Pairs Of Items Research Articles

On the Relationship between Explanation and Recommendation: Learning to Rank Explanations for Improved Performance

Explaining to users why some items are recommended is critical, as it can help users to make better decisions, increase their satisfaction, and gain their trust in recommender systems (RS). However, existing explainable RS usually consider explanation as a side output of the recommendation model, which has two problems: (1) It is difficult to evaluate the produced explanations, because they are usually model-dependent, and (2) as a result, how the explanations impact the recommendation performance is less investigated. In this article, explaining recommendations is formulated as a ranking task and learned from data, similarly to item ranking for recommendation. This makes it possible for standard evaluation of explanations via ranking metrics (e.g., Normalized Discounted Cumulative Gain). Furthermore, this article extends traditional item ranking to an item–explanation joint-ranking formalization to study if purposely selecting explanations could reach certain learning goals, e.g., improving recommendation performance. A great challenge, however, is that the sparsity issue in the user-item-explanation data would be inevitably severer than that in traditional user–item interaction data, since not every user–item pair can be associated with all explanations. To mitigate this issue, this article proposes to perform two sets of matrix factorization by considering the ternary relationship as two groups of binary relationships. Experiments on three large datasets verify the solution’s effectiveness on both explanation ranking and item recommendation.

Speech intelligibility of English spoken by Indonesian and Taiwanese speakers and judged by Taiwanese listeners: application of TOCS+ automatic software

Former investigations were about the familiarity advantage that people with the same language backgrounds are more intelligible. Besides, particular English was most intelligible to participants with high certain English familiarity. This study investigated how Taiwanese listeners judge the English speech intelligibility of Taiwanese and Indonesian speakers. Thirty Indonesian speakers and thirty Taiwanese speakers participated in this study. Ninety Taiwanese listeners were recruited to judge speech intelligibility. The recording and judging process used the TOCS+ software (Hodge et al., 2009). The software provided 124 contrast items of minimal pairs for the recognition task, including contrast of syllable shape, vowels, and consonants. The listeners' judgments were then analyzed by the TOCS+ software automatically. It was found that Taiwanese English speaker is more intelligible to Taiwanese listeners. The variables that predict intelligibility for Indonesian speakers are syllable shape contrast item correct (SSIC) and consonant item correct (CIC), while Taiwanese speakers’ SSIC and vowel item correct (VIC) did not predict intelligibility. Only CIC predicts intelligibility. Both groups made similar errors in consonant voicing. The study's findings contribute to the teaching materials for the English preparation of students who will study abroad.  Â

Diagnostics and treatment for HER2+ mG&CRC: Learning outcomes of GetSMART.

82 Background: Gastric/colorectal cancer (G&CRC) are among the most prevalent cancers. Efforts are ongoing to better assess their clinical/genomic factors. HER2 (human epidermal growth factor receptor 2) testing is strongly recommended in clinical practice to guide treatment. An accredited education program (GetSMART) was launched online (January 2021–April 2022) to equip healthcare providers (HCPs) with knowledge, skills, and confidence to accurately assess/apply molecular testing and therapy options for metastatic G&CRC patients. Methods: A knowledge/confidence-based assessment was embedded in four modules to measure learners’ knowledge mastery (% answering correctly with confidence) pre/post-activity. Self-reported confidence and anticipated performance in clinical practice were measured before/after each module. At the end of program, HCPs were asked about their commitment to change. McNemar’s test was applied to assess statistical significance in pre-post for paired items, and descriptive analysis for non-paired items. Results: Overall, 925 unique HCPs engaged in GetSMART, 31% (n=284) completed all modules, of which 57% (n=163) consented analysis of their data. A statistically significant increase (p<0.05) was measured in percentage of HCPs demonstrating knowledge mastery, confidence, and commitment to change their practice from pre to post across four competencies, except for percentage of HCPs “often”/“always” engaging patients in shared decision-making (SDM) (3, Performance). Largest gains in knowledge mastery/confidence were noted for HCPs’ ability to anticipate/manage adverse events (AE) ( 4A); and for anticipated performance, in HCPs considering appropriate treatment. Post-activity, 66% (98/149) of HCPs intended to change practice based on module content, 83% (70/84) of whom were “highly”/“somewhat” confident to make changes. The activity’s anticipated impact on patients was rated “major”/“moderate” by 69% (103/149) of HCPs. Conclusions: Identification of HER2 aberrations in mG&CRC and its appropriate treatment selection remains challenging in oncology care. This evaluation demonstrated that GetSMART positively impacted HCPs’ knowledge, confidence, and anticipated performance in clinical practice to enhance health outcomes of HER2+ mG&CRC patients. [Table: see text]

Multilevel SEM with random slopes in discrete data using the pairwise maximum likelihood.

