Learning To Rank Research Articles

Doubly Robust Estimation for Correcting Position Bias in Click Feedback for Unbiased Learning to Rank

Clicks on rankings suffer from position bias: generally items on lower ranks are less likely to be examined—and thus clicked—by users, in spite of their actual preferences between items. The prevalent approach to unbiased click-based learning-to-rank (LTR) is based on counterfactual inverse-propensity-scoring (IPS) estimation. In contrast with general reinforcement learning, counterfactual doubly robust (DR) estimation has not been applied to click-based LTR in previous literature. In this article, we introduce a novel DR estimator that is the first DR approach specifically designed for position bias. The difficulty with position bias is that the treatment—user examination—is not directly observable in click data. As a solution, our estimator uses the expected treatment per rank, instead of the actual treatment that existing DR estimators use. Our novel DR estimator has more robust unbiasedness conditions than the existing IPS approach, and in addition, provides enormous decreases in variance: our experimental results indicate it requires several orders of magnitude fewer datapoints to converge at optimal performance. For the unbiased LTR field, our DR estimator contributes both increases in state-of-the-art performance and the most robust theoretical guarantees of all known LTR estimators.

User Behavior Simulation for Search Result Re-ranking

Result ranking is one of the major concerns for Web search technologies. Most existing methodologies rank search results in descending order of relevance. To model the interactions among search results, reinforcement learning (RL algorithms have been widely adopted for ranking tasks. However, the online training of RL methods is time and resource consuming at scale. As an alternative, learning ranking policies in the simulation environment is much more feasible and efficient. In this article, we propose two different simulation environments for the offline training of the RL ranking agent: the Context-aware Click Simulator (CCS) and the Fine-grained User Behavior Simulator with GAN (UserGAN). Based on the simulation environment, we also design a User Behavior Simulation for Reinforcement Learning (UBS4RL) re-ranking framework, which consists of three modules: a feature extractor for heterogeneous search results, a user simulator for collecting simulated user feedback, and a ranking agent for generation of optimized result lists. Extensive experiments on both simulated and practical Web search datasets show that (1) the proposed user simulators can capture and simulate fine-grained user behavior patterns by training on large-scale search logs, (2) the temporal information of user searching process is a strong signal for ranking evaluation, and (3) learning ranking policies from the simulation environment can effectively improve the search ranking performance.

Descriptive Analysis of Learner-to-Learner Engagement in Online Learning Survey

COVID-19 pandemic which began in early 2020 has impacted the education landscape when many countries were forced to implement lockdown. As a result, many moved to online learning mode throughout 2020 and 2021. There are proven success of online learning delivery when designed using the theory of cognitive constructivism which focus on combination of authentic and active learning. One aspect of active learning is learner-to-learner engagement. This study investigates the survey questions on learner-to-learner engagement from 117 respondents of UiTM bachelor’s degree students by carrying out descriptive analysis. The descriptive analysis involves finding the measures of central tendency, calculation of Relative Important Index, classifying the distribution plot and determining the correlation between variables using Spearman correlation method. Results of the survey showed that peer support to finish task ranks first among while collaborative learning rank last.

LtRFT: Mitigate the Low-Rate Data Plane DDoS Attack With Learning-To-Rank Enabled Flow Tables

Software-Defined Networking (SDN) switches typically have limited ternary content addressable memory (TCAM) that caches the flow entries on the data plane. The scarcity and strong resource competitiveness of TCAM space put the flow tables at the risk of malicious Distributed Denial-of-Service (DDoS) attacks. In this paper, we propose LtRFT, a Learning-To-Rank (LtR) based scheme for mitigating the low-rate DDoS attacks targeted at flow tables. LtRFT consists of three modules: <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">monitor</i> , <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">ranker</i> , and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">mitigator</i> . <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Monitor</i> manages the flow table status and sends alerts to other modules after detecting attacks. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Ranker</i> models the attack mitigation problem as a flow entry ranking task, and ranks malicious flows with a high eviction priority using a pairwise-based LtR algorithm. The <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">mitigator</i> frees up the flow table space by deleting malicious flow entries according to the flow entry ranking sequence generated by <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">ranker</i> . We introduce LtR to network attack detection innovatively and use both classification and information retrieval metrics to describe and evaluate LtRFT. Extensive experiments were conducted to validate the effectiveness and robustness of LtRFT in detecting and mitigating the low-rate data plane DDoS attacks. LtRFT can detect malicious attack flows with an accuracy of over 96%, and can reduce the attack flow duration by 97.7% with an average extra latency of 0.5 seconds, which proves that LtRFT is practicable in SDN deployments.

