, hyperparameter optimization, and data augmentation in enhancing machine learning models for regression tasks, particularly in scenarios with class imbalance or limited data.

Universal approximation theorem for vector- and hypercomplex-valued neural networks

The universal approximation theorem states that a neural network with one hidden layer can approximate continuous functions on compact sets with any desired precision. This theorem supports using neural networks for various applications, including regression and classification tasks. Furthermore, it is valid for real-valued neural networks and some hypercomplex-valued neural networks such as complex-, quaternion-, tessarine-, and Clifford-valued neural networks. However, hypercomplex-valued neural networks are a type of vector-valued neural network defined on an algebra with additional algebraic or geometric properties. This paper extends the universal approximation theorem for a wide range of vector-valued neural networks, including hypercomplex-valued models as particular instances. Precisely, we introduce the concept of non-degenerate algebra and state the universal approximation theorem for neural networks defined on such algebras.

Airport time profile construction driven by flight delay prediction

Slot structure is the equilibrium result of market demand side and slot resource supply side, while slot parameters reflect the operational support capacity of the aviation system. Time parameters reflect the operational support capability of the aviation system. Time structure should not only reflect changes in market demand, but also meet the constraints of operational efficiency. Constructing a reasonable 18–24 h timetable profile for busy airports that meets normal expectations for declared capacity and seasonal scheduling is a challenge in civil aviation slot management. This study utilizes historical data on airport flights and weather conditions to establish a regression prediction model for the time structure using K-means clustering and partial least squares regression. Additionally, ensemble learning is employed to forecast flight delay levels. The findings demonstrate that random forest yields favorable results in regression and prediction tasks, allowing for the integration of upper (good weather) and lower (severse weather) limits of the time profile with delay predictions as time parameter intervals. Consequently, the flights falling within these intervals achieve an average delay level of less than 15 min which meets the expectations of normal flight.

Dynamic Feature Focusing Network for small object detection

Deep learning has driven research in object detection and achieved proud results. Despite its significant advancements in object detection, small object detection still struggles with low recognition rates and inaccurate positioning, primarily attributable to their miniature size. The location deviation of small objects induces severe feature misalignment, and the disequilibrium between classification and regression tasks hinders accurate recognition. To address these issues, we propose a Dynamic Feature Focusing Network (DFFN), which contains a duo of crucial modules: Visual Perception Enhancement Module (VPEM) and Task Association Module (TAM). Drawing upon the deformable convolution and attention mechanism, the VPEM concentrates on sparse key features and perceives the misalignment via positional offset. We aggregate multi-level features at identical spatial locations via layer average operation for learning a more discriminative representation. Incorporating class alignment and bounding box alignment parts, the TAM promotes classification ability, refines bounding box regression, and facilitates the joint learning of classification and localization. We conduct diverse experiments, and the proposed method considerably enhances the small object detection performance on four benchmark datasets of MS COCO, VisDrone, VOC, and TinyPerson. Our method has improved by 3.4 and 2.2 in mAP and APs, making solid improvements on COCO. Compared to other classic detection models, DFFN exhibits a high level of competitiveness in precision.

Average Voltage Prediction of Battery Electrode Materials Using Transformer Deep Learning Architecture

Batteries are ubiquitous, with their presence ranging from electric vehicles to portable electronics. Research focused on increasing average voltage, improving stability, and extending cycle longevity of batteries is pivotal for the advancement of battery technology. These advancements can be accelerated through research into battery chemistries. The traditional approach, which examines each material combination individually, poses significant challenges in terms of resources and financial investment. Physics-based simulations, while detailed, are both time-consuming and resource-intensive. Researchers are aiming to mitigate these concerns by employing Machine Learning (ML) techniques. We propose the use of a Transformer ML model for predicting the average voltage of battery electrode materials. A Transformer is a Deep Learning (DL) architecture based on the attention mechanism; it dynamically focuses on different parts of the input sequence, thus capturing complex relationships and dependencies among input features. The absence of recurrent units allows the model to require less training time compared to other DL models. We have adapted the Transformer model for tabular data and regression tasks. Our dataset was sourced from the Materials Project Database. To ensure the model's compatibility with our voltage prediction regression task, we successfully trained and tested it using four features across 99 samples. Our ongoing research focuses on incorporating more relevant features and expanding the training and testing of the model to the entire dataset. Our work demonstrates the effectiveness of ML, especially the Transformer model, in predicting the average voltage of battery electrode materials.