Pairwise maximum likelihood (PML) estimation is a promising method for multilevel models with discrete responses. Multilevel models take into account that units within a cluster tend to be more alike than units from different clusters. The pairwise likelihood is then obtained as the product of bivariate likelihoods for all within-cluster pairs of units and items. In this study, we investigate the PML estimation method with computationally intensive multilevel random intercept and random slope structural equation models (SEM) in discrete data. In pursuing this, we first reconsidered the general 'wide format' (WF) approach for SEM models and then extend the WF approach with random slopes. In a small simulation study we the determine accuracy and efficiency of the PML estimation method by varying the sample size (250, 500, 1000, 2000), response scales (two-point, four-point), and data-generating model (mediation model with three random slopes, factor model with one and two random slopes). Overall, results show that the PML estimation method is capable of estimating computationally intensive random intercept and random slopes multilevel models in the SEM framework with discrete data and many (six or more) latent variables with satisfactory accuracy and efficiency. However, the condition with 250 clusters combined with a two-point response scale shows more bias.

A large neighborhood search algorithm and lower bounds for the variable-Sized bin packing problem with conflicts

In this paper, we study the Variable-Sized Bin Packing Problem with Conflicts (VSBPPC). In VSBPPC, a set of items each with a certain size has to be packed into bins of various types. Bin types differ in terms of their capacity and cost, and certain pairs of items cannot be packed into the same bin due to conflicts. The goal is to pack the items into the bins such that the total cost of the used bins is minimized. VSBPPC generalizes both the Variable-Sized Bin Packing Problem (VSBPP) and Bin Packing Problem with Conflicts (BPPC). We propose new lower bounds and develop a large neighborhood search algorithm for the problem. In the proposed solution approach, we destroy the solution by unpacking some of the bins and then repair the solution by a greedy method considering the unit cost of packing each item followed by a local search procedure. In the local search phase, we improve the repaired solution by (i) transferring items from its current bin to another bin, and (ii) swapping the items between bins. We evaluate the performance of the proposed solution approach not only against a lower bound but also against the benchmark algorithms from the literature. The proposed solution approach outperforms the benchmark algorithms with at least a margin of 4.39% on average. Moreover, the solutions obtained by the proposed approach have an average optimality gap of 2.77% with respect to the lower bound.

The creation of the Global Scales for Early Development (GSED) for children aged 0–3 years: combining subject matter expert judgements with big data

IntroductionWith the ratification of the Sustainable Development Goals, there is an increased emphasis on early childhood development (ECD) and well-being. The WHO led Global Scales for Early Development (GSED) project...

ASEAN+3 캠프에 참여한 과학영재들의 과학적 탐구의 본성에 대한 이해수준

This study aimed to investigate what extent science gifted students understand about the nature of scientific inquiry and how their understanding levels differ between Korea and other six Asian countries and among three science majors, physics, chemistry, and biology. For this end, 101 students (60.5%) from the 167 participants at ASEAN+3 science gifted students camp in January 2021 completed the questionnaire of VOSI (Views of Scientific Inquiry; Swartz et al., 2008). The quantitative data collected from the questionnaire were coded as two levels, naive and informed. And two paired items about each view were combined and coded as naive, transitional, and informed. For the results, science gifted students in this study showed that the average ratio of the three views on the nature of scientific inquiry was 42.8% for naive level of understanding and 57.2% for informed level. Among three views with five items, the highest rate of understanding as informed was for the view, ‘Inquiry procedures can influence results.’ while the lowest rate was appeared with the view ‘There is no single scientific method.’ Science gifted students of Korea showed higher rate of informed level about the view, ‘There is no single scientific method.’ than other six Asian countries. But the rest two views were opposite as the understanding ratio of the informed level by other countries were higher than those by Korea. There were distinct differences among three science majors and between Korea and other Asian countries. It was called that further research about scientific gifted students’ understanding of the nature of scientific inquiry should be conducted.

