Large Set Of Features Research Articles

A hybrid features fusion-based framework for classification of breast micronodules using ultrasonography

BackgroundBreast cancer is one of the leading diseases worldwide. According to estimates by the National Breast Cancer Foundation, over 42,000 women are expected to die from this disease in 2024.ObjectiveThe prognosis of breast cancer depends on the early detection of breast micronodules and the ability to distinguish benign from malignant lesions. Ultrasonography is a crucial radiological imaging technique for diagnosing the illness because it allows for biopsy and lesion characterization. The user’s level of experience and knowledge is vital since ultrasonographic diagnosis relies on the practitioner’s expertise. Furthermore, computer-aided technologies significantly contribute by potentially reducing the workload of radiologists and enhancing their expertise, especially when combined with a large patient volume in a hospital setting.MethodThis work describes the development of a hybrid CNN system for diagnosing benign and malignant breast cancer lesions. The models InceptionV3 and MobileNetV2 serve as the foundation for the hybrid framework. Features from these models are extracted and concatenated individually, resulting in a larger feature set. Finally, various classifiers are applied for the classification task.ResultsThe model achieved the best results using the softmax classifier, with an accuracy of over 95%.ConclusionComputer-aided diagnosis greatly assists radiologists and reduces their workload. Therefore, this research can serve as a foundation for other researchers to build clinical solutions.

Putting computational models of immunity to the test - an invited challenge to predict B. pertussis vaccination outcomes.

Systems vaccinology studies have been used to build computational models that predict individual vaccine responses and identify the factors contributing to differences in outcome. Comparing such models is challenging due to variability in study designs. To address this, we established a community resource to compare models predicting B. pertussis booster responses and generate experimental data for the explicit purpose of model evaluation. We here describe our second computational prediction challenge using this resource, where we benchmarked 49 algorithms from 53 scientists. We found that the most successful models stood out in their handling of nonlinearities, reducing large feature sets to representative subsets, and advanced data preprocessing. In contrast, we found that models adopted from literature that were developed to predict vaccine antibody responses in other settings performed poorly, reinforcing the need for purpose-built models. Overall, this demonstrates the value of purpose-generated datasets for rigorous and open model evaluations to identify features that improve the reliability and applicability of computational models in vaccine response prediction.

Feature selection by replicate reproducibility and non-redundancy.

A fundamental step in many analyses of high-dimensional data is dimension reduction. Two basic approaches are introduction of new synthetic coordinates and selection of extant features. Advantages of the latter include interpretability, simplicity, transferability, and modularity. A common criterion for unsupervized feature selection is variance or dynamic range. However, in practice, it can occur that high-variance features are noisy, that important features have low variance, or that variances are simply not comparable across features because they are measured in unrelated numeric scales or physical units. Moreover, users may want to include measures of signal-to-noise ratio and non-redundancy into feature selection. Here, we introduce the RNR algorithm, which selects features based on (i) the reproducibility of their signal across replicates and (ii) their non-redundancy, measured by linear dependence. It takes as input a typically large set of features measured on a collection of objects with two or more replicates per object. It returns an ordered list of features, i1,i2,…,ik, where feature i1 is the one with the highest reproducibility across replicates, i2 that with the highest reproducibility across replicates after projecting out the dimension spanned by i1, and so on. Applications to microscopy-based imaging of cells and proteomics highlight benefits of the approach. The RNR method is available via Bioconductor (Huber W, Carey VJ, Gentleman R et al. (Orchestrating high-throughput genomic analysis with bioconductor. Nat Methods 2015;12:115-21.) in the R package FeatSeekR. Its source code is also available at https://github.com/tcapraz/FeatSeekR under the GPL-3 open source license.

Detach-ROCKET: sequential feature selection for time series classification with random convolutional kernels