An ensemble learning-based prediction model for the compressive strength degradation of concrete containing superabsorbent polymers (SAP)

Super absorbent polymer (SAP) has a capacity to enhance the characteristics of cementitious composites in both their fresh and hardened forms. However, it is essential to recognize that the strength of SAP concrete may decrease. By altering the concrete composition and selecting the appropriate type of SAP, it is possible to reduce this reduction. This work employs machine learning (ML) to tackle the issue of strength degradation. The analysis considers ten distinct variables linked to concrete composition and the type of SAP. The study uses machine learning approaches that involve both regression and classification tasks. The use of ensemble learning greatly improves the quality and accuracy of the results, showing its superiority in combining several models to produce more precise predictions. The findings demonstrate that the Support Vector Machines (SVM) and Extreme Gradient Boosting (XGBoost) regression algorithms accurately forecasted the percentage of reduction in strength in SAP concrete. These predictions were based on the concrete composition and SAP details, resulting in R2 values of 0.90 and 0.88, respectively. Furthermore, XGBoost exhibited the highest accuracy, reaching 0.94, when compared to the various categorization algorithms. According to the results, the mean squared error (MSE) of the ensemble model demonstrated superior outcomes. Furthermore, the SHapley Additive exPlanations (SHAP) reveal that some variables, including SAP%, SAP size, and compressive strength, have a significant influence on the strength reduction model. This study aims to bridge the gap between academic research and practical application by developing a web application that employs ensemble learning to precisely forecast the reduction in compressive strength caused by the usage of SAP.

Denoising Drug Discovery Data for Improved Absorption, Distribution, Metabolism, Excretion, and Toxicity Property Prediction.

Predicting absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties of small molecules is a key task in drug discovery. A major challenge in building better ADMET models is the experimental error inherent in the data. Furthermore, ADMET predictors are typically regression tasks due to the continuous nature of the data, which makes it difficult to apply existing denoising methods from other domains as they largely focus on classification tasks. Here, we develop denoising schemes based on deep learning to address this. We find that the training error (TE) can be used to identify the noise in regression tasks while ensemble-based and forgotten event-based metrics fail to detect the noise. The most significant performance increase occurs when the original model is finetuned with the denoised data using TE as the noise detection metric. Our method has the ability to improve models with medium noise and does not degrade the performance of models with noise outside this range (low noise and high noise regimes). To our knowledge, our denoising scheme is the first to improve model performance for ADMET data and has implications for improving models for experimental assay data in general.

Active learning approaches in molecule pKi prediction.

During the early stages of drug design, identifying compounds with suitable bioactivities is crucial. Given the vast array of potential drug databases, it's feasible to assay only a limited subset of candidates. The optimal method for selecting the candidates, aiming to minimize the overall number of assays, involves an active learning (AL) approach. In this work, we benchmarked a range of AL strategies with two main objectives: (1) to identify a strategy that ensures high model performance and (2) to select molecules with desired properties using minimal assays. To evaluate the different AL strategies, we employed the simulated AL workflow based on "virtual" experiments. These experiments leveraged ChEMBL datasets, which come with known biological activity values for the molecules. Furthermore, for classification tasks, we proposed the hybrid selection strategy that unified both exploration and exploitation AL strategies into a single acquisition function, defined by parameters n and c. We have also shown that popular minimal margin and maximal variance selection approaches for exploration selection correspond to minimization of the hybrid acquisition function with n=1 and 2 respectively. The balance between the exploration and exploitation strategies can be adjusted using a coefficient (c), making the optimal strategy selection straightforward. The primary strength of the hybrid selection method lies in its adaptability; it offers the flexibility to adjust the criteria for molecule selection based on the specific task by modifying the value of the contribution coefficient. Our analysis revealed that, in regression tasks, AL strategies didn't succeed at ensuring high model performance, however, they were successful in selecting molecules with desired properties using minimal number of tests. In analogous experiments in classification tasks, exploration strategy and the hybrid selection function with a constant c<1 (for n=1) and c≤0.2 (for n=2) were effective in achieving the goal of constructing a high-performance predictive model using minimal data. When searching for molecules with desired properties, exploitation, and the hybrid function with c≥1 (n=1) and c≥0.7 (n=2) demonstrated efficiency identifying molecules in fewer iterations compared to random selection method. Notably, when the hybrid function was set to an intermediate coefficient value (c=0.7), it successfully addressed both tasks simultaneously.