Zero the hero: Evidence for involvement of the ventromedial prefrontal cortex in affective bias for free items

Recent evidence from psycho-economics shows that when the price of an item decreases to the extent that it becomes available for free, one can observe a remarkable increase of subjective utility toward this item. This phenomenon, which is not observed for any other price but zero, has been termed the zero-price effect (ZPE). The ZPE is attributed to an affective heuristic where the positive affect elicited by the free status of an item provides a mental shortcut biasing choice towards that item. Given that the ZPE relies on affective processing, a key role of the ventromedial prefrontal cortex (vmPFC) has been proposed, yet neuroscientific studies of the ZPE remain scarce. This study aimed to explore the role of the vmPFC in the ZPE using a novel, within-subject assessment in participants with either an acquired (lesion patients) or degenerative (behavioural-variant frontotemporal dementia patients) lesion of the vmPFC, and age-matched healthy controls. All participants were asked to make a series of choices between pairs of items that varied in price. One choice trial involved an equal decrease of both item prices, such that one of the items was priced zero. In contrast to controls, patients with both vmPFC-lesion and behavioural-variant frontotemporal dementia showed marked reductions in zero-related changes of preference in pairs of gift-cards, but not for pairs of food items. Our findings suggest that affective evaluations driving the ZPE are altered in patients with focal or degenerative damage to the vmPFC. This supports the notion of a key role of the vmPFC in the ZPE and, more generally, the importance of this region in value-based affective decision-making. Our findings also highlight the potential utility of affective heuristic tasks in future clinical assessments.

Do financial education workshops enhance parent–adolescent communication about money?

PurposeThe purpose of this study is to evaluate the effect of financial education workshops (FEWs) on parent–adolescent communication about money by controlling for a parent's gender.Design/methodology/approachThis study utilized a pre- and post-survey-based experimental research design for impact evaluation. Assuming that parents often claim that they frequently communicate with their children about money, the researcher asked children to rate their perception of their mothers' financial communication with them. Their mothers completed the pre-survey before agreeing to participate in FEWs. A follow-up survey was conducted for both study groups six months after completing the FEW series. The data consisted of 300 responses on 19 pairs of money communication items from both study groups. Neither the mothers nor the children were aware that data were collected from both the groups.FindingsThe results of the paired t-tests support the notion that financial education enhances monetary communication between mothers and adolescents.Research limitations/implicationsThis study is helpful to policymakers and financial educators not only to understand the need for “family-based financial education workshops” but also to design and implement such programs to open up the line of “money communication” between parents and children.Social implicationsThis important outcome provides a likely assumption that the enhancement in communication that had been previously constrained by factors such as a “parent's inability or unwillingness to discuss financial matters” is improved by empowering the parent on the subject of personal finance. Second, financial educators and policymakers need to understand that parents play a crucial role in the socialization process of their children. Parents' instructions and communications with their children not only impact the children's financial choices but also make them feel more competent about managing their finances. So, importance of financial socialization strategies should be included in the FEWs designed for the adults.Originality/valueExisting research studies evaluated the FEW outcomes by reporting a positive change in various financial behaviors of the participants by considering only one unit of the household. This study extends the impact evaluation of FEWs to measure the behavioral outcomes at the household level by considering two units of the household, the mother parent and adolescent child by studying their communication about money.

Session-Based Recommendations for e-Commerce with Graph-Based Data Modeling

Conventional recommendation methods such as collaborative filtering cannot be applied when long-term user models are not available. In this paper, we propose two session-based recommendation methods for anonymous browsing in a generic e-commerce framework. We represent the data using a graph where items are connected to sessions and to each other based on the order of appearance or their co-occurrence. In the first approach, called Hierarchical Sequence Probability (HSP), recommendations are produced using the probabilities of items’ appearances on certain structures in the graph. Specifically, given a current item during a session, to create a list of recommended next items, we first compute the probabilities of all possible sequential triplets ending in each candidate’s next item, then of all candidate item pairs, and finally of the proposed item. In our second method, called Recurrent Item Co-occurrence (RIC), we generate the recommendation list based on a weighted score produced by a linear recurrent mechanism using the co-occurrence probabilities between the current item and all items. We compared our approaches with three state-of-the-art Graph Neural Network (GNN) models using four session-based datasets one of which contains data collected by us from a leather apparel e-shop. In terms of recommendation effectiveness, our methods compete favorably on a number of datasets while the time to generate the graph and produce the recommendations is significantly lower.

Monitoring independence in daily life activities after trauma in humanitarian settings: Item reduction and assessment of content validity of the Activity Independence Measure-Trauma (AIM-T).