Time Series Classification (TSC) is essential in fields like medicine, environmental science, and finance, enabling tasks such as disease diagnosis, anomaly detection, and stock price analysis. While machine learning models like Recurrent Neural Networks and InceptionTime are successful in numerous applications, they can face scalability issues due to computational requirements. Recently, ROCKET has emerged as an efficient alternative, achieving state-of-the-art performance and simplifying training by utilizing a large number of randomly generated features from the time series data. However, many of these features are redundant or non-informative, increasing computational load and compromising generalization. Here we introduce Sequential Feature Detachment (SFD) to identify and prune non-essential features in ROCKET-based models, such as ROCKET, MiniRocket, and MultiRocket. SFD estimates feature importance using model coefficients and can handle large feature sets without complex hyperparameter tuning. Testing on the UCR archive shows that SFD can produce models with better test accuracy using only 10% of the original features. We named these pruned models Detach-ROCKET. We also present an end-to-end procedure for determining an optimal balance between the number of features and model accuracy. On the largest binary UCR dataset, Detach-ROCKET improves test accuracy by 0.6% while reducing features by 98.9%. By enabling a significant reduction in model size without sacrificing accuracy, our methodology improves computational efficiency and contributes to model interpretability. We believe that Detach-ROCKET will be a valuable tool for researchers and practitioners working with time series data, who can find a user-friendly implementation of the model at https://github.com/gon-uri/detach_rocket.

Deep learning for Bitcoin price direction prediction: models and trading strategies empirically compared

This paper applies deep learning models to predict Bitcoin price directions and the subsequent profitability of trading strategies based on these predictions. The study compares the performance of the convolutional neural network–long short-term memory (CNN–LSTM), long- and short-term time-series network, temporal convolutional network, and ARIMA (benchmark) models for predicting Bitcoin prices using on-chain data. Feature-selection methods—i.e., Boruta, genetic algorithm, and light gradient boosting machine—are applied to address the curse of dimensionality that could result from a large feature set. Results indicate that combining Boruta feature selection with the CNN–LSTM model consistently outperforms other combinations, achieving an accuracy of 82.44%. Three trading strategies and three investment positions are examined through backtesting. The long-and-short buy-and-sell investment approach generated an extraordinary annual return of 6654% when informed by higher-accuracy price-direction predictions. This study provides evidence of the potential profitability of predictive models in Bitcoin trading.

FVDB : A Deep-Learning Framework for Sparse, Large Scale, and High Performance Spatial Intelligence

We present f VDB, a novel GPU-optimized framework for deep learning on large-scale 3D data. f VDB provides a complete set of differentiable primitives to build deep learning architectures for common tasks in 3D learning such as convolution, pooling, attention, ray-tracing, meshing, etc. f VDB simultaneously provides a much larger feature set (primitives and operators) than established frameworks with no loss in efficiency: our operators match or exceed the performance of other frameworks with narrower scope. Furthermore, f VDB can process datasets with much larger footprint and spatial resolution than prior works, while providing a competitive memory footprint on small inputs. To achieve this combination of versatility and performance, f VDB relies on a single novel VDB index grid acceleration structure paired with several key innovations including GPU accelerated sparse grid construction, convolution using tensorcores, fast ray tracing kernels using a Hierarchical Digital Differential Analyzer algorithm (HDDA), and jagged tensors. Our framework is fully integrated with PyTorch enabling interoperability with existing pipelines, and we demonstrate its effectiveness on a number of representative tasks such as large-scale point-cloud segmentation, high resolution 3D generative modeling, unbounded scale Neural Radiance Fields, and large-scale point cloud reconstruction.

Calibration of Transformer-Based Models for Identifying Stress and Depression in Social Media

In today’s fast-paced world, the rates of stress and depression present a surge. People use social media for expressing their thoughts and feelings through posts. Therefore, social media provide assistance for the early detection of mental health conditions. Existing methods mainly introduce feature extraction approaches and train shallow machine learning (ML) classifiers. For addressing the need of creating a large feature set and obtaining better performance, other research studies use deep neural networks or language models based on transformers. Despite the fact that transformer-based models achieve noticeable improvements, they cannot often capture rich factual knowledge. Although there have been proposed a number of studies aiming to enhance the pretrained transformer-based models with extra information or additional modalities, no prior work has exploited these modifications for detecting stress and depression through social media. In addition, although the reliability of a machine learning (ML) model’s confidence in its predictions is critical for high-risk applications, there is no prior work taken into consideration the model calibration. To resolve the above issues, we present the first study in the task of depression and stress detection in social media, which injects extra-linguistic information in transformer-based models, namely, bidirectional encoder representations from transformers (BERT) and MentalBERT. Specifically, the proposed approach employs a multimodal adaptation gate for creating the combined embeddings, which are given as input to a BERT (or MentalBERT) model. For taking into account the model calibration, we apply label smoothing. We test our proposed approaches in three publicly available datasets and demonstrate that the integration of linguistic features into transformer-based models presents a surge in performance. Also, the usage of label smoothing contributes to both the improvement of the model’s performance and the calibration of the model. We finally perform a linguistic analysis of the posts and show differences in language between stressful and nonstressful texts, as well as depressive and nondepressive posts.