Influence Maximization Based on Adaptive Graph Convolution Neural Network in Social Networks

The influence maximization problem is a hot issue in the research on social networks due to its wide application. The problem aims to find a small subset of influential nodes to maximize the influence spread. To tackle the challenge of striking a balance between efficiency and effectiveness in traditional influence maximization algorithms, deep learning-based influence maximization algorithms have been introduced and have achieved advancement. However, these algorithms still encounter two key problems: (1) Traditional deep learning models are not well-equipped to capture the latent topological information of networks with varying sizes and structures. (2) Many deep learning-based methods use the influence spread of individual nodes as labels to train a model, which can result in an overlap of influence among the seed nodes selected by the model. In this paper, we reframe the influence maximization problem as a regression task and introduce an innovative approach to influence maximization. The method adopts an adaptive graph convolution neural network which can explore the latent topology information of the network and can greatly improve the performance of the algorithm. In our approach, firstly, we integrate several network-level attributes and some centrality metrics into a vector as the presentation vector of nodes in the social network. Next, we propose a new label generation method to measure the influence of nodes by neighborhood discount strategy, which takes full account of the influence overlapping problem. Subsequently, labels and presentation vectors are fed into an adaptive graph convolution neural network model. Finally, we use the well-trained model to predict the importance of nodes and select top-K nodes as a seed set. Abundant experiments conducted on various real-world datasets have confirmed that the performance of our proposed algorithm surpasses that of several current state-of-the-art algorithms.

When Do Quantum Mechanical Descriptors Help Graph Neural Networks to Predict Chemical Properties?

Deep graph neural networks are extensively utilized to predict chemical reactivity and molecular properties. However, because of the complexity of chemical space, such models often have difficulty extrapolating beyond the chemistry contained in the training set. Augmenting the model with quantum mechanical (QM) descriptors is anticipated to improve its generalizability. However, obtaining QM descriptors often requires CPU-intensive computational chemistry calculations. To identify when QM descriptors help graph neural networks predict chemical properties, we conduct a systematic investigation of the impact of atom, bond, and molecular QM descriptors on the performance of directed message passing neural networks (D-MPNNs) for predicting 16 molecular properties. The analysis surveys computational and experimental targets, as well as classification and regression tasks, and varied data set sizes from several hundred to hundreds of thousands of data points. Our results indicate that QM descriptors are mostly beneficial for D-MPNN performance on small data sets, provided that the descriptors correlate well with the targets and can be readily computed with high accuracy. Otherwise, using QM descriptors can add cost without benefit or even introduce unwanted noise that can degrade model performance. Strategic integration of QM descriptors with D-MPNN unlocks potential for physics-informed, data-efficient modeling with some interpretability that can streamline de novo drug and material designs. To facilitate the use of QM descriptors in machine learning workflows for chemistry, we provide a set of guidelines regarding when and how to best leverage QM descriptors, a high-throughput workflow to compute them, and an enhancement to Chemprop, a widely adopted open-source D-MPNN implementation for chemical property prediction.

Machine-Learning-Assisted Design of Novel TiZrNbVAl Refractory High-Entropy Alloys with Enhanced Ductility

Refractory high-entropy alloys (RHEAs) typically exhibit excellent high-temperature strength but limited ductility. In this study, a comprehensive machine learning strategy with integrated material knowledge is proposed to predict the elongation of TiZrNbVAl RHEAs. By referring to the ductility theories, a set of cost-effective material features is developed with various mathematical forms of thermodynamic parameters. These features are proven to effectively incorporate material knowledge into ML modeling. They also offer potential alternatives to those obtained from costly first-principles calculations. Based on Pearson correlation coefficients, the linear relationships between pairwise features were compared, and the seven key features with the greatest impact on the model were selected for ML modeling. Regression tasks were performed to predict the ductility of TiZrNbVAl, and the CatBoost gradient boosting algorithm exhibiting the best performance was eventually selected. The established optimized model achieves high predictive accuracies exceeding 0.8. These key features were further analyzed using interpretable ML methods to elucidate their influences on various ductility mechanisms. According to the ML results, different compositions of TiZrNbVAl with excellent tensile properties were prepared. The experimental results indicate that Ti44Zr24Nb17V5Al10 and Ti44Zr26Nb8V13Al9 both exhibited ultimate tensile strengths of approximately 1180 MPa and elongations higher than 21%. They verified that the ML strategy proposed in this study is an effective approach for predicting the properties of RHEAs. It is a potential method that can replace costly first-principles calculations. Thermodynamic parameters have been shown to effectively predict alloy ductility to a certain extent.