Human Behavior Recognition Based on Multiscale Convolutional Neural Network

The key problem in human behavior recognition is how to build a spatiotemporal feature extraction and classification network. Aiming at the problem that the existing channel attention mechanism directly pools the global average information of each channel and ignores its local spatial information, this paper proposes two improved channel attention modules, namely the space-time (ST) interaction module of matrix operation and the depth separable convolution module, combined with the research of human behavior recognition. Combined with the superior performance of convolutional neural network (CNN) in image and video processing, a multi-scale convolutional neural network method for human behavior recognition is proposed. Firstly, the behavior video is segmented, and low rank learning is performed on each video segment to extract the corresponding Low rank behavior information, and then these Low rank behavior information are connected on the time axis to obtain the Low rank behavior information of the whole video, so as to effectively capture the behavior information in the video, avoiding tedious extraction steps and various assumptions. The ability of neural network to model human behavior can be transferred and reused in networks with different structures. According to the different characteristics of data features at different network levels, two effective feature difference measurement functions are introduced to reduce the difference between features extracted from different network structures. Experiments on several public datasets show that the proposed method has a good classification effect. The experimental results show that the method has a good accuracy in human behavior recognition. It is proved that the proposed model not only improves the recognition accuracy, but also effectively reduces the computational complexity of output weights and improves the compactness of the model structure.

Bipartite Ranking Fairness Through a Model Agnostic Ordering Adjustment.

Recently, with the applications of algorithms in various risky scenarios, algorithmic fairness has been a serious concern and received lots of interest in machine learning community. In this article, we focus on the bipartite ranking scenario, where the instances come from either the positive or negative class and the goal is to learn a ranking function that ranks positive instances higher than negative ones. We are interested in whether the learned ranking function can cause systematic disparity across different protected groups defined by sensitive attributes. While there could be a trade-off between fairness and performance, we propose a model agnostic post-processing framework xOrder for achieving fairness in bipartite ranking and maintaining the algorithm classification performance. In particular, we optimize a weighted sum of the utility as identifying an optimal warping path across different protected groups and solve it through a dynamic programming process. xOrder is compatible with various classification models and ranking fairness metrics, including supervised and unsupervised fairness metrics. In addition to binary groups, xOrder can be applied to multiple protected groups. We evaluate our proposed algorithm on four benchmark data sets and two real-world patient electronic health record repositories. xOrder consistently achieves a better balance between the algorithm utility and ranking fairness on a variety of datasets with different metrics. From the visualization of the calibrated ranking scores, xOrder mitigates the score distribution shifts of different groups compared with baselines. Moreover, additional analytical results verify that xOrder achieves a robust performance when faced with fewer samples and a bigger difference between training and testing ranking score distributions.

DRONet: effectiveness-driven drug repositioning framework using network embedding and ranking learning.

As one of the most vital methods in drug development, drug repositioning emphasizes further analysis and research of approved drugs based on the existing large amount of clinical and experimental data to identify new indications of drugs. However, the existing drug repositioning methods didn't achieve enough prediction performance, and these methods do not consider the effectiveness information of drugs, which make it difficult to obtain reliable and valuable results. In this study, we proposed a drug repositioning framework termed DRONet, which make full use of effectiveness comparative relationships (ECR) among drugs as prior information by combining network embedding and ranking learning. We utilized network embedding methods to learn the deep features of drugs from a heterogeneous drug-disease network, and constructed a high-quality drug-indication data set including effectiveness-based drug contrast relationships. The embedding features and ECR of drugs are combined effectively through a designed ranking learning model to prioritize candidate drugs. Comprehensive experiments show that DRONet has higher prediction accuracy (improving 87.4% on Hit@1 and 37.9% on mean reciprocal rank) than state of the art. The case analysis also demonstrates high reliability of predicted results, which has potential to guide clinical drug development.

LncRNA-disease association identification using graph auto-encoder and learning to rank.