Creating Location-Based Mobile Applications for Tourism: A Virtual AR Guide for Western Macedonia

Augmented reality (AR) applications are currently used in many fields for communication and educational purposes. Tourism is also a sector where augmented reality is used for destination marketing and cultural heritage promotion. This study will focus on mobile location-based AR applications and their potential in tourism. Such applications can guide tourists to places of interest and enhance their overall experience. The aim of this paper is to present a mobile application that was created for tourists visiting the region of Western Macedonia, Greece. The application was developed in order to guide the users in the region, entertain them, and educate them about the region’s sights, cultural heritage, and other special characteristics. The paper also aims to present a large set of features that are present in the application, including various types of AR (marker-based, markerless, and location-based) in order to provide designers who wish to create AR applications for tourism with new ideas. The application was validated by a usability test, and its features were evaluated by 39 participants who completed a questionnaire with 29 Likert-scale items. This procedure revealed the level of acceptance for the application features, and valuable feedback was also received during a discussion with the participants about how the application could be upgraded in the future.

Functionality and Usability of the PCU Learning Management System for Online Learning: An Evaluation

A learning management system (LMS) is a highly adaptable “online education hub” that provides a large and indispensable set of features to efficiently support teaching and learning activities for schools. It can be used to support traditional face-toface instruction, as well as blended/hybrid and distance learning environments. According to studies, the functionality and usability of LMS are important aspects of web design to ensure quality education for students. This study is descriptive since the data gathered aims to know the extent of functionality and usability of the Bluebook, the Learning Management System used by Philippine Christian University. A total of 53 teachers responded, 29 from Manila and 24 from Dasmariñas. The findings of the study revealed the following: Course Navigation, 3.48, satisfactory; Course work, 3.33, satisfactory; Communication tools, 3.21satisfactory and Gradebook, 3.22 satisfactory. An overall weighted mean of 3.31 showed that the overall performance in terms of functionality and usability of PCU-LMS is satisfactory. As much as we wanted to give the best to our students, we aim for excellence. Since the main aim of this paper is to help teachers and institutions to continue providing care for students by providing them with the best education possible. The researcher recommended the retooling, retraining, and upskilling of all the faculty members with regard to the utilization of the Learning Management System with an emphasis on harmonizing the grading system in the LMS with the grading system used in Philippine Christian University

Analysis of a Fuzzy Based Intrusion Detection System in Wireless Ad Hoc Networks

Technology and its growth is considerably enormous. This massive growth allows the opening of new fields of application in the domain of wireless networking and mobile ad-hoc networks (MANET) is one of its kinds. Mobile ad hoc network is widely used from collaborative computing to time critical applications in indoor and outdoor environment. Mobility of ad hoc network makes very attractive in all areas of mobile applications. Nodes participating in mobile ad hoc networks are autonomous, self-configurable and act as a router as well. These types of networks are dynamic in nature and have sort life time. Dynamic nature and limitations of the wireless transmission medium make MANET unsecured and vulnerable to various attacks. It is very tough to implement security for this networks and it opens up doors for further research work on this area. There is a great scope for designing a system to identify attacks and take countermeasures to minimize it and keep the performance of the network within acceptable limits. Intrusion detection system is one of its kinds and considered as security mechanism for MANET. This thesis explores different types of intrusion detection system like misuse detection and anomaly detection for mobile ad hoc networks. Anomaly detection techniques for ad hoc networks depend on the characterization of normal behavior pattern of wireless nodes. This research work focuses on wireless node behavior based detection technique. Most of anomaly intrusion detection systems are focusing on upper layer traffic to a profile normal behavior of wireless node. This research work focus on media access control (MAC) layer and network layer of wireless node. It is inefficient to use a large feature set of MAC layer and network layer due to energy limitation in ad-hoc network. A minimal feature set from MAC layer and network layer were proposed. This research work proposed an anomaly intrusion detection system for mobile ad hoc network using fuzzy logic and weighted average method. The network performance of mobile ad hoc network with intrusion detection system was analyzed using various network parameters. Further the performance of the performance of the intrusion detection system was analyzed using detection rate and false alarms. Results show good improvement in detection rate and other performance metrics

Priming effect of individual similarity and ensemble perception in visual search and working memory.