Prediction of segmental motor outcomes in traumatic spinal cord injury: Advances beyond sum scores

Background and objectivesNeurological and functional recovery after traumatic spinal cord injury (SCI) is highly challenged by the level of the lesion and the high heterogeneity in severity (different degrees of in/complete SCI) and spinal cord syndromes (hemi-, ant-, central-, and posterior cord). So far outcome predictions in clinical trials are limited in targeting sum motor scores of the upper (UEMS) and lower limb (LEMS) while neglecting that the distribution of motor function is essential for functional outcomes. The development of data-driven prediction models of detailed segmental motor recovery for all spinal segments from the level of lesion towards the lowest motor segments will improve the design of rehabilitation programs and the sensitivity of clinical trials. MethodsThis study used acute-phase International Standards for Neurological Classification of SCI exams to forecast 6-month recovery of segmental motor scores as the primary evaluation endpoint. Secondary endpoints included severity grade improvement, independent walking, and self-care ability. Different similarity metrics were explored for k-nearest neighbor (kNN) matching within 1267 patients from the European Multicenter Study about Spinal Cord Injury before validation in 411 patients from the Sygen trial. The kNN performance was compared to linear and logistic regression models. ResultsWe obtained a population-wide root-mean-squared error (RMSE) in motor score sequence of 0.76(0.14, 2.77) and competitive functional score predictions (AUCwalker = 0.92, AUCself-carer = 0.83) for the kNN algorithm, improving beyond the linear regression task (RMSElinear = 0.98(0.22, 2.57)). The validation cohort showed comparable results (RMSE = 0.75(0.13, 2.57), AUCwalker = 0.92). We deploy the final historic control model as a web tool for easy user interaction (https://hicsci.ethz.ch/). DiscussionOur approach is the first to provide predictions across all motor segments independent of the level and severity of SCI. We provide a machine learning concept that is highly interpretable, i.e. the prediction formation process is transparent, that has been validated across European and American data sets, and provides reliable and validated algorithms to incorporate external control data to increase sensitivity and feasibility of multinational clinical trials.

Feature decomposition-based gaze estimation with auxiliary head pose regression

Recognition and understanding of facial images or eye images are critical for eye tracking. Recent studies have shown that the simultaneous use of facial and eye images can effectively lower gaze errors. However, these methods typically consider facial and eye images as two unrelated inputs, without taking into account their distinct representational abilities at the feature level. Additionally, implicitly learned head pose from highly coupled facial features would make the trained model less interpretable and prone to the gaze-head overfitting problem. To address these issues, we propose a method that aims to enhance task-relevant features while suppressing other noises by leveraging feature decomposition. We disentangle eye-related features from the facial image via a projection module and further make them distinctive with an attention-based head pose regression task, which could enhance the representation of gaze-related features and make the model less susceptible to task-irrelevant features. After that, the mutually separated eye features and head pose are recombined to achieve more accurate gaze estimation. Experimental results demonstrate that our method achieves state-of-the-art performance, with an estimation error of 3.90° on the MPIIGaze dataset and 5.15° error on the EyeDiap dataset, respectively.