Discovering the relationships between long non-coding RNAs (lncRNAs) and diseases is significant in the treatment, diagnosis and prevention of diseases. However, current identified lncRNA-disease associations are not enough because of the expensive and heavy workload of wet laboratory experiments. Therefore, it is greatly important to develop an efficient computational method for predicting potential lncRNA-disease associations. Previous methods showed that combining the prediction results of the lncRNA-disease associations predicted by different classification methods via Learning to Rank (LTR) algorithm can be effective for predicting potential lncRNA-disease associations. However, when the classification results are incorrect, the ranking results will inevitably be affected. We propose the GraLTR-LDA predictor based on biological knowledge graphs and ranking framework for predicting potential lncRNA-disease associations. Firstly, homogeneous graph and heterogeneous graph are constructed by integrating multi-source biological information. Then, GraLTR-LDA integrates graph auto-encoder and attention mechanism to extract embedded features from the constructed graphs. Finally, GraLTR-LDA incorporates the embedded features into the LTR via feature crossing statistical strategies to predict priority order of diseases associated with query lncRNAs. Experimental results demonstrate that GraLTR-LDA outperforms the other state-of-the-art predictors and can effectively detect potential lncRNA-disease associations. Availability and implementation: Datasets and source codes are available at http://bliulab.net/GraLTR-LDA.

Are Neural Ranking Models Robust?

Recently, we have witnessed the bloom of neural ranking models in the information retrieval (IR) field. So far, much effort has been devoted to developing effective neural ranking models that can generalize well on new data. There has been less attention paid to the robustness perspective. Unlike the effectiveness, which is about the average performance of a system under normal purpose, robustness cares more about the system performance in the worst case or under malicious operations instead. When a new technique enters into the real-world application, it is critical to know not only how it works in average, but also how would it behave in abnormal situations. So, we raise the question in this work: Are neural ranking models robust? To answer this question, first, we need to clarify what we refer to when we talk about the robustness of ranking models in IR. We show that robustness is actually a multi-dimensional concept and there are three ways to define it in IR: (1) the performance variance under the independent and identically distributed (I.I.D.) setting; (2) the out-of-distribution (OOD) generalizability ; and (3) the defensive ability against adversarial operations. The latter two definitions can be further specified into two different perspectives, respectively, leading to five robustness tasks in total. Based on this taxonomy, we build corresponding benchmark datasets, design empirical experiments, and systematically analyze the robustness of several representative neural ranking models against traditional probabilistic ranking models and learning-to-rank (LTR) models. The empirical results show that there is no simple answer to our question. While neural ranking models are less robust against other IR models in most cases, some of them can still win two out of five tasks. This is the first comprehensive study on the robustness of neural ranking models. We believe the way we study the robustness as well as our findings would be beneficial to the IR community. We will also release all the data and codes to facilitate the future research in this direction.

Report on the 1st Workshop on Reaching Efficiency in Neural Information Retrieval (ReNeuIR 2022) at SIGIR 2022

As Information Retrieval (IR) researchers, we not only develop algorithmic solutions to hard problems, but we also insist on a proper, multifaceted evaluation of ideas. The IR literature on the fundamental topic of retrieval and ranking, for instance, has a rich history of studying the effectiveness of indexes, retrieval algorithms, and complex machine learning rankers and, at the same time, quantifying their computational costs, from creation and training to application and inference. This is evidenced, for example, by more than a decade of research on efficient training and inference of large decision forest models in Learning to Rank (LtR). As we move towards even more complex, deep learning models in a wide range of applications, questions on efficiency have once again become relevant with renewed urgency. Indeed, efficiency is no longer limited to time- and space-efficiency; instead it has found new, challenging dimensions that stretch to resource-, sample- and energy-efficiency with ramifications for researchers, users, and the environment. As a step towards bringing together experts from industry and academia and creating a forum for a critical discussion and the promotion of efficiency in the era of Neural Information Retrieval (NIR), we held the ReNeuIR workshop on July 15 th , 2022 as a hybrid event---in person in Madrid, Spain along with online attendees---in conjunction with ACM SIGIR 2022. Recognizing the importance of this topic, approximately 80 participants answered our call and attended the workshop over three sessions. The event included a total of two keynotes and eight paper presentations, and concluded with a lively discussion where participants helped identify gaps in existing research and brainstormed future research directions. We had consensus in recognizing that efficiency is not simply latency, that a holistic, concrete definition of efficiency is needed to guide researchers and reviewers alike, and that more research is necessary in the development of efficiency-centered evaluation metrics and standard benchmark datasets, platforms, and tools. Date: 15 July, 2022. Website: https://ReNeuIR.org.