A standardized set of measures to assess functioning after trauma in humanitarian settings has been called for. The Activity Independence Measure for Trauma (AIM-T) is a clinician-rated measure of independence in 20 daily activities among patients after trauma. Designed in Afghanistan, it has since been used in other contexts. Before recommending the AIM-T for wider use, its measurement properties required confirmation. This study aims at item reduction followed by content validity assessment of the AIM-T. Using a two-step revision process, first, routinely collected data from 635 patients at five facilities managing patients after trauma in Haiti, Burundi, Yemen, and Iraq were used for item reduction. This was performed by analyzing inter-item redundancy and distribution of the first version of the AIM-T (AIM-T1) item scores, resulting in a shortened version (AIM-T2). Second, content validity of the AIM-T2 was assessed by item content validity indices (I-CVI, 0-1) based on structured interviews with 23 health care professionals and 60 patients in Haiti, Burundi, and Iraq. Through the analyses, nine pairs of redundant items (r≥0.90) were identified in the AIM-T1, leading to the removal of nine items, and resulting in AIM-T2. All remaining items were judged highly relevant, appropriate, clear, feasible and representative by most of participants (I-CVI>0.5). Ten items with I-CVI 0.5-0.85 were revised to improve their cultural relevance or appropriateness and one item was added, resulting in the AIM-T3. In conclusion, the proposed 12-item AIM-T3 is overall relevant, clear, and representative of independence in daily activity after trauma and it includes items appropriate and feasible to be observed by clinicians across different humanitarian settings. While some additional measurement properties remain to be evaluated, the present version already has the potential to serve as a routine measure to assess patients after trauma in humanitarian settings.

Fair Allocation of Indivisible Items with Conflict Graphs

We consider the fair allocation of indivisible items to several agents and add a graph theoretical perspective to this classical problem. Namely, we introduce an incompatibility relation between pairs of items described in terms of a conflict graph. Every subset of items assigned to one agent has to form an independent set in this graph. Thus, the allocation of items to the agents corresponds to a partial coloring of the conflict graph. Every agent has its own profit valuation for every item. Aiming at a fair allocation, our goal is the maximization of the lowest total profit of items allocated to any one of the agents. The resulting optimization problem contains, as special cases, both Partition and Independent Set. In our contribution we derive complexity and algorithmic results depending on the properties of the given graph. We show that the problem is strongly NP-hard for bipartite graphs and their line graphs, and solvable in pseudo-polynomial time for the classes of chordal graphs, cocomparability graphs, biconvex bipartite graphs, and graphs of bounded treewidth. Each of the pseudo-polynomial algorithms can also be turned into a fully polynomial approximation scheme (FPTAS).

Intra-list similarity and human diversity perceptions of recommendations: the details matter

The diversity of the generated item suggestions can be an important quality factor of a recommender system. In offline experiments, diversity is commonly assessed with the help of the intra-list similarity (ILS) measure, which is defined as the average pairwise similarity of the items in a list. The similarity of each pair of items is often determined based on domain-specific meta-data, e.g., movie genres. While this approach is common in the literature, it in most cases remains open if a particular implementation of the ILS measure is actually a valid proxy for the human diversity perception in a given application. With this work, we address this research gap and investigate the correlation of different ILS implementations with human perceptions in the domains of movie and recipe recommendation. We conducted several user studies involving over 500 participants. Our results indicate that the particularities of the ILS metric implementation matter. While we found that the ILS metric can be a good proxy for human perceptions, it turns out that it is important to individually validate the used ILS metric implementation for a given application. On a more general level, our work points to a certain level of oversimplification in recommender systems research when it comes to the design of computational proxies for human quality perceptions and thus calls for more research regarding the validation of the corresponding metrics.

Measurement invariance of the Broad Conceptualization of Beauty Scale (BCBS) across Black, Hispanic, and White women in the United States

The Broad Conceptualization of Beauty Scale (BCBS) assesses the degree to which women perceive diverse appearances and internal qualities as being beautiful. Although the instrument is increasingly used in diverse national and linguistic contexts, no previous study has examined measurement invariance of the BCBS across racial groups. To rectify this, we asked 395 Black, 406 Hispanic, and 423 White women from the United States to complete the BCBS. Confirmatory factor analysis indicated that a unidimensional model of BCBS scores had poor fit to the data in the total sample, but freely estimating error covariances between six pairs of items resulted in adequate fit. Additionally, full configural and scalar invariance was supported, but metric invariance was not, with further testing indicating that the item loading for one item differed across groups. Comparison of latent means indicated that all between-groups comparisons in BCBS scores were non-significant. However, medium-sized group differences in BCBS scores emerged once group differences in key demographics were controlled for. Overall, these results suggest that the BCBS largely achieves measurement invariance across Black, Hispanic, and White women in the United States, suggestive of similarity in how the construct of broad conceptualisation of beauty is understood and experienced.