Perceptual priming is a well-known phenomenon showing that the repetition of an object's feature can facilitate subsequent detection of that item. Although the priming effect has been rigorously studied in visual search, less is known about its effect on working memory and it is unclear whether the repetition of similar features, and furthermore, ensemble perception created by a large set of similar features, can induce priming. In this study, we investigated the priming effects of individual similarity and ensemble perception in visual search and visual working memory (VWM). We replicated the classic perceptual priming effect (Experiment 1a) and found that visual search was enhanced when the current target had a similar color to the previous target (Experiment 1b), but not when the similar color had been shown as a distractor before (Experiment 1c). However, if the target and distractors of similar colors formed ensemble perception, the search efficiency was again promoted even when the current target shared the same color with the previous distractor (Experiment 1d). For VWM, repeating the ensembles of the target- and nontarget-color subsets did not significantly affect the memory capacity, while switching the two harmed the memory fidelity but not capacity (Experiment 2). We suggest different underlying mechanisms for priming in visual search and VWM: in the former, the perception history of individual similarity and stimuli ensemble exert their effects on through the priority map, by forming a gradient distribution of attentional weights that peak at the previous target feature and diminish as stimulus diverges from the previously selected one; while in the latter, perception history of memory ensemble may influence the deployment of existing memory resources across trials, thereby affecting the memory fidelity but not its capacity.

CAGSI: A Classification Approach towards Gait Speed Identification

The last few decades have witnessed a remarkable amount of research addressing numerous challenges in the domain of human activity recognition. One popular problem in this domain has been that of gait analysis. A subproblem in this domain is to identify the speed of a mobile object through gait analysis. Apart from clinical diagnostic applications, the detection of the speed of a person is also important in remote health monitoring, tracking of the mentally incompetent, and determining proper ambulatory assistive devices for the orthopaedically impaired. Gait analysis-related problems commonly deal with large volumes of interrelated data for which machine-learning techniques have been proven effective. However, the size of the feature set used in such problems is a crucial factor. The choice of a large feature set may complicate the approach for long-term analysis. The present work addresses the problem of human walking speed classification through the machine learning approach. Data was experimentally collected with the mobile phone sensors carried by volunteers of different physiques. Only the acceleration readings along the three axes of the accelerometer are considered for further experimentation. Although walking speed is a personal trait, four classes of data have been curated, namely, slow walking, moderate walking, fast walking, and sitting. The speeds of the walks were not pre-defined so the volunteers performed the walks as per their own comfort, which enhances the challenge of distinguishing between sensor signals of varying speed. Experiments have been performed using different supervised learning algorithms with only acceleration data. The performance of the learning models has been analyzed with the help of accuracy, precision, recall, f1-score, and the ROC curve in a One-vs-Rest approach. The results demonstrate that the performance of this system for walking speed identification is comparable to state-of-the-art works. Our work has a unique perspective as it uses a primary dataset comprising only three features.

3D morphometry of De Geer Moraines and Crevasse-Squeeze Ridges: Differentiating between pushing and squeezing mechanisms from remotely sensed data

De Geer Moraines (DGM) and Crevasse-Squeeze Ridges (CSR) are important landforms that can provide useful insights regarding palaeo-glacial processes. Specifically, these landforms can provide information concerning ice-marginal dynamics, and/or subglacial processes, depending on the context in which they are formed. The extraction of 3D morphometric data from these ridges can help to elucidate their formational processes, and potentially enable landform differentiation. We develop a new Python-based ArcGIS toolbox that can automatically extract 3D morphometric data from large sample sets of linear features. The morphometry toolbox may be applied to a wide range of research disciplines that are concerned with quantifying the morphometry of any elongated landforms. This is particularly useful for DGM and CSR studies, where visual similarities can result in confusion over landform type and/or formation. Here we present a case study from southwest Finland and the Northwest Territories, Canada, whereby high-resolution 3D morphometric data is used to analyse and classify DGMs and CSRs. The results reveal key differences in morphometric properties between the landforms which enables a quantified foundation by which to differentiate them. The studied CSRs are found to be higher, wider, steeper, more symmetrical, less sinuous and more voluminous than the studied prominent DGM. In contrast, a tendency for cross-sectional asymmetry in DGM supports an origin by ice-marginal pushing, rather than basal squeeze-up into crevasses. This is further supported by CSRs being less sinuous than DGM due to them being constrained to the dimensions and planform of the (relatively straight) host crevasses, whereas DGM follow a more sinuous path related to the ice margin shape. Future work should include sedimentological and geophysical studies to constrain DGM internal architecture and formation processes. The results may then be used to validate the application of DGM for detailed ice marginal reconstructions.