A Ranking Learning Model by K-Means Clustering Technique for Web Scraped Movie Data

Business organizations experience cut-throat competition in the e-commerce era, where a smart organization needs to come up with faster innovative ideas to enjoy competitive advantages. A smart user decides from the review information of an online product. Data-driven smart machine learning applications use real data to support immediate decision making. Web scraping technologies support supplying sufficient relevant and up-to-date well-structured data from unstructured data sources like websites. Machine learning applications generate models for in-depth data analysis and decision making. The Internet Movie Database (IMDB) is one of the largest movie databases on the internet. IMDB movie information is applied for statistical analysis, sentiment classification, genre-based clustering, and rating-based clustering with respect to movie release year, budget, etc., for repository dataset. This paper presents a novel clustering model with respect to two different rating systems of IMDB movie data. This work contributes to the three areas: (i) the “grey area” of web scraping to extract data for research purposes; (ii) statistical analysis to correlate required data fields and understanding purposes of implementation machine learning, (iii) k-means clustering is applied for movie critics rank (Metascore) and users’ star rank (Rating). Different python libraries are used for web data scraping, data analysis, data visualization, and k-means clustering application. Only 42.4% of records were accepted from the extracted dataset for research purposes after cleaning. Statistical analysis showed that votes, ratings, Metascore have a linear relationship, while random characteristics are observed for income of the movie. On the other hand, experts’ feedback (Metascore) and customers’ feedback (Rating) are negatively correlated (−0.0384) due to the biasness of additional features like genre, actors, budget, etc. Both rankings have a nonlinear relationship with the income of the movies. Six optimal clusters were selected by elbow technique and the calculated silhouette score is 0.4926 for the proposed k-means clustering model and we found that only one cluster is in the logical relationship of two rankings systems.

Using clarification questions to improve software developers’ Web search

Context: Recent research indicates that Web queries written by software developers are not very successful in retrieving relevant results, performing measurably worse compared to general purpose Web queries. Most approaches up to this point have addressed this problem with software engineering-specific automated query reformulation techniques, which work without developer involvement but are limited by the content of the original query. In other words, these techniques automatically improve the existing query but can not contribute new, previously unmentioned, concepts. Objective: In this paper, we propose a technique to guide software developers in manually improving their own Web search queries. We examine a conversational approach that follows unsuccessful queries with a clarification question aimed at eliciting additional query terms, thus providing to the developer a clear dimension along which the query could be improved. Methods: We describe a set of clarification questions derived from a corpus of software developer queries and a neural approach to recommending them for a newly issued query. Results: Our evaluation indicates that the recommendation technique is accurate, predicting a valid clarification question 80% of the time and outperforms simple baselines, as well as, state-of-the-art Learning To Rank (LTR) baselines. Conclusion: As shown in the experimental results, the described approach is capable at recommending appropriate clarification questions to software developers and considered useful by a sample of developers ranging from novices to experienced professionals.

Deep learning under mass-to-charge ratio pre-retrieval to realize electron ionization mass spectrometry library retrieval.

Gas chromatography-mass spectrometry (GC-MS) is an analytical technique widely used in materials science, biomedicine, and other fields. The target compound in the experiment is identified by searching for its mass spectrum in a large mass spectrum database using some algorithms. This work introduces the use of deep learning ranking for the identification of small molecules using low-resolution electron ionization MS. Because different spectra are often very similar, the algorithm produces wrong search results, and the search accuracy needs improvement. Due to the library's large amount of data, the algorithm sometimes requires a large amount of calculation and is very time consuming. Given these two problems, this work aims to develop a model for ranking based on mass-to-charge ratio (m/z) pre-retrieval method combined with deep learning to improve search accuracy and reduce the algorithm's computational time. The master spectral library maintained by the National Institute of Standards and Technology is used as the reference library for all the experiments, and the replicate library is used as the query library to evaluate the method's performance. Compared with non-machine learning algorithms, the combination of m/z matching pre-retrieval and deep learning significantly improves library retrieval accuracy by about 4%. Moreover, compared with the deep learning sorting algorithm that does not use the pre-retrieval process, it improves the accuracy of spectral library retrieval by about 0.1% and reduces the computational time of the algorithm by more than 2h. This method identifies compounds more efficiently and accurately than non-machine learning and deep learning algorithms without a pre-retrieval process.