Cross-cultural adaptation of discrimination and vigilance scales in ELSA-Brasil.

To describe the process of cross-cultural adaptation for the use in Brazil of the everyday discrimination scale (EDS) and the heightened vigilance scale (HVS) applied in the Longitudinal Study of Adult Health (ELSA-Brasil). Conceptual, item and semantic equivalence analyses were conducted by a group of four epidemiologists; evaluation of measurement equivalence (factorial analysis of configural, metric and scalar structures, according to sociodemographic characteristics) and reliability. A total of 11,987 participants responded to the discrimination scale, and a subsample of 260 people participated in the test-retest study. In the case of HVS, 8,916 people responded, while 149 individuals did so in the test-retest study. The scales presented conceptual, item and semantic equivalence pertinent in the Brazilian context, in addition to adequate correspondence of referential/denotative meaning of terms and also of the general/connotative of the items. The confirmatory factor analysis of EDS revealed a unidimensional structure, with residual correlations between two pairs of items, presenting configural and metric invariance among the four subgroups evaluated. Scalar invariance was identified according to sex and age group, but it was not observed for race/color and education. Heightened vigilance showed low loads and high residuals, with inadequate adjustment indicators. For the items of the discrimination scale the weighted kappa coefficient (Kp) ranged from 0.44 to 0.78, and the intraclass correlation coefficient (ICC) was 0.87. For HVS items, the Kp ranged from 0.47 to 0.59 and the ICC was 0.83. Although there are correlated items, it was concluded that the EDS is a promising scale to evaluate experiences of perceived discrimination in Brazilian daily life. However, the heightened vigilance scale did not present equivalence of measurement in the current format.

Measuring brain activity associated with subsequent relational memory in typically developing periadolescent children: preliminary data from the PRANK study

AbstractBackgroundThe developing brain undergoes significant structural and functional changes that are associated with substantial cognitive changes, and this maturational trajectory continues during the periadolescent epoch. One such change is improvement in relational memory (RM) abilities. Normal RM requires the hippocampus, as demonstrated by the distinct memory deficits of Alzheimer’s disease (AD) associated with AD‐related hippocampal pathology. Lifelong risk factors for late onset AD may affect childhood brain development in ways that affect vulnerability to AD and other memory deficits later in life. This possibility motivates investigation into the brain’s developmental trajectory and correlations with AD‐vulnerable cognitive abilities, especially memory.MethodOur ongoing NIA‐funded study, the Polygenic Risk of Alzheimer’s Disease in Nebraska Kids (PRANK) study, measures brain structure, function, and cognitive abilities in periadolescent children (age 8–13 years). Here, we report preliminary data using task‐based fMRI and a subsequent memory (SM) paradigm to measure hippocampal‐dependent RM. A periadolescent sample from the PRANK study provided data for the current project (N = 90). MRI data were collected using protocols adapted from the Human Connectome Project. Hippocampal‐dependent RM was tested using a subsequent memory task where participants were asked to study pairs of common objects while fMRI‐BOLD data were collected. A memory test was completed for the studied materials where participants were presented two types of stimuli (studied match pairs or mismatched pairs of studied items).ResultSuccessful SM was associated with increased activity consistent with the frontoparietal networks in the lateral prefrontal and lateral parietal regions, increased activity in the dorsal attention network, and decreased activity in the so‐called memory retrieval network. There was evidence of increased anterior hippocampal activity associated with successful SM.ConclusionUsing task‐based fMRI to study brains of periadolescent children while they studied pairs of items, we observed SM effects that implicated anterior hippocampus and large‐scale functional brain networks (frontoparietal, dorsal attention, and memory retrieval). Findings from this preliminary analysis are novel in this task and in this population but concordant with previous studies of older and younger individuals. Future work will test whether there is an association in the magnitude of SM effects with polygenic risk for AD among PRANK participants.