Improved feature selection using a hybrid side-blotched lizard algorithm and genetic algorithm approach

<span lang="EN-US">Feature selection entails choosing the significant features among a wide collection of original features that are essential for predicting test data using a classifier. Feature selection is commonly used in various applications, such as bioinformatics, data mining, and the analysis of written texts, where the dataset contains tens or hundreds of thousands of features, making it difficult to analyze such a large feature set. Removing irrelevant features improves the predictor performance, making it more accurate and cost-effective. In this research, a novel hybrid technique is presented for feature selection that aims to enhance classification accuracy. A hybrid binary version of side-blotched lizard algorithm (SBLA) with genetic algorithm (GA), namely SBLAGA, which combines the strengths of both algorithms is proposed. We use a sigmoid function to adapt the continuous variables values into a binary one, and evaluate our proposed algorithm on twenty-three standard benchmark datasets. Average classification accuracy, average number of selected features and average fitness value were the evaluation criteria. According to the experimental results, SBLAGA demonstrated superior performance compared to SBLA and GA with regards to these criteria. We further compare SBLAGA with four wrapper feature selection methods that are widely used in the literature, and find it to be more efficient.</span>

An Instance- and Label-Based Feature Selection Method in Classification Tasks

Feature selection is crucial in classification tasks as it helps to extract relevant information while reducing redundancy. This paper presents a novel method that considers both instance and label correlation. By employing the least squares method, we calculate the linear relationship between each feature and the target variable, resulting in correlation coefficients. Features with high correlation coefficients are selected. Compared to traditional methods, our approach offers two advantages. Firstly, it effectively selects features highly correlated with the target variable from a large feature set, reducing data dimensionality and improving analysis and modeling efficiency. Secondly, our method considers label correlation between features, enhancing the accuracy of selected features and subsequent model performance. Experimental results on three datasets demonstrate the effectiveness of our method in selecting features with high correlation coefficients, leading to superior model performance. Notably, our approach achieves a minimum accuracy improvement of 3.2% for the advanced classifier, lightGBM, surpassing other feature selection methods. In summary, our proposed method, based on instance and label correlation, presents a suitable solution for classification problems.

Automatic detection of ictal activity in EEG using synchronization and chaos-based attributes

Automatic seizure onset detectors (SODs) have been proposed to alert epileptic patients when a seizure is about to happen and in turn improve their quality of life. Yet, the detectors proposed in literature are complex and difficult to implement in real-time as they utilize large feature sets with redundant and irrelevant features. Hence, the aim of this work is to propose a simple and lightweight SOD that exploits two characteristics that reflect the neuronal behavior during a seizure. Namely, the synchronization between EEG channels and the chaoticity of the EEG; synchronization was measured by the condition number while the recurrence period density entropy estimated the chaoticity of an EEG signal. A support vector machine was trained and tested on 10 patients from a scalp EEG dataset and was able to detect the considered seizures with a sensitivity of 100% and a false positives rate of 0.5 per hour. The results indicate that synchronization and chaos attributes can reflect the manifestation of seizures in EEG data and can be used to develop SODs. This work emphasizes that even a single relevant feature can produce an SOD with comparable performance to SODs that use many features.Graphical

A Deep Learning Approach for Network Intrusion Detection Using a Small Features Vector

With the growth in network usage, there has been a corresponding growth in the nefarious exploitation of this technology. A wide array of techniques is now available that can be used to deal with cyberattacks, and one of them is network intrusion detection. Artificial Intelligence (AI) and Machine Learning (ML) techniques have extensively been employed to identify network anomalies. This paper provides an effective technique to evaluate the classification performance of a deep-learning-based Feedforward Neural Network (FFNN) classifier. A small feature vector is used to detect network traffic anomalies in the UNSW-NB15 and NSL-KDD datasets. The results show that a large feature set can have redundant and unuseful features, and it requires high computation power. The proposed technique exploits a small feature vector and achieves better classification accuracy.