Visualization Method of Key Knowledge Points of Nursing Teaching Management System Based on SOM Algorithm and Biomedical Diagnosis.

The traditional nursing teaching knowledge point recommendation algorithm based on collaborative filtering is difficult to deal with the problem of data sparsity, while the traditional recommendation algorithm based on matrix decomposition has poor scalability in dealing with high-dimensional data, and their recommendation results are only determined according to the prediction score, resulting in low recommendation accuracy. In view of this, a nursing teaching knowledge point recommendation method based on a SOM neural network and ranking factor decomposition machine is proposed. Firstly, the SOM neural network is used to cluster users based on users' academic background information, then the partial order relationship of nursing teaching knowledge points is constructed by using users' explicit and implicit web access behavior, and finally, the factor decomposition machine is used as the ranking function to classify users' academic background web access behavior, borrowing nursing teaching introduction text, and other characteristic information were modeled, and the peer-to-peer ranking learning algorithm was used to accurately recommend nursing teaching knowledge points. Experimental results show that the proposed method can effectively alleviate the problem of data sparsity and improve the accuracy and efficiency of recommendations.

Distance-based positive and unlabeled learning for ranking

Learning to rank – producing a ranked list of items specific to a query and with respect to a set of supervisory items – is a problem of general interest. The setting we consider is one in which no analytic description of what constitutes a good ranking is available. Instead, we have a collection of representations and supervisory information consisting of a (target item, interesting items set) pair. We demonstrate analytically, in simulation, and in real data examples that learning to rank via combining representations using an integer linear program is effective when the supervision is as light as “these few items are similar to your item of interest.” While this nomination task is quite general, for specificity we present our methodology from the perspective of vertex nomination in graphs. The methodology described herein is model agnostic.

Treatment Discontinuation Prediction in Patients With Diabetes Using a Ranking Model: Machine Learning Model Development

Background Treatment discontinuation (TD) is one of the major prognostic issues in diabetes care, and several models have been proposed to predict a missed appointment that may lead to TD in patients with diabetes by using binary classification models for the early detection of TD and for providing intervention support for patients. However, as binary classification models output the probability of a missed appointment occurring within a predetermined period, they are limited in their ability to estimate the magnitude of TD risk in patients with inconsistent intervals between appointments, making it difficult to prioritize patients for whom intervention support should be provided. Objective This study aimed to develop a machine-learned prediction model that can output a TD risk score defined by the length of time until TD and prioritize patients for intervention according to their TD risk. Methods This model included patients with diagnostic codes indicative of diabetes at the University of Tokyo Hospital between September 3, 2012, and May 17, 2014. The model was internally validated with patients from the same hospital from May 18, 2014, to January 29, 2016. The data used in this study included 7551 patients who visited the hospital after January 1, 2004, and had diagnostic codes indicative of diabetes. In particular, data that were recorded in the electronic medical records between September 3, 2012, and January 29, 2016, were used. The main outcome was the TD of a patient, which was defined as missing a scheduled clinical appointment and having no hospital visits within 3 times the average number of days between the visits of the patient and within 60 days. The TD risk score was calculated by using the parameters derived from the machine-learned ranking model. The prediction capacity was evaluated by using test data with the C-index for the performance of ranking patients, area under the receiver operating characteristic curve, and area under the precision-recall curve for discrimination, in addition to a calibration plot. Results The means (95% confidence limits) of the C-index, area under the receiver operating characteristic curve, and area under the precision-recall curve for the TD risk score were 0.749 (0.655, 0.823), 0.758 (0.649, 0.857), and 0.713 (0.554, 0.841), respectively. The observed and predicted probabilities were correlated with the calibration plots. Conclusions A TD risk score was developed for patients with diabetes by combining a machine-learned method with electronic medical records. The score calculation can be integrated into medical records to identify patients at high risk of TD, which would be useful in supporting diabetes care and preventing TD.

An Attention-Based Interactive Learning-to-Rank Model for Document Retrieval

The core issue of learning-to-rank (LTR) for document retrieval lies in finding an optimal ranking policy to meet the search intent of the user. The majority of proposed LTR approaches treat the ranking as a static process, employing a fixed ranking policy to immediately assign scores to documents. By contrast, ranking is not a static but an interactive process where the user continues interacting with the document retrieval system through information exchange such as search intent (e.g., rating or clicking for the retrieved items). We model the interactive ranking process (IRP), and propose an Attention-Based Interactive LTR model (AIRank) to constitute an intent-aware flexible ranking policy to gratify the user’s need. To enhance the ranking quality, the inherent relations among documents are procured by the self-attention method to contribute to an enriched user intent representation. Furthermore, we mend the policy gradient learning method to train the AIRank in the IRP. Experiments demonstrate the effectiveness of AIRank compared to the state-of-the-art methods in terms of normalized discounted cumulative gain and expected reciprocal rank.