The Knapsack Problem with forfeit sets

This work introduces a novel extension of the 0/1 Knapsack Problem in which we consider the existence of so-called forfeit sets. A forfeit set is a subset of items of arbitrary cardinality, such that including a number of its elements that exceeds a predefined allowance threshold implies some penalty costs to be paid in the objective function value. A global upper bound on these allowance violations is also considered. We show that the problem generalizes both the Knapsack Problem with conflicts among item pairs and the Knapsack Problem with forfeit pairs, that have been previously introduced in the literature. We present a polynomial subcase by proving the integrality of its LP relaxation polytope and, we introduce three heuristic approaches, namely a constructive greedy, an algorithm based on the recently introduced Carousel Greedy paradigm and a hybrid Memetic/Carousel Greedy algorithm. Finally, we validate the performances for the proposed algorithms on a set of benchmark instances that consider both random and correlated data.

Dual-track feedback aggregation recommendation model for programming training

Programming online judges (POJs) are widely used to train programming skills, and exercise recommendation algorithms in POJs have attracted wide attention. The current programming recommendation algorithms cannot make full use of the feedback of user–item pairs and cannot effectively express students’ mastery of exercises. Therefore, we propose a dual-track feedback aggregation recommendation model for programming training (DTFARec). In this model, multiple types of feedback fusion mechanism (MTFFM) and dual-track method (DTM) are proposed to solve this problem and can better express students’ mastery of exercises. The MTFFM uses an attention mechanism to learn different feedback information, and the DTM is able to fuse information from both feedback and interactive aspects. The experimental results on a real-world dataset show that the model has better recommendation performance than the best performing benchmark and that our method can effectively model students’ mastery of exercises.

Characterizing Standardized Functional Data at Inpatient Rehabilitation Facilities.

The Improving Medicare Post-Acute Care Transformation Act of 2014 mandates using standardized patient functional data across post-acute settings. This study characterized similarities and differences in clinician-observed scores of self-care and transfer items for the standardized section GG functional domain and the functional independent measure (FIM) at inpatient rehabilitation facilities. We conducted secondary analyses of 2017 Uniform Data System for Medical Rehabilitation national data. Patients were assessed by clinicians on both section GG and FIM at admission and discharge. We identified 7 self-care items and 6 transfer items in section GG conceptually equivalent with FIM. Clinician-assessed scores for each pair of items were examined using score distributions, Bland-Altman plot, correlation (Pearson coefficients), and agreement (kappa and weighted kappa) analyses. In all, 408,491 patients were admitted to Uniform Data System for Medical Rehabilitation-affiliated inpatient rehabilitation facilities with one of the following impairments: stroke, brain dysfunction, neurologic condition, orthopedic disorders, and debility. Section GG and FIM. Patients were scored as more functionally independent in section GG compared with FIM, but change score distributions and score orders within impairment groups were similar. Total scores in section GG had strong positive correlations (self-care: r= 0.87 and 0.95; transfer: r= 0.82 and 0.90 at admission and discharge, respectively) with total FIM scores. Weak to moderate ranking agreements with total FIM scores were observed (self-care: kappa= 0.49 and 0.60; transfers: kappa= 0.43 and 0.52 at admission and discharge, respectively). Lower agreements were observed for less able patients at admission and for higher ability patients of their change scores. Overall, response patterns were similar in section GG and FIM across impairments. However, variations exist in score distributions and ranking agreement. Future research should examine the use of GG codes to maintain effective care, outcomes, and unbiased reimbursement across post-acute settings.

Cross-Domain Explicit–Implicit-Mixed Collaborative Filtering Neural Network

Collaborative filtering (CF) is a classical model for recommender systems. Though neural network-based CF models have been shown to perform well in some cases, they still suffer from the sparsity and cold-start issues. To address these issues, we propose a novel neural network-based CF model, termed the cross-domain explicit–implicit-mixed CF neural network (CEICFNet). The proposed model utilizes deep neural networks to learn latent factors not only from the explicit ratings and the implicit interactions but also in a cross-domain manner. In particular, domain-shared multilayer perception (MLP) networks are designed to learn the user rating latent factors and the user interaction latent factors from the explicit ratings and the implicit interactions, respectively, which act as bridges for transferring knowledge across domains. On the other hand, domain-specific MLP networks are designed to learn the item rating latent factors and the item interaction latent factors from the explicit ratings and the implicit interactions, respectively. Then, in each domain, based on the user rating latent factors and the item rating latent factors, the rating predictive representation for each user–item pair can be learned by an MLP. Similarly, the interaction predictive representation for each user–item pair can be learned. For integrating the explicit ratings and the implicit interactions, in each domain, a fully connected layer is used to automatically assign different weights to the rating predictive representations and the interaction predictive representations, based on which the final interaction probability can be generated. Extensive experiments are conducted on five datasets and the results have confirmed the effectiveness of our model.