A Double machine learning trend model for citizen science data

Abstract Citizen and community science datasets are typically collected using flexible protocols. These protocols enable large volumes of data to be collected globally every year; however, the consequence is that these protocols typically lack the structure necessary to maintain consistent sampling across years. This can result in complex and pronounced interannual changes in the observation process, which can complicate the estimation of population trends because population changes over time are confounded with changes in the observation process. Here we describe a novel modelling approach designed to estimate spatially explicit species population trends while controlling for the interannual confounding common in citizen science data. The approach is based on Double machine learning, a statistical framework that uses machine learning (ML) methods to estimate population change and the propensity scores used to adjust for confounding discovered in the data. ML makes it possible to use large sets of features to control for confounding and to model spatial heterogeneity in trends. Additionally, we present a simulation method to identify and adjust for residual confounding missed by the propensity scores. To illustrate the approach, we estimated species trends using data from the citizen science project eBird. We used a simulation study to assess the ability of the method to estimate spatially varying trends when faced with realistic confounding and temporal correlation. Results demonstrated the ability to distinguish between spatially constant and spatially varying trends. There were low error rates on the estimated direction of population change (increasing/decreasing) at each location and high correlations on the estimated magnitude of population change. The ability to estimate spatially explicit trends while accounting for confounding inherent in citizen science data has the potential to fill important information gaps, helping to estimate population trends for species and/or regions lacking rigorous monitoring data.

O-243 Personalized pregnancy odds estimation can be obtained using AI embryo evaluation and individual patient characteristics

Abstract Study question Can EMATM (AIVF, Israel) artificial intelligence (AI) platform provide personalized success estimates based on the patient’s metadata and embryonic development? Summary answer Individual patients can be given an accurate estimation of their chances for a clinical pregnancy using AI-based embryo evaluation and patient metadata. What is known already: AI models for embryo evaluation are trained on diverse datasets using data from multiple clinics and from patients with varying ages and clinical history. Precision medicine can be attained by AI models that provide individual patients with personalized success rates per embryo transfer based on their characteristics. Study design, size, duration A large dataset (9,812 embryos) from 3 geographically diverse IVF Clinics (Israel, Spain, USA) with known clinical pregnancy (fetal heartbeat) and patient characteristics obtained from Electronic Medical Records (EMR) were used to evaluate the importance of individual features contributing to pregnancy. Participants/materials, setting, methods Machine learning models were trained to predict clinical pregnancy on a large training and feature set which combines AI scores (EMATM, AIVF) with additional information obtained from EMR systems and intrinsic features that can be derived from the embryo cohort. The importance of each one of the features for the prediction ability of the models was evaluated. Main results and the role of chance AI embryo score, maternal age and cohort features were found to be the major contributing factors for the prediction models, thus improving the accuracy of the estimates for pregnancy probability. Significant contributing factors were: cohort size, number of viable embryos, patient age, number of past treatments, BMI and EMA score. The AUC of the AI embryo score model was 0.7, when factoring in the patient metadata and the cohort features, the AUC of the combined model was 0.75. Limitations, reasons for caution This study is limited by its retrospective design. A prospective study is needed to validate the results. Wider implications of the findings This study validates the use of AI-based scores for embryo evaluation, for relative grading of the embryos within the cohort, and also to estimate the true pregnancy odds of each embryo based on individual patient features. This can provide a superior decision support tool for doctors, embryologists, and patients. Trial registration number not applicable

Making it into a successful series A funding: An analysis of Crunchbase and LinkedIn data

Startups are a key force driving economic development, and the success of these high-risk ventures can bring huge profits to venture capital firms. The ability to predict the success of startups is a major advantage for investors to outperform their competitors. In this study, we explore the potential of using publicly available LinkedIn profiles as an alternative and complementary data source to Crunchbase for predicting startup success. We provide a comprehensive review of the existing literature on the factors that influence startup success to create a large set of features for predictive modeling. We train two models for predicting startup success employing light gradient boosting that use LinkedIn data as a standalone and as a complementary data source, and compare them to baseline models based on Crunchbase data. We show that using LinkedIn as a complementary data source yields the best result with a mean area under the curve (AUC) value of 84%. We also provide a thorough analysis of what types of information contribute most to modeling startup success using the Shapley value method. Our models and analysis can be used to develop a decision support system to facilitate startup screening and the due diligence process for venture capital firms.