Phenotype prediction using biologically interpretable neural networks on multi-cohort multi-omics data

Integrating multi-omics data into predictive models has the potential to enhance accuracy, which is essential for precision medicine. In this study, we developed interpretable predictive models for multi-omics data by employing neural networks informed by prior biological knowledge, referred to as visible networks. These neural networks offer insights into the decision-making process and can unveil novel perspectives on the underlying biological mechanisms associated with traits and complex diseases. We tested the performance, interpretability and generalizability for inferring smoking status, subject age and LDL levels using genome-wide RNA expression and CpG methylation data from the blood of the BIOS consortium (four population cohorts, Ntotal = 2940). In a cohort-wise cross-validation setting, the consistency of the diagnostic performance and interpretation was assessed. Performance was consistently high for predicting smoking status with an overall mean AUC of 0.95 (95% CI: 0.90–1.00) and interpretation revealed the involvement of well-replicated genes such as AHRR, GPR15 and LRRN3. LDL-level predictions were only generalized in a single cohort with an R2 of 0.07 (95% CI: 0.05–0.08). Age was inferred with a mean error of 5.16 (95% CI: 3.97–6.35) years with the genes COL11A2, AFAP1, OTUD7A, PTPRN2, ADARB2 and CD34 consistently predictive. For both regression tasks, we found that using multi-omics networks improved performance, stability and generalizability compared to interpretable single omic networks. We believe that visible neural networks have great potential for multi-omics analysis; they combine multi-omic data elegantly, are interpretable, and generalize well to data from different cohorts.

Variational data encoding and correlations in quantum-enhanced machine learning

Abstract Leveraging the extraordinary phenomena of quantum superposition and quantum correlation, quantum computing offers unprecedented potential for addressing challenges beyond the reach of classical computers. This paper tackles two pivotal challenges in the realm of quantum computing: firstly, the development of an effective encoding protocol for translating classical data into quantum states, a critical step for any quantum computation. Different encoding strategies can significantly influence quantum computer performance. Secondly, we address the need to counteract the inevitable noise that can hinder quantum acceleration. Our primary contribution is the introduction of a novel variational data encoding method, grounded in quantum regression algorithm models. By adapting the learning concept from machine learning, we render data encoding a learnable process. This allowed us to study the role of quantum correlation in data encoding. Through numerical simulations of various regression tasks, we demonstrate the efficacy of our variational data encoding, particularly post-learning from instructional data. Moreover, we delve into the role of quantum correlation in enhancing task performance, especially in noisy environments. Our findings underscore the critical role of quantum correlation in not only bolstering performance but also in mitigating noise interference, thus advancing the frontier of quantum computing.

Short-term load forecasting of mixed industrial users in distribution networks using DARTS-LSTM method

Abstract Load forecasting has perennially stood as a fundamental pillar within the power industry, with precision in prognostication constituting a paramount necessity for critical functions including grid management, optimization of energy storage, and response to demand fluctuations. In the present study, we introduce a comprehensive global modeling prediction framework aimed at forecasting large-scale power demand within time series data, alongside the development of a novel short-term power load forecasting model. By integrating all time series data pertaining to the same regression task and employing a unified univariate prediction function across the entire time series dataset, we efficaciously forecast large-scale heterogeneous loads aggregated at various levels within distribution networks. The proposed framework leverages the cutting-edge DARTS (Differentiable Architecture Search) deep learning framework for implementation. Through comprehensive experimentation on real-world data, our model framework demonstrates superior performance in terms of overall prediction accuracy, scalability, and robustness in handling missing and non-stationary time series data.

Towards Better Modeling With Missing Data: A Contrastive Learning-Based Visual Analytics Perspective.

Missing data can pose a challenge for machine learning (ML) modeling. To address this, current approaches are categorized into feature imputation and label prediction and are primarily focused on handling missing data to enhance ML performance. These approaches rely on the observed data to estimate the missing values and therefore encounter three main shortcomings in imputation, including the need for different imputation methods for various missing data mechanisms, heavy dependence on the assumption of data distribution, and potential introduction of bias. This study proposes a Contrastive Learning (CL) framework to model observed data with missing values, where the ML model learns the similarity between an incomplete sample and its complete counterpart and the dissimilarity between other samples. Our proposed approach demonstrates the advantages of CL without requiring any imputation. To enhance interpretability, we introduce CIVis, a visual analytics system that incorporates interpretable techniques to visualize the learning process and diagnose the model status. Users can leverage their domain knowledge through interactive sampling to identify negative and positive pairs in CL. The output of CIVis is an optimized model that takes specified features and predicts downstream tasks. We provide two usage scenarios in regression and classification tasks and conduct quantitative experiments, expert interviews, and a qualitative user study to demonstrate the effectiveness of our approach. In short, this study offers a valuable contribution to addressing the challenges associated with ML modeling in the presence of missing data by providing a practical solution that achieves high predictive accuracy and model interpretability.