AMRank: An adversarial Markov ranking model combining short- and long-term returns

Learning to rank (LTR) is a method of ranking search results using machine learning techniques. Currently, the reinforcement-learning-based ranking models have achieved some success in LTR task. However, these models have disadvantages like high variance gradient estimates and train inefficiency, which bring great challenges to the convergence and accuracy of the ranking model. Combining short- and long-term returns, this paper proposes AMRank, an adversarial Markov ranking model, which is based on reinforcement learning and formalizes the ranking task as a Markov decision process. To address the aforementioned weaknesses, in AMRank, we present a sequence discriminator to output a long-term return with a smaller variance and conduct single step updates, and use a document discriminator to yield a short-term return. The two discriminators are trained simultaneously before the decision is made. In the training process, the policy network is applied as a generator to sample candidate documents and get negative samples. At the beginning of the decision, the discriminator outputs the returns based on the environment state and the policy, and finally updates the parameters of the policy network using the policy gradient method. Experimental results on three LETOR benchmark datasets, OHUSMED, MQ2007 and MQ2008, demonstrate that the proposed AMRank outperforms the baseline models in document ranking task.

Learning to branch in the generation maintenance scheduling problem

To maximize the reliability index of a power system, this study modeled a generation maintenance scheduling problem that considers the network security constraints and rationality constraints of the generation maintenance practice in a power system. In view of the computational complexity of the generation maintenance scheduling model, a variable selection method based on a support vector machine (SVM) is proposed to solve the 0–1 mixed integer programming problem (MIP). The algorithm observes and collects data from the decisions made by strong branching (SB) and then learns a surrogate function that mimics the SB strategy using a support vector machine. The learned ranking function is then used for variable branching during the solution process of the model. The test case showed that the proposed variable selection algorithm — based on the features of the proposed generation maintenance scheduling problem during branch-and-bound — can increase the solution efficiency of the generation-scheduling model on the premise of guaranteed accuracy.

NICU Admission for Term Neonates in a Large Single-Center Population: A Comprehensive Assessment of Risk Factors Using a Tandem Analysis Approach.

Objective: Neonatal intensive care unit (NICU) admission among term neonates is associated with significant morbidity and mortality, as well as high healthcare costs. A comprehensive NICU admission risk assessment using an integrated statistical approach for this rare admission event may be used to build a risk calculation algorithm for this group of neonates prior to delivery. Methods: A single-center case–control retrospective study was conducted between August 2005 and December 2019, including in-hospital singleton live born neonates, born at ≥37 weeks’ gestation. Analyses included univariate and multivariable models combined with the machine learning gradient-boosting model (GBM). The primary aim of the study was to identify and quantify risk factors and causes of NICU admission of term neonates. Results: During the study period, 206,509 births were registered at the Shaare Zedek Medical Center. After applying the study exclusion criteria, 192,527 term neonates were included in the study; 5292 (2.75%) were admitted to the NICU. The NICU admission risk was significantly higher (ORs [95%CIs]) for offspring of nulliparous women (1.19 [1.07, 1.33]), those with diabetes mellitus or hypertensive complications of pregnancy (2.52 [2.09, 3.03] and 1.28 [1.02, 1.60] respectively), and for those born during the 37th week of gestation (2.99 [2.63, 3.41]; p < 0.001 for all), adjusted for congenital malformations and genetic syndromes. A GBM to predict NICU admission applied to data prior to delivery showed an area under the receiver operating characteristic curve of 0.750 (95%CI 0.743–0.757) and classified 27% as high risk and 73% as low risk. This risk stratification was significantly associated with adverse maternal and neonatal outcomes. Conclusion: The present study identified NICU admission risk factors for term neonates; along with the machine learning ranking of the risk factors, the highly predictive model may serve as a basis for individual risk calculation algorithm prior to delivery. We suggest that in the future, this type of planning of the delivery will serve different health systems, in both high- and low-resource environments, along with the NICU admission or transfer